Fine-grained decentralized server architecture

Disaggregating ICs into smaller dies with high-density interconnects addresses performance and yield issues in monolithic ICs, enhancing efficiency and customization in server architectures.

JP2024544814APending Publication Date: 2024-12-05INTEL CORP
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
JP2023579672
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2021-12-10
Filing Date
2022-11-03
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

Current monolithic ICs face performance degradation due to circuits being far apart, leading to longer delays and reduced manufacturing yields, as they are optimized equally for all circuits on a planar surface, and traditional interconnect technologies have low interconnect densities.

Method used

Disaggregate circuits into smaller IC dies connected via high-density interconnects, using silicon interposers, through-silicon vias, and embedded multi-die interconnect bridges to optimize each die for specific functions, achieving silicon-level interconnect densities with hybrid bonding.

Benefits of technology

Improves overall yield and cost by optimizing each die for its function, reducing latency and power consumption, and enabling modular server architectures with fine-grained customization and reconfigurability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024544814000001_ABST
    Figure 2024544814000001_ABST
Patent Text Reader

Abstract

A microelectronic assembly is provided having a first plurality of integrated circuit (IC) dies in a first layer, a second plurality of IC dies in a second layer between the first layer and a third layer, and a third plurality of IC dies in a third layer. In some embodiments, the second plurality of IC dies has IC dies in an array of rows and columns, each IC die of the second plurality of IC dies is coupled to two or more IC dies of the first plurality of IC dies, and the third plurality of IC dies provides electrical coupling between adjacent IC dies of the second plurality of IC dies.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present disclosure relates to techniques, methods, and apparatus directed to fine-grained, disaggregated server architectures in semiconductor integrated circuit (IC) packaging. [Background technology]

[0002] Electronic circuits are called ICs when they are commonly fabricated on a wafer of semiconductor material, such as silicon. A wafer with such ICs is typically cut into a number of individual dies. The dies may be packaged into an IC package that contains one or more dies along with other electronic components, such as resistors, capacitors, and inductors. The IC package may be integrated onto an electronics system, such as a consumer electronics system, or a server, such as a mainframe. [Brief description of the drawings]

[0003] BRIEF DESCRIPTION OF THE DRAWINGS The following detailed description, taken in conjunction with the accompanying drawings, will facilitate an understanding of the embodiments, in which like structural elements are designated with like reference numerals, and in which: FIG. [Figure 1] 1 is a simplified top view of a portion of an example microelectronic assembly, according to some embodiments of the present disclosure. [Figure 2A] 1 is a simplified top view of a portion of an example microelectronic assembly, according to some embodiments of the present disclosure. [Figure 2B] FIG. 2B is a simplified cross-sectional view of a portion of the example microelectronic assembly of FIG. 2A. [Figure 2C] FIG. 2C is a simplified cross-sectional view of a detail of a portion of the example microelectronic assembly of FIG. 2B. [Figure 3A] 3A and 3B are simplified top views of example IC die configurations in a portion of a microelectronic assembly, according to some embodiments of the present disclosure. [Figure 3B] 3A and 3B are simplified top views of example IC die configurations in a portion of a microelectronic assembly, according to some embodiments of the present disclosure. [Figure 4] 1 is a simplified top view of another example microelectronic assembly, according to some embodiments of the present disclosure. [Diagram 5] 1 is a simplified cross-sectional view of yet another example microelectronic assembly, in accordance with some embodiments of the present disclosure. [Figure 6] 5 is a simplified cross-sectional view of a portion of the example microelectronic assembly of FIG. 4, according to some embodiments of the present disclosure. [Figure 7A] 7A-7B are schematic block diagrams of different network-on-chip configurations in a microelectronic assembly, according to some embodiments of the present disclosure. [Figure 7B] 7A-7B are schematic block diagrams of different network-on-chip configurations in a microelectronic assembly, according to some embodiments of the present disclosure. [Figure 8A] 8A-8B are block diagrams of portions of a network-on-chip configuration in a microelectronic assembly according to various embodiments of the present disclosure. [Figure 8B] 8A-8B are block diagrams of portions of a network-on-chip configuration in a microelectronic assembly according to various embodiments of the present disclosure. [Figure 8C] 1 is a simplified cross-sectional view of an example network-on-chip within a portion of a microelectronic assembly, according to some embodiments of the present disclosure. [Figure 9] FIG. 1 is a block diagram of a microelectronic assembly having IC dies fabricated by two or more different processes in accordance with an embodiment of the present disclosure. [Figure 10] 1A-1C are simplified cross-sectional views of example microelectronic assemblies having IC dies fabricated by two or more different processes in accordance with embodiments of the present disclosure. [Figure 11] FIG. 2 is a simplified cross-sectional view of an example network-on-chip in a microelectronic assembly having IC dies fabricated by two or more different processes in accordance with an embodiment of the present disclosure. [Figure 12] 1 is a simplified cross-sectional view of a network-on-chip in another example microelectronic assembly having IC dies fabricated by two or more different processes according to an embodiment of the present disclosure. [Figure 13] 1 is a simplified cross-sectional view of a network-on-chip in another example microelectronic assembly having IC dies fabricated by two or more different processes according to an embodiment of the present disclosure. [Figure 14] FIG. 1 is a cross-sectional view of a device package including one or more microelectronic assemblies according to any of the embodiments disclosed herein. [Figure 15] FIG. 1 is a cross-sectional side view of a device assembly including one or more microelectronic assemblies according to any of the embodiments disclosed herein. [Figure 16] FIG. 1 is a block diagram of an example computing device including one or more microelectronic assemblies according to any of the embodiments disclosed herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0004] Overview For purposes of illustrating the IC packages described herein, it is important to understand the phenomena that may be at work during assembly and packaging of ICs. The following basic information may be viewed as a basis upon which the present disclosure may be adequately described. Such information is provided for illustrative purposes only and, therefore, should not be construed in any way as limiting the broad scope of the present disclosure and its potential applications.

[0005] Advances in semiconductor processing and logic design have allowed an increase in the amount of logic that can be included in processors and other IC devices. As a result, many processors now have multiple cores monolithically integrated on a single die. Generally, these types of monolithic ICs are also described as planar because they take the form of a flat surface and are built on a single silicon wafer that is typically made from a single crystal silicon boule. The typical manufacturing process for such monolithic ICs is called a planar process, in which photolithography, etching, thermal diffusion, oxidation, and other such processes are performed on the surface of the wafer to allow active circuit elements (e.g., transistors and diodes) to be formed on the planar surface of the silicon wafer.

[0006] Current technology allows hundreds and thousands of such active circuit elements to be formed on a single die, on which a large number of logic circuits may be enabled. In such a monolithic die, the manufacturing process must be optimized equally for all circuits, resulting in trade-offs between the different circuits. Furthermore, the constraints of having to place circuits on a planar surface mean that some circuits are far away from others, resulting in performance degradation such as longer delays. Manufacturing yields can also be severely affected, since even a malfunctioning circuit may mean that the entire die must be scrapped.

[0007] One solution to overcome these adverse effects of monolithic dies is to disaggregate circuits into smaller IC dies (e.g., chiplets, tiles) that are electrically coupled by interconnect bridges. The smaller dies become part of an assembly of interconnected dies that together form ICs that are complete in terms of application and / or function, such as memory chips, microprocessors, microcontrollers, general-purpose ICs (e.g., chips used for repetitive processing routines, simple tasks, application-specific ICs, etc.), and systems-on-chips (SOCs). In other words, the individual dies are connected together to create the function of the monolithic IC. By using separate dies, each individual die can be optimally designed and manufactured for a specific function. For example, a processor circuit, including logic circuits, may be performance targeted and therefore require a very speed-optimized layout. This has different manufacturing requirements compared to a Universal Serial Bus (USB) controller, which is built to meet a specific USB standard rather than with respect to processing speed. Therefore, by separating different parts of the overall design onto different die, each optimized for design and manufacturing, the overall yield and cost of the combined die solution may be improved.

[0008] The connections between these dies can be achieved by a number of different techniques. For example, in 2.5D packaging solutions, silicon interposers and through-substrate vias (TSVs), also called through-silicon vias when the substrate is silicon, connect the dies at silicon interconnect speeds in the smallest footprints. In another example, called an Embedded Multi-Die Interconnect Bridge (EMIB), a silicon bridge embedded under the edges of the two interconnecting dies facilitates the electrical connection between them. In three-dimensional (3D) architectures, dies are stacked one on top of the other, creating an overall smaller footprint. Typically, the electrical connections and mechanical bonds in such 3D architectures are achieved using TSVs and high-pitch solder-based bumps (e.g., C2 interconnects). EMIB and 3D stacked architectures can also be combined with omni-directional interconnects (ODIs), which allow the top-packaged chip to communicate with other chips horizontally using EMIBs and vertically using through-mold vias (TMVs), which are typically larger than TSVs. However, these current interconnect technologies use solder or its equivalent for connections and, as a result, have low interconnect densities in the vertical and horizontal directions.

[0009] One approach to alleviate low vertical interconnect density is to use interposers, which improve vertical interconnect density but suffer from low lateral interconnect density when the base wafer of the interposer is passive. In a general sense, "interposer" is commonly used to refer to a base piece of silicon that interconnects two dies. By including active circuitry in the interposer, lateral speed can be improved, but it requires a more expensive manufacturing process, especially when a larger base die is used to interconnect the smaller dies. Furthermore, not all interfaces require fine pitch connections, which can lead to additional manufacturing and processing overhead without the benefits of fine pitch. Thus, typical modular server architectures for complex and customized server configurations continue to use large monolithic dies to avoid de-aggregation overhead.

[0010] In this regard, a quasi-monolithic hierarchical integration architecture using multiple dies recursively bonded to form a microelectronic assembly helps to mitigate some of the above-mentioned shortcomings. The multiple dies can include active and / or passive dies, and at least some of the multiple dies are bonded using high density interconnects. As used herein, "high density interconnects" includes sub-10 micrometer pitch die-to-die (DTD) interconnects, also referred to as "hybrid bond," "hybrid interconnects," or "direct bond interconnects."

[0011] Some embodiments of the present disclosure provide a microelectronic assembly having a first plurality of IC dies in a first layer, a second plurality of IC dies in a second layer between the first layer and a third layer, and a third plurality of IC dies in a third layer, where the second plurality of IC dies have IC dies in an array of rows and columns, each IC die of the second plurality of IC dies is coupled to two or more IC dies of the first plurality of IC dies, and the third plurality of IC dies provides electrical coupling between adjacent IC dies of the second plurality of IC dies.

[0012] The first plurality of IC dies, in many embodiments, comprises intellectual property (IP) cores. As used herein, the terms "IP core", "IP block", "core complex", or "core" are used interchangeably to refer to circuits that comprise logic, cells, or reusable units of IC layout design with specific functionality and defined interfaces, and serve as building blocks in IC chip designs. For example, an IP core may generally comprise a set of memory registers, arithmetic logic units (ALUs), power converters, high-speed input / output (I / O) interfaces, peripherals, programmable microprocessors, microcontrollers, digital signal processors, analog-digital mixed signal processing blocks, configurable computing architectures, and the like. Because computing cores typically require additional components to create a fully functional chip or SOC, these complementary components are assumed to be inherently present in the microelectronic assemblies of the various embodiments disclosed herein, either directly coupled to the IP core in question or coupled via other IP cores or circuit blocks (e.g., portions or "blocks" of circuitry).

[0013] Some embodiments of the present disclosure also provide an IC package having a first IC die in a first layer of the IC package, a second IC die and a third IC die in a second layer of the IC package, a fourth IC die in a third layer of the IC package, and a package substrate coupled to the third layer of the IC package, where the second layer is between the first layer and the third layer, the first IC die is attached to the second IC die, the third IC die is adjacent to the second IC die, and the fourth IC die has a conductive path between the second IC die and the third IC die.

[0014] Some embodiments of the present disclosure also provide an IC die configuration (e.g., a server architecture) having a first plurality of IC dies having IC dies with IP cores, a second plurality of IC dies having IC dies, each of the second plurality of IC dies having at least one circuit conductively coupled to one or more of the IP cores, and a third plurality of IC dies having IC dies, each of the third plurality of IC dies having at least one conductive path between the second plurality of IC dies having interconnects with a pitch of less than 10 micrometers or “microns.” The first plurality of IC dies are on a first layer, the second plurality of IC dies are on a second layer, and the third plurality of IC dies are on a third layer, the second layer being between the first layer and the third layer.

[0015] The interconnects between adjacent layers may be distributed at silicon-level interconnect density. As used herein, the term "silicon-level" interconnect density when used with respect to interconnects has an interconnect density greater than 10,000 connections per square millimeter. This term refers to the trace pitch and / or via density typically found within an IC die (e.g., within the metallization stack of an IC die over an active area) as opposed to the interconnect density between IC dies of older packaging technologies (e.g., solder-based C4 or larger interconnects). Interconnects having silicon level interconnect density may have a pitch ranging between 0.5 micrometers and 10 micrometers (i.e., in one embodiment, the interconnects may have a pitch of 0.5 micrometers, in another embodiment, the interconnects may have a pitch of 2 micrometers, in some embodiments, the interconnects may have a pitch of less than 10 micrometers, in some embodiments, the interconnects may have a pitch of less than 9 micrometers, in some embodiments, the interconnects may have a pitch of less than 8 micrometers, in some embodiments, the interconnects may have a pitch ranging between 2 micrometers and 10 micrometers, in some embodiments, the interconnects may have a pitch ranging between 4 micrometers and 8 micrometers, etc.). Any range of possible values ​​between 2 micrometers and 10 micrometers may be included in the pitch of the interconnects within the broad scope of the embodiments described herein.

[0016] Some embodiments of the present disclosure also provide a microelectronic assembly having a first plurality of IC dies in a first layer, a second plurality of IC dies in a second layer, and a third plurality of IC dies in a third layer, the second layer being between the first layer and a third layer, a first portion of the first plurality of IC dies having general purpose processor circuitry and a second portion of the first plurality of IC dies having accelerator circuitry, the first portion and the second portion being physically exclusive of each other, the second plurality of IC dies having IC dies including circuitry conductively coupled to the general purpose processor circuitry and the accelerator circuitry, and the third plurality of IC dies having conductive paths providing electrical coupling between adjacent ones of the second plurality of IC dies.

[0017] Some embodiments of the present disclosure also provide an IC package having a first IC die having a general purpose processor circuit, a second IC die having an accelerator circuit, a third IC die having circuitry conductively coupled to the general purpose processor circuit, a fourth IC die having circuitry conductively coupled to the accelerator circuit, and a fifth IC die having a conductive path between the third IC die and the fourth IC die, where the first IC die and the second IC die are on a first layer, the third IC die and the fourth IC die are on a second layer, the fifth IC die is on a third layer, the second layer is between the first layer and the third layer, the first IC die is electrically and mechanically coupled to the third IC die, and the second IC die is electrically and mechanically coupled to the fourth IC die.

[0018] Some embodiments of the present disclosure also provide an IC die configuration (e.g., a server architecture) having a general purpose processor circuit in a first plurality of IC dies, an accelerator circuit in a second plurality of IC dies, support circuitry in a third plurality of IC dies conductively coupled to the general purpose processor circuit and the accelerator circuit, and a router circuit of a network-on-chip (NOC) for routing signals between the general purpose processor circuit and / or the accelerator circuit, the router circuit conductively coupled to the general purpose processor circuit and the accelerator circuit. In some embodiments, the router circuit, the conductive paths between the general purpose processor circuit and / or the accelerator circuit, and the router circuit may be, and will be, referred to as a NOC in the remainder of the specification for clarity. The router circuit is in the third plurality of IC dies, and the accelerator circuit is distributed among the general purpose processor circuits, and the conductive paths of the NOC between the general purpose processor circuits may not pass through the accelerator circuit in some embodiments. In some other embodiments, the conductive paths between the general purpose processor circuits may pass through the accelerator circuit.

[0019] Some embodiments of the present disclosure also provide a microelectronic assembly having a first plurality of IC dies in a first tier, a second plurality of IC dies in a second tier, a third plurality of IC dies in a third tier, and a NOC spanning the microelectronic assembly, the NOC having a plurality of conductively coupled router circuits, the second tier being between the first and third tier, the first plurality of IC dies having IP cores conductively coupled through the router circuits, the router circuits being of the second plurality of IC dies, and the third plurality of IC dies having conductive paths between adjacent ones of the second plurality of IC dies.

[0020] Some embodiments of the present disclosure also provide an IC package having a first IC die, a second IC die, and a third IC die in a first layer, a first IP core in the first IC die, a second IP core in the second IC die, and a third IP core in the third IC die, a fourth IC die and a fifth IC die in the second layer, a sixth IC die in the third layer, a package substrate coupled to the third layer, and a NOC coupling the first IP core, the second IP core, the third IP core, and the fourth IP core, wherein the second layer is between the first layer and the third layer, the NOC having a plurality of conductively coupled router circuits, the first IC die and the second IC die are attached to the fourth IC die, the third IC die is attached to the fifth IC die, and the sixth IC die has a conductive path between the fourth IC die and the fifth IC die.

[0021] Some embodiments of the present disclosure also provide an IC die configuration (e.g., a server architecture) having IP cores in a first plurality of IC dies in a first tier and router circuits in a second plurality of IC dies in a second tier, where the second tier is not coplanar with the first tier, and where each IP core is conductively coupled by one or more conductors over the NOC to one of the router circuits.

[0022] Some embodiments of the present disclosure also provide a microelectronic assembly having a first layer and a second layer of multiple IC dies, the first layer and the second layer being non-coplanar, the first layer and the second layer being electrically and mechanically coupled with interconnects having a pitch of less than 10 micrometers between adjacent ones of the interconnects, a first IC die of the multiple IC dies fabricated using a complementary metal-oxide-semiconductor (CMOS) process and does not have transistors with ferromagnetic material, and a second IC die of the multiple IC dies fabricated using processes and materials not typically used in CMOS fabrication (i.e., a beyond-CMOS process) and has transistors with ferromagnetic material.

[0023] Some embodiments of the present disclosure also provide an IC package having a first IC die on a first layer, a second IC die on a second layer, a third IC die on a third layer, and a package substrate coupled to the third layer, the second layer being between the first layer and the third layer, the interconnects between the first layer and the second layer having an interconnect with a pitch of less than 10 micrometers between adjacent ones of the interconnects, one of the first IC die and the second IC die having a transistor with a ferromagnetic material, and the other of the first IC die and the second IC die having a first circuit for voltage conversion, a second circuit for frequency shifting, and a third circuit for voltage regulation.

[0024] Some embodiments of the present disclosure also provide an IC die configuration (e.g., a server architecture) having a plurality of IP cores distributed across at least a first tier and a second tier, a router circuit conductively coupled to the plurality of IP cores, and a circuit distributed across an array of the IC die and conductively coupled to the plurality of IP cores and the router circuit, where the first tier and the second tier are not coplanar, at least one IP core of the plurality of IP cores is fabricated using a Beyond CMOS process and has transistors with ferromagnetic material, and the router circuit is distributed across the array of the IC die.

[0025] Each of the structures, assemblies, packages, methods, devices, and systems of the present disclosure may have several innovative aspects, no single one of which is solely responsible for all of the desirable attributes disclosed herein. Details of one or more implementations of the subject matter described in this specification are set forth in the following description and accompanying drawings.

[0026] In the following detailed description, various aspects of exemplary implementations may be described using terms commonly used by those skilled in the art to convey the substance of their work to others skilled in the art.

[0027] The terms "circuit" and "circuitry" mean one or more passive and / or active electric and / or electronic components configured to cooperate with each other to provide a desired function. These terms also refer to analog circuits, digital circuits, hardwired circuits, programmable circuits, microcontroller circuits, and / or any other type of physical hardware electric and / or electronic components.

[0028] The term "integrated circuit" means a circuit integrated into a semiconductor or similar material, such as a semiconductor die.

[0029] In some embodiments, the IC die disclosed herein may have a substantially single crystalline semiconductor, such as silicon or germanium, as a substrate on which an integrated circuit is fabricated using traditional semiconductor processing methods. The semiconductor substrate may include, for example, an N-type or P-type material. The die may include, for example, a crystalline substrate formed using bulk silicon (or other bulk semiconductor material) or a semiconductor-on-insulator (SOI, e.g., silicon-on-insulator) structure. In some other embodiments, the substrate of one or more of the IC dies may have alternative materials, which may or may not be combined with silicon, including, but not limited to, germanium, indium antimony, lead telluride, indium arsenide, indium phosphide, gallium arsenide, indium gallium arsenide, gallium antimony, or other combinations of Group III-N, III-V, II-VI, or IV materials. In yet other embodiments, the substrate may comprise a compound semiconductor having, for example, a first sublattice of at least one element from group III of the periodic table (e.g., Al, Ga, In) and a second sublattice of at least one element from group V of the periodic table (e.g., P, As, Sb). In yet other embodiments, the substrate may comprise an intrinsic group IV or III-V semiconductor material or alloy that is not intentionally doped with electrically active impurities, and in alternative embodiments, nominal impurity dopant levels may be present. In still other embodiments, the die may comprise an amorphous material, such as, for example, a polymer, and for example, the substrate may comprise a silica-filled epoxy. In other embodiments, the substrate may comprise a high mobility oxide semiconductor material, such as tin oxide, antimony oxide, indium oxide, indium tin oxide, titanium oxide, zinc oxide, indium zinc oxide, indium gallium zinc oxide (IGZO), gallium oxide, titanium oxynitride, ruthenium oxide, or tungsten oxide.In general, the substrate may comprise one or more of tin oxide, cobalt oxide, copper oxide, antimony oxide, ruthenium oxide, tungsten oxide, zinc oxide, gallium oxide, titanium oxide, indium oxide, titanium oxynitride, indium tin oxide, zinc indium oxide, nickel oxide, niobium oxide, copper peroxide, IGZO, indium telluride, molybdenite, molybdenum diselenide, tungsten diselenide, tungsten disulfide, N-type or P-type amorphous or polycrystalline silicon, germanium, indium gallium arsenide, silicon germanium, gallium nitride, aluminum gallium nitride, indium phosphide, and black phosphorus, each of which may be optionally doped with one or more of gallium, indium, aluminum, fluorine, boron, phosphorus, arsenic, nitrogen, tantalum, tungsten, magnesium, and the like.

[0030] Unless otherwise specified, an IC die as described herein includes one or more IC structures (or simply "ICs") that implement (i.e., are configured to perform) a particular function. In one such example, the term "memory die" may be used to describe a die that includes one or more ICs that implement memory circuitry (e.g., ICs that implement one or more of memory devices, memory arrays, control logic configured to control the memory devices and arrays, etc.). In another such example, the term "compute die" may be used to describe a die that includes one or more ICs that implement logic / computational circuitry (e.g., ICs that implement one or more of I / O functions, arithmetic operations, data pipelining, etc.).

[0031] In another example, the terms "package" and "IC package" are synonymous, as are the terms "die" and "IC die." Note that the terms "chip," "die," and "IC die" are used interchangeably herein.

[0032] Unless otherwise noted, the term "insulating" means "electrically insulating" and the term "conducting" means "electrically conducting." With respect to optical signals and / or devices, components, and elements that act on or use optical signals, the term "conducting" can also mean "optically conducting."

[0033] The terms "oxide", "carbide", "nitride", etc. refer to compounds that contain oxygen, carbon, nitrogen, etc., respectively.

[0034] The term "high-k dielectric" refers to a material that has a higher dielectric constant than silicon oxide, and the term "low-k dielectric" refers to a material that has a lower dielectric constant than silicon oxide.

[0035] The term "insulating material" or "insulator" (also referred to herein as "dielectric material" or "dielectric") refers to solid materials (and / or liquid materials that solidify after processing as described herein) that are substantially non-conductive. They may include, by way of example and not limitation, organic polymers and plastics, as well as inorganic materials such as ionic crystals, porcelain, glass, silicon, silicon oxide, silicon nitride, and alumina, or combinations thereof. They may include dielectric materials, high polarizability materials, and / or piezoelectric materials. They may be transparent or opaque without departing from the scope of the present disclosure. Further examples of insulating materials are underfills and mold or mold-like materials used in packaging applications, including, for example, materials used in organic interposers, package supports, and other such components.

[0036] In various embodiments, the elements associated with an IC may include, for example, transistors, diodes, power supplies, resistors, capacitors, inductors, sensors, transceivers, receivers, antennas, etc. In various embodiments, the elements associated with an IC may include elements monolithically integrated within an IC, implemented on an IC, or connected to an IC. The ICs described herein may be either analog or digital and may be used in a number of applications, for example, as microprocessors, optoelectronics, logic blocks, audio amplifiers, etc., depending on the components associated with the IC. The ICs described herein may be used in a single IC die or as part of a chipset to perform one or more related functions in a computer.

[0037] In various embodiments of the present disclosure, the transistors described herein may be field effect transistors (FETs), such as metal oxide semiconductor (MOS) FETs (MOSFETs). In general, FETs are three-terminal devices that include a source terminal, a drain terminal, and a gate terminal, and that use an electric field to control the current through the device. A FET typically includes a channel material, source and drain regions disposed in and / or on the channel material, and a gate stack including a gate electrode material, alternatively referred to as a "work function" material, disposed on a portion of the channel material between the source and drain regions (the "channel portion"), with the gate stack also optionally including a gate dielectric material between the gate electrode material and the channel material.

[0038] In a general sense, an "interconnect" refers to any element that provides a physical connection between two other elements. For example, an electrical interconnect provides an electrical connection between two electrical components to facilitate communication of electrical signals therebetween, and an optical interconnect provides an optical connection between two optical components to facilitate communication of optical signals therebetween. As used herein, both electrical and optical interconnects are included in the term "interconnect." The nature of the interconnect being described will be understood herein with reference to its associated signal medium. Thus, when used with reference to an electronic device, such as an IC, that operates with electrical signals, the term "interconnect" describes any element formed of a conductive material to provide electrical connection to one or more elements associated with the IC and / or between various such elements. In such cases, the term "interconnect" may refer to both conductive traces (which may also be referred to as "lines," "wires," "metal lines," or "trenches") and conductive vias (which may also be referred to as "vias" or "metal vias"). Sometimes, conductive traces and vias may be referred to as "conductive traces" and "conductive vias," respectively, to emphasize that these elements include a conductive material, such as, for example, a metal. Similarly, when used in reference to devices that operate on optical signals, such as, for example, photonic integrated circuits (PICs), "interconnect" may also describe any element formed of a material that is optically conductive to provide an optical connection to one or more elements associated with a PCI. In such cases, the term "interconnect" may refer to optical waveguides (e.g., structures that guide and confine optical waves), including optical fibers, optical splitters, optical combiners, optical couplers, and optical vias.

[0039] The term "conductive trace" may be used to describe conductive elements isolated by insulating materials. In IC dies, such insulating materials include interlayer low-k dielectrics disposed within the IC die. In package substrates and printed circuit boards (PCBs), such insulating materials include organic materials such as Ajinomoto Build-Up Film (ABF), polyimide, or epoxy resins. Such conductive lines are typically disposed at several levels, i.e., several layers, of a metallization stack.

[0040] The term "package substrate" may be used to describe any substrate material that facilitates packaging together a semiconductor die and / or any collection of other electrical components, such as passive electrical components. As used herein, a package substrate may be formed of materials including, but not limited to, insulating materials such as resin impregnated fiberglass (e.g., PCB or printed wiring board (PWB)), glass, ceramic, silicon, silicon carbide, etc. As used herein, a package substrate may also refer to a substrate that includes build-up layers (e.g., multiple ABF layers).

[0041] The term "conductive via" may be used to describe a conductive element that interconnects two or more conductive lines on different levels of a metallization stack. To this end, the via may be disposed substantially perpendicular to a surface of an IC die / chip or a supporting structure on which an IC structure is disposed, and may interconnect two conductive lines on adjacent levels or two conductive lines on non-adjacent levels.

[0042] The term "metallization stack" may be used to refer to a stack of one or more interconnects to provide connections to different circuit components of an IC die / chip and / or a package substrate.

[0043] As used herein, the term "pitch" of an interconnect refers to the center-to-center distance between adjacent interconnects.

[0044] In the context of a stack of dies bonded together or a die bonded to a package substrate, the term "interconnect" can also refer to a DTD interconnect and a die-to-package substrate (DTPS) interconnect, respectively. A DTD interconnect is also called a First Level Interconnect (FLI). A DTPS interconnect is also called a Second Level Interconnect (SLI).

[0045] Although not specifically shown throughout this description to avoid cluttering the drawings, when a DTD or DTPS interconnect is described, a surface of a first die can include a first set of conductive contacts and a surface of a second die or package substrate can include a second set of conductive contacts, and one or more of the conductive contacts of the first set can be electrically and mechanically coupled to a portion of the conductive contacts of the second set by the DTD or DTPS interconnect.

[0046] In some embodiments, the pitch of the DTD interconnects may be different from the pitch of the DTPS interconnects, while in other embodiments, the pitches may be substantially the same.

[0047] The DTPS interconnects disclosed herein may take any suitable form. In some embodiments, the set of DTPS interconnects may include solder (e.g., solder bumps or balls that undergo thermal reflow to form the DTPS interconnects). The DTPS interconnects including solder may include any suitable solder material, such as, for example, lead / tin, tin / bismuth, eutectic tin / silver, ternary tin / silver / copper, eutectic tin / copper, tin / nickel / copper, tin / bismuth / copper, tin / indium / copper, tin / zinc / indium / bismuth, or other alloys. In some embodiments, the set of DTPS interconnects may include an anisotropic conductive material, such as, for example, an anisotropic conductive film or an anisotropic conductive paste. The anisotropic conductive material may include a conductive material dispersed within a non-conductive material. In some embodiments, the anisotropic conductive material may include fine conductive particles embedded in a binder or thermosetting adhesive film (e.g., a thermosetting biphenyl-type epoxy resin, or an acrylic material). In some embodiments, the conductive particles may include a polymer and / or one or more metals (e.g., nickel or gold). For example, the conductive particles may include nickel-coated gold or silver-coated copper, which is then coated with a polymer. In another example, the conductive particles may include nickel. When the anisotropic conductive material is not compressed, there may be no conductive path from one side of the material to the other. However, when the anisotropic conductive material is appropriately compressed (e.g., by conductive contacts on either side of the anisotropic conductive material), the conductive materials near the compressed area may contact each other to form a conductive path from one side of the film to the other side in the compressed area.

[0048] The DTD interconnects disclosed herein may take any suitable form. In some embodiments, some or all of the DTD interconnects in the microelectronic assemblies or IC packages described herein may be metal-to-metal interconnects (e.g., copper-to-copper interconnects or plated interconnects). In such embodiments, the conductive contacts on either side of the DTD interconnect may be joined together (e.g., under elevated pressure and / or temperature) without the use of an intervening solder or anisotropic conductive material. In some metal-to-metal interconnects, a dielectric material (e.g., silicon oxide, silicon nitride, silicon carbide) may be present between the metals being joined together (e.g., between copper pads or posts that provide the associated conductive contacts). In some embodiments, one side of the DTD interconnect may include a metal pillar (e.g., a copper pillar) and the other side of the DTD interconnect may include a metal contact (e.g., a copper contact) recessed in a dielectric. In some embodiments, the intermetallic interconnect (e.g., copper-to-copper interconnect) may include a noble metal (e.g., gold) or a metal whose oxide is conductive (e.g., silver). In some embodiments, the intermetallic interconnect may include metal nanostructures (e.g., nanorods) that have a reduced melting point. Intermetallic interconnects may be capable of reliably conducting higher currents than other types of interconnects; for example, some solder interconnects may form brittle intermetallic compounds when current flows through them, and the maximum current provided through such interconnects may be constrained to mitigate mechanical failure.

[0049] In some embodiments, the die on either side of a set of DTD interconnects may be bare die.

[0050] In some embodiments, the DTD interconnects may include solder. For example, the DTD interconnects may include conductive bumps or pillars (e.g., copper bumps or pillars) attached to the respective conductive contacts by solder. In some embodiments, a thin cap of solder may be used in the intermetallic interconnects to accommodate planarity, and the solder may become intermetallic during processing. In some embodiments, the solder used in some or all of the DTD interconnects may have a higher melting point than the solder included in some or all of the DTPS interconnects. For example, if the DTD interconnects in an IC package are formed before the DTPS interconnects are formed, the solder-based DTD interconnects may use a higher temperature solder (e.g., having a melting point above 200 degrees Celsius), while the DTPS interconnects may use a lower temperature solder (e.g., having a melting point below 200 degrees Celsius). In some embodiments, the higher temperature solder may include tin, tin and gold, or tin, silver, and copper (e.g., 96.5% tin, 3% silver, and 0.5% copper), while the lower temperature solder may include tin and bismuth (e.g., eutectic tin bismuth), tin, silver, bismuth, indium, indium and tin, or gallium.

[0051] In some embodiments, the set of DTD interconnects may include anisotropic conductive materials, such as any of the materials described above for DTPS interconnects, In some embodiments, the DTD interconnects may be used as data transfer lanes, while the DTPS interconnects may be used for power and ground lines, among other things.

[0052] In the microelectronic assemblies or IC packages described herein, some or all of the DTD interconnects may have a finer pitch than the DTPS interconnects. In some embodiments, the DTPS interconnects disclosed herein may have a pitch between about 80 microns and 300 microns, while the DTD interconnects disclosed herein may have a pitch between about 0.5 microns and 100 microns, depending on the type of DTD interconnect. An example of silicon-level interconnect density is given by the density of some DTD interconnects. In some embodiments, the DTD interconnects may have a pitch that is too fine to be directly bonded to the package substrate (e.g., too fine to function as a DTPS interconnect). The DTD interconnects may have a smaller pitch than the DTPS interconnects due to greater similarity of materials between the different dies on each side of a set of DTD interconnects than between the dies on each side of the set of DTPS interconnects and the package substrate. In particular, differences in the material composition of the die and the package substrate can result in differential expansion and contraction between the die and the package substrate due to heat generated during operation (and heat applied during various manufacturing processes.) To mitigate damage caused by this differential expansion and contraction (e.g., crack formation, solder bridging, etc.), the DTPS interconnects in any of the microelectronic assemblies or IC packages described herein may be formed larger and further apart than DTD interconnects, which may experience less thermal stress due to the greater material similarity of the die pair on each side of the DTD interconnect.

[0053] It is recognized that the IC packages described herein may be provided with one or more levels of underfill (e.g., organic polymeric materials such as benzotriazole, imidazole, polyimide, or epoxy), which may not be labeled to avoid cluttering the drawings. In various embodiments, the levels of underfill may have the same or different insulating materials. In some embodiments, the levels of underfill may have a thermal curable epoxy with silicon oxide particles, and in some embodiments, the levels of underfill may have any suitable material capable of performing an underfill function, such as, for example, supporting the die and reducing thermal stress on the interconnect. In some embodiments, the selection of the underfill material may be based on design considerations, such as, for example, form factor, size, stress, operating conditions, and the like; in other embodiments, the selection of the underfill material may be based on material properties and processing conditions, such as, for example, cure temperature, glass transition temperature, viscosity, and chemical resistance, among other factors; and in some embodiments, the selection of the underfill material may be based on both design and processing considerations.

[0054] In some embodiments, the IC packages described herein may be provided with one or more levels of soldermask (e.g., epoxy liquid, liquid photopolymer, dry film photopolymer, acrylic, solvent), which may not be labeled or shown to avoid cluttering the drawings. The soldermask may be a liquid or dry film material that includes a photopolymer. In some embodiments, the soldermask may be non-photosensitive.

[0055] The terms "substantially," "close," "approximately," "approximately," and "about" generally refer to within + / - 20% of a target value (e.g., within + / - 5% or 10% of a target value) based on the context of a particular value as described herein or known in the art.

[0056] Terms referring to the orientation of various elements, such as "coplanar," "vertical," "orthogonal," "parallel," or any other angle between elements, generally refer to within + / - 5% to 20% of a target value based on the context of the particular value described herein or known in the art.

[0057] The term "connected" means a direct connection (which can be one or more of a mechanical, electrical, and / or thermal connection) between the things that are connected, without any intermediate devices, while the term "coupled" means either a direct connection between the things that are connected, or an indirect connection via one or more passive or active intermediate devices.

[0058] This description uses the phrases "in one embodiment" or "in an embodiment," each of which may refer to one or more of the same or different embodiments.

[0059] Additionally, the terms "have," "include," "have," and the like, as used with respect to the embodiments of the present disclosure, are synonymous.

[0060] The present disclosure may use perspective-based descriptions, such as "upper," "lower," "top," "bottom," and "side," and such descriptions are used for ease of description and are not intended to limit the application of the disclosed embodiments.

[0061] As used herein, the terms "above," "below," "between," and "on" refer to the relative location of one layer of material or component with respect to another layer or component. For example, a layer disposed above or below another layer may be in direct contact with the other layer or may have one or more intervening layers. Also, a layer disposed between two layers may be in direct contact with one or both of the two layers or may have one or more intervening layers. In contrast, a first layer described as being "on" a second layer refers to the layer in direct contact with the second layer. Similarly, unless expressly stated otherwise, a feature disposed between two features may be in direct contact with the adjacent feature or may have one or more intervening layers.

[0062] As used herein, the term "disposed" refers to a position, location, arrangement, and / or alignment, rather than any particular method of formation.

[0063] The term "between," when used in reference to a measurement range, includes both ends of the measurement range.

[0064] For purposes of this disclosure, the phrase "A and / or B" means (A), (B), or (A and B). For purposes of this disclosure, the phrase "A, B, and / or C" means (A), (B), (C), (A and B), (A and C), (B and C), or (A, B and C). As used herein, the notation "A / B / C" means (A), (B), and / or (C).

[0065] Although certain elements may be referred to herein in the singular, such elements may include multiple subelements. For example, a "conductive material" may include one or more conductive materials. In another example, a "dielectric material" may include one or more dielectric materials.

[0066] Unless otherwise noted, the use of ordinal adjectives "first," "second," "third," etc. to describe a common object is intended merely to indicate that different instances of a similar object are being referred to, and is not intended to imply that the objects so described must be in a given sequence, temporally, spatially, in ranking, or in any other way.

[0067] In the following detailed description, references are made to the accompanying drawings, which form a part hereof, and in which are shown by way of illustration embodiments that may be practiced. It is to be understood that other embodiments may be utilized and structural or logical changes may be made without departing from the scope of the present disclosure. Therefore, the following detailed description is not to be taken in a limiting sense.

[0068] The accompanying drawings are not necessarily drawn to scale.

[0069] In the figures, the same reference numbers refer to the same or similar elements / materials shown, and therefore, unless otherwise stated, a description of an element / material with a given reference number given in the context of one of the figures may be applicable to other figures in which the element / material with the same reference number may be shown. Also, the singular and plural forms of labels may be used with a reference number to denote single and multiple, respectively, of the same or similar type, species, or class of elements.

[0070] Also, in the drawings, schematic views of some of the example structures of the various devices and assemblies described herein may be shown with precise right angles and straight lines, but it is understood that such schematic views may not reflect real process limitations that may cause features to appear less than "ideal" when any of the structures described herein are inspected using images from a suitable characterization tool, such as, for example, a scanning electron microscope (SEM) image, a transmission electron microscope (TEM) image, or a non-contact surface profilometer. In such images of actual structures, possible processing and / or surface defects may also be visible, such as, for example, surface roughness, curvature or profile deviations, pits or scratches, imperfect straight edges of materials, tapered vias or other openings, inadvertent rounding of corners or variations in thickness of different material layers, accidental screw, edge, or combination dislocations within a crystalline region(s), and / or accidental dislocation defects of single atoms or clusters of atoms. Other defects not listed here but common in the field of device manufacturing and / or packaging may also be present.

[0071] In the drawings, specific numbers and arrangements of structures and components are presented for illustrative purposes; any desired number or arrangement of such structures and components may be present in various embodiments.

[0072] Also, unless otherwise noted, the structures shown in the figures may take any suitable form or shape depending on material properties, manufacturing processes, and operating conditions.

[0073] For convenience, where there is a collection of figures designated with different letters (e.g., FIGS. 10A-10C), such collection may also be referred to herein without those letters (e.g., as "FIG. 10"). Similarly, where there is a collection of reference numbers designated with different letters (e.g., 110a-110e), such collection may also be referred to herein without those letters (e.g., as "110").

[0074] Various operations may be described as multiple separate actions or operations, in the order that is most helpful in understanding the subject examples. However, the order of description should not be construed as to imply that the operations are necessarily order dependent. In particular, the operations may not be performed in the order presented. The operations described may be performed in a different order than in the described embodiment. Various additional operations may be performed and / or the operations described may be omitted in further embodiments.

[0075] Example embodiment 1 is a schematic top view of a portion of a microelectronic assembly 100 according to some embodiments of the present disclosure. The microelectronic assembly 100 includes multiple base dies 102 electrically coupled via one or more bridge dies 104. In various embodiments, the bridge die 104 includes a relatively small piece of silicon (e.g., smaller than the base dies 102) embedded or otherwise disposed under the edges of two adjacent base dies 102. In some embodiments, the bridge die 104 may be below the base dies 102, in some other embodiments, the bridge die 104 may be above the base dies 102, and in still other embodiments, some bridge dies 104 may be below some base dies 102 and other bridge dies 104 may be above other base dies 102. In a general sense, two or more base dies 102 may be coplanar and disposed in the same layer (e.g., tier, step, row, section, level, etc.) of a three-dimensional (3D) multi-layer package having multiple layers of dies.

[0076] The multiple base dies 102 may have circuitry that collectively functions as a high-bandwidth SOC-level coherent fabric (e.g., interconnected circuits, network) that facilitates data and control transmission across linked components. Such a silicon interconnect fabric allows bare IC dies to be directly connected to wiring on modular IC dies in the multiple base dies 102, facilitating high-density wiring between connected IC dies with the same small interconnect pitch as wiring in a single IC die. Thus, a larger number of inter-die connections are possible, and those connections can transmit data faster while using less energy. In various embodiments, the base die 102 may have IC dies with circuitry for low-power functions and subsystems for always-on and ultra-low power standby functions. The base die 102 may include interfaces for Peripheral Interconnect Express (PCIe), USB (e.g., USB3 Type-C), audio, debug, secure digital input / output (SDIO), and various other low-power I / O components. In various embodiments, the base die 102 includes support circuitry for other IC dies in the microelectronic assembly 100 .

[0077] In various embodiments, multiple base dies 102 can be arranged in a tiled architecture as an array of rows and columns, with the base dies 102 being substantially identical in shape (but not necessarily in function) to one another. In some embodiments, the base dies 102 can be substantially identical to one another in terms of circuitry and structure, and thus can be replicated in function as well. In other embodiments, one or more base dies 102 may have different circuitry and structure than some other base dies 102. In an array configuration, since the SOC fabric is composed of multiple base dies 102, rather than one monolithic die that spans the entire footprint of the microelectronic assembly 100, single base dies 102, and collections of multiple base dies 102, can be cheaper to manufacture, test, and assemble than single monolithic dies with the same combined footprint. Due to the lower costs associated with manufacturing smaller sized IC dies, the base dies 102 can also include active circuit blocks that are not necessarily provided in traditional silicon interposers. Thus, in various embodiments, the base die 102 may have an active interposer, although some base die 102 may also have a passive interposer without any active circuitry.

[0078] In some embodiments, the base die 102 may be manufactured using a manufacturing process tailored for low power applications, such as, for example, a 22 nanometer FinFET Low Power (22FFL), 14 nanometer process, or any other semiconductor manufacturing process suitable for the type of circuitry and functionality included in the base die 102. In some embodiments, the separate individual base die 102 may be manufactured using a corresponding desired process to optimize performance (e.g., for functionality such as I / O, Full Integrated Voltage Regulator (FIVR), passives, etc.) or cost, or both. For example, the small size of the base die 102 may enable manufacturing with the latest process nodes for maximum performance in terms of power, frequency, leakage, cache capacity, etc. Thus, a particular base die 102 used for caches may be manufactured using a process optimized for cache capacity, while a particular other base die 102 used for high speed routing may be manufactured using another process suitable for high speed, etc. Although an array of base dies 102 may occupy a relatively large footprint (e.g., 92 square millimeters), the individual base dies 102 can be much smaller (e.g., 10 square millimeters), allowing for higher manufacturing yields and resulting in lower manufacturing costs than monolithic large dies.

[0079] As shown in FIG. 2A, one or more IP dies 202 with reusable IP cores 204 may be combined in another layer on top of multiple base dies 102. Although only a few IP cores 204 are shown in the figure, it may be understood that every IP die 202 has a portion that is one or more such IP cores 204. The IP cores 204 and the IP dies 202 may be related in one-to-one, many-to-one, and / or one-to-many configurations in various embodiments. For example, one IP die 202 may have one IP core 204, one IP core 204 may be distributed across multiple IP dies 202, or one IP die 202 may have two or more IP cores 204. Multiple IP dies 202 may together form one or more core complexes.

[0080] In various embodiments, the microelectronic assembly 100 may embody a server architecture. When the microelectronic assembly 100 is used in a microprocessor, an example IP core 204 may include the smallest physical hardware unit (e.g., electronic circuitry) capable of executing a processing task, such as an ALU and one or two sets of supporting memory registers. A logical hardware unit capable of processing a single execution thread may include a central processing unit (CPU) as one or more IP cores 204, a single-core processor that does not support hyperthreading is the equivalent of a single CPU, a hyperthreading processor with a single core is the functional equivalent of two CPUs, and so on. The CPUs included in one or more IP cores 204 may be dedicated (e.g., graphics processing unit (GPU), accelerator) or general purpose (e.g., general purpose processor).

[0081] In various embodiments, a collection of base dies 102 may provide a resource pool shared by some or all of the IP cores 204 in the IP die 202. Beneficially, when new IP core technologies are introduced, the IP cores 204 may be replaced in the manufacturing process without requiring a redesign of the base die 102, reducing manufacturing costs and improving manufacturing flexibility and market responsiveness. The base die 102 may provide multiple generic or customizable interfaces to the IP die 202, and the IP die 202 may be coupled to an array of base dies 102 in a 3D architecture to optimize performance (e.g., bandwidth latency) among various components. Although the IP die 202 and the base die 102 appear in a many-to-one relationship in the figures, other relationships may be used within the broad scope of the embodiments of the present disclosure. For example, a single base die 102 may be attached to a single IP die 202, several base dies 102 may be attached to a single IP die 202, and so on. The high bandwidth, low latency SOC-level fabric facilitated by the array of base die 102 in the microelectronic assembly 100 does not compete for die area with the IP cores 204 provided in the IP die 202. In such an embodiment, the base die 102 may include circuitry to support the collective operation of the IP cores 204 provided in the IP die 202 ("support circuitry"). For example, the base die 102 may include data storage circuitry, controller circuitry, clock circuitry, voltage regulation circuitry, and / or I / O circuitry conductively coupled in a NOC configuration to the multiple IP cores 204 disposed in the IP die 202. In various embodiments, the support circuitry in the base die 102 may provide power, data, control commands, and the like to the IP die 202. Such a configuration may usefully and advantageously allow for the selection of one or more IP cores 204 that address specific needs or functions while still retaining a "standard" or "generic" die configuration for the base die 102.Furthermore, when new IP core technologies are introduced, the IP die 202 can be replaced in the manufacturing process without requiring a redesign of the base die 102, thereby reducing manufacturing costs and improving manufacturing flexibility and market responsiveness. In various embodiments, any one base die 102 may be relatively larger than each of the IP dies 202 directly attached to it.

[0082] Traditionally, circuits included in IP cores are designed to be used with circuits in other IP cores to build complex ICs, such as SOCs, application specific integrated circuits (ASICs), application specific standard products (ASSPs), and field programmable gate arrays (FPGAs), on a single monolithic chip. Such complex ICs are then used in products such as computers, cell phones, television desktop boxes, digital cameras, digital audio players, automobile engines and industrial process controllers, toys, smart cards, hearing aids, heart monitors, and other devices that use or process data. In an embodiment of the present disclosure, the IP cores 204 are included in a "chiplet," which is an IP die 202. In other words, the circuits represented by the IP cores 204 can be realized in hardware form as the IP die 202. Thus, a chiplet comprises a modular IC die that can be coupled to other modular IC dies, similar to the virtual IP cores in a monolithic SOC.

[0083] For example, a traditional SOC can be a monolithic IC die that incorporates a CPU and hundreds of additional IP cores on the same IC die. The design is then scaled by moving to the next processing node, which is an expensive process. In a chiplet or tile model, hundreds of IP cores are packed into smaller chiplets that can be mixed and matched and then assembled into a package to build a SOC. The chiplets can have different functions and can be fabricated in different process nodes and reused in different designs. The combination of IP cores or chiplets (used interchangeably herein to refer to a circuit or its hardware implementation in IC die form, respectively) in the IP die 202 of the microelectronic assembly 100 can, in some embodiments, have the functionality of a traditional compute die together, including compute cores, graphics, and display engines for monitors, with, for example, quad channel (16-bit) low power double data rate (LPDDR) memory, CPU cores, clocks, system agents, GPU, display engine, infrastructure processing unit (IPU), high speed I / O interfaces, etc.

[0084] In various embodiments, the array of generic base dies 102 may share a common distributed cache for suitable IP dies 202. Any one IP die 202 may be placed on the array of base dies 102 in a location that best suits its respective function (e.g., package pin location) or affinity to other IP cores 204 included in the other IP dies 202 (e.g., based on the functionality of the other IP cores 204). In other words, in some embodiments, for mounting of a particular one of the IP dies 202, a particular one of the base dies 102 may be selected based on the function of the particular one of the IP dies 202, in other embodiments, for mounting of a particular one of the IP dies 202, a particular one of the base dies 102 may be selected based on the relative position of the particular base die 102 within the array, and in yet other embodiments, for mounting of a particular one of the IP dies 202, a particular one of the base dies 102 may be selected based on the affinity (e.g., based on function) of the IP cores 204 in the particular one of the IP dies 202 to other IP cores in the other IP dies 202.

[0085] For example, the IP die 202 may have an IP core 204 with a processor circuit, and one of the base dies 102 may have a second circuit (e.g., a memory controller) while another one of the base dies 102 may have a third circuit (e.g., an I2C interface circuit). The IP core 204 may have more electrical connections with the second circuit than with the third circuit. Thus, the IP die 202 may be directly coupled to the base die 102 with the second circuit with a DTD interconnect having a pitch of less than 10 micrometers in an example embodiment.

[0086] In another example, the IP die 202 can have an IP core 204 with IO interface circuitry, and one of the base dies 102 can be placed near the periphery of the microelectronic assembly 100 or near an external interface of the microelectronic assembly 100, while another one of the base dies 102 can be placed toward the center of the microelectronic assembly 100. In an example embodiment, based on the circuitry of the IP core 204, the IP die 202 can be coupled to the base die 102 closer to the periphery with a DTD interconnect having a pitch of less than 10 micrometers, such that the electrical path from the IP core 204 to the external interface or periphery can be shorter than from the base die 102 in the center of the microelectronic assembly 100.

[0087] In yet another example, one of the IP dies 202 may have a processor circuit and another of the IP dies 202 may have a memory block. The processor circuit may be configured to operate with the memory block. The IP die 202 with the memory block may be coupled to a particular one of the base dies 102. In an example embodiment, the IP die 202 with the processor circuit may be coupled to the same IC die 102 so as to have a shorter electrical path and be closer to the memory block.

[0088] The IP die 202 may be attached to the base die 102 at one or more network connection points depending on its size and bandwidth requirements. Each IP die 202 may have different IP cores 204 in some embodiments. In other embodiments, a particular IP die 202 located on a particular base die 102 may have the same IP cores 204 that are different from the IP cores 204 in other IP dies 202 located on another base die 102. In yet other embodiments, the different IP cores 204 in separate IP dies 202 may be located anywhere on the array of the base die 102.

[0089] In current technology, to reduce the cost of building custom products for a variety of different applications, a SOC may be built to handle multiple different applications, with some IP cores in the SOC being required for some applications but not others. The excess IP cores may be disabled and unused in those applications, resulting in “dark silicon” which refers to the area of ​​the SOC that is not usable. Essentially, dark silicon refers to the amount of circuitry in an IC that cannot be powered at its nominal operating voltage. Dark silicon costs in terms of silicon area, both in terms of actual dollars spent on that area and opportunity cost. In the built-to-workload approach enabled by the embodiments of the present disclosure, IP cores that are not required for an application based on customer needs do not need to be assembled on the base die 102, as shown by the empty space 205 in the figure, thereby reducing and / or eliminating dark silicon. The empty space 205 may be utilized for other IP cores 204, or the remaining required IP cores 204 may be rearranged to reduce the overall footprint of the microelectronic assembly 100.

[0090] For example, a first IP die 202(1) may have a first IP core 204(1) having a first function. A second IP die 202(2) may have a second IP core 204(2) having a second function. A first microprocessor having a first function and a second function would include both IP dies 202(1) and 202(2), while a second microprocessor having a first function but not a second function may include IP die 202(1) without IP die 202(2), thus allowing fine-grained feature selection without significant loss in usable silicon area. Additionally, both the first and second microprocessors may use the same array of base die 102 and bridge die 104 (i.e., base die 102 may be the same between the first and second microprocessors) (i.e., bridge die 104 may be the same between the first and second microprocessors). In other words, the size, number, and circuitry of the base die 102 and bridge die 104 can be the same between the first microprocessor and the second microprocessor.

[0091] Some embodiments of the microelectronic assembly 100 may enable fine-grained de-consolidation, allowing individual IP cores 204 to be removed or replaced as needed based on customer needs and / or manufacturing constraints (among other reasons). Such fine-grained de-consolidation also allows process selection on a per-IP-core basis to optimize power-performance-area-cost per IP die. For example, an IP die 202 with an IP core 204 that is a high-power CPU may require the latest process node for very high density transistors and lowest power per square area or low voltage operation, while another IP die 202 with another IP core 204 that is a voltage regulator may perform better in a process that supports a much higher input voltage to maximize efficiency. Thus, in some embodiments, some IP dies 202 may be fabricated using an advanced process node (e.g., a 10 nm process) and have smaller transistors (e.g., transistors having a 64 nm transistor gate pitch) than transistors (e.g., transistors having a 160 nm transistor gate pitch) in other IP dies 202 fabricated using older processes (e.g., a 45 nm process). In some embodiments, a particular IP die 202 (e.g., having IP cores 204 with logic and / or memory circuits) may have transistors with thinner gate oxides, and therefore lower breakdown voltages, than transistors in other IP dies 202 (e.g., transistors used for higher voltage applications such as power routing).

[0092] This heterogeneous integration allows for the construction of entirely new SOCs from chiplets of different generations and process technologies that are not normally compatible with CMOS. By constructing a base die complex with an array of base dies 102 to route inter-base die traffic between different base dies 102 and inter-chiplet traffic between different IP dies 202 and distribute bandwidth between the connected IP cores 204, the individual IP cores 204 of the IP die 202 are not burdened with wire costs and extraneous logic complexity. This also further decouples the IP die 202 from process selection interdependencies with the base die 102.

[0093] Any suitable combination, layout, configuration, or arrangement of various IP cores 204 and corresponding IC dies, such as base die 102, bridge die 104, and IP die 202, may be used in the microelectronic assembly 100 within the broad scope of embodiments of the present disclosure. For example, multiple such microelectronic assemblies may be stacked in a single package. The microelectronic assembly 100 may have all of the functionality of a monolithic IC, such as a microprocessor, in some embodiments. In other embodiments, the microelectronic assembly 100 may form part of a larger IC (e.g., a system controller block), such as a microprocessor, a CPU, a memory device, such as a high bandwidth memory device, a logic circuit, an input / output circuit, a transceiver, such as a field programmable gate array transceiver, a gate array logic, such as a field programmable gate array logic, a power supply circuit, a III-V or III-nitride device, such as a III-nitride or III-nitride amplifier (e.g., a GaN amplifier), a peripheral interconnect express circuit, a double data rate transfer circuit, or other electronic components known in the art.

[0094] A portion of the cross section of the microelectronic assembly 100 along axis BB' is shown in FIG. 2B. The microelectronic assembly 100 is an example of a quasi-monolithic packaging architecture. The microelectronic assembly 100 may have at least three layers: a first layer 206 containing the IP die 202, a second layer 208 containing the base die 102, and a third layer 210 containing the bridge die 104. This three-layer structure may be mounted on a package substrate 212. In some embodiments, the package substrate 212 may have a PCB with multiple layers of conductive traces embedded in one or more layers of organic dielectric. For example, the package substrate 212 may have a laminate substrate with several layers of metal planes or traces interconnected with each other by microvias and / or through-hole plated vias, with input / output routing planes on the top and bottom layers and inner layers used as ground and power planes. In other embodiments, the package substrate 212 may include an organic interposer, and in yet other embodiments, the package substrate may include an inorganic interposer (e.g., made from glass, ceramic, or semiconductor materials.) In yet other embodiments, the package substrate 212 may include a composite of organic and inorganic materials, for example, with a semiconductor die embedded in an organic substrate.

[0095] Any IP die 202 may have several layers, namely, a semiconductor substrate 214 with active devices (e.g., transistors, diodes, etc.) and a metallization stack 216 with layers of dielectric (e.g., interlayer dielectric (ILD)) and metallization routing (e.g., metal layers between ILD layers and vias through the ILD). The IP die 202 may be electrically and mechanically coupled to the base die 102 using DTD interconnects 218. In various embodiments, the DTD interconnects 218 have hybrid junctions consisting of metal-to-metal and oxide-to-oxide (e.g., silicon oxide-to-silicon) junctions, enabling silicon-level interconnect densities and low-pitch interconnects, e.g., greater than 10,000 connections per square millimeter, at the interface between layers 206 and 208 (e.g., between the base die 102 and the IP die 202). As used herein, the term “interface” when used with respect to a structural connection between two components refers to the boundary, joint, or attachment surface of dissimilar materials of those components.

[0096] In an example embodiment, the pitch of the DTD interconnects 218 may be about 2 microns or less. In other embodiments, the pitch may be about 2 microns or more. In the example embodiment shown, the IP die 202 is bonded to the base die 102 in a face-to-face (FTF) configuration. In other embodiments, the IP die 202 may be bonded to the base die 102 in a face-to-back (FTB) or back-to-back (BTB) configuration with TSVs providing electrical coupling to the active devices and metallization stack 216 in the substrate 214, depending on the particular needs. In some embodiments, some IP die 202 may be bonded to some base die 102 in a FTF configuration, other IP die 202 may be bonded to other base die 102 in a BTB configuration, and still other IP die 202 may be bonded to the base die 102 in a FTB configuration.

[0097] Any base die 102 in the second tier 208 may have several layers, namely a semiconductor substrate 220 with active devices (e.g., transistors, diodes, etc.) and a metallization stack 222 with layers of dielectric (e.g., ILD) and metallization routing (e.g., metal layers between ILD layers and vias through the ILD). The base die 102 may be electrically and mechanically coupled to the IP die 202 using DTD interconnects 218. TSVs 224 that penetrate the substrate 220 of the base die 102 may provide electrical coupling between two opposing interfaces 226 (between layers 206 and 208) and 228 (between layers 208 and 210). The base die 102 may be coupled to the IP die 202 at the interface 226 and to the bridge die 104 in the third tier 210 at another interface 228 opposite the interface 226.

[0098] In many embodiments, the base die 102 may be electrically and mechanically coupled to the bridge die 104 at the interface 228 with the DTD interconnect 230. In some embodiments, the pitch of the DTD interconnect 218 may be smaller than the pitch of the DTD interconnect 230, and in other embodiments, the pitch of the DTD interconnect 218 may be the same as the pitch of the DTD interconnect 230. In various embodiments, the DTD interconnect 230 has a hybrid junction consisting of metal-to-metal and oxide-to-oxide (e.g., silicon oxide-to-silicon) junctions, enabling silicon-level interconnect density and low pitch interconnects. In the illustrated example embodiment, the base die 102 is coupled to the bridge die 104 in a FTB configuration, separating the metallization stack 222 of the base die 102 from the metallization stack 238 of the bridge die 104. In other embodiments, the base die 102 may be coupled to the bridge die 104 in a FTF or BTB configuration, depending on the particular needs. For example, the base die 102 may be inverted so that the metallization 222 is adjacent to the bridge die 104 instead of the IP die 202 .

[0099] In some embodiments, the base die 102 can be embedded in or surrounded by a dielectric 232 (e.g., silicon oxide, silicon nitride, mold compound). Through dielectric vias (TDVs) 234, also referred to as TMVs when the dielectric is a mold compound, can facilitate electrical coupling between the two sides of the second layer 208, for example, for power delivery and high speed signaling. In many embodiments, the thickness of the second layer 208 including the base die 102 can be less than 40 micrometers.

[0100] Any bridge die 104 in the third tier 210 may have an IC die with several layers, namely, a semiconductor substrate 236, a metallization stack 238 having layers of dielectric (e.g., ILD) and metallization routing (e.g., metal layers between ILD layers and vias through the ILD). In some embodiments, the bridge die 104 may not have any active devices, such as transistors, and instead simply serves as a routing medium for high speed and high density interconnects through the metallization stack 238. In some embodiments, the bridge die 104 may have active devices in the semiconductor substrate 236. The bridge die 104 may be electrically and mechanically coupled to the base die 102 at an interface 228 (between layers 208 and 210) using DTD interconnects 230, and a face 240 opposite the interface 228 may be proximate to the package substrate 212. In some embodiments, the bridge die 104 may have TSVs 242 that provide electrical coupling between the interface 228 and the opposing surface 240 .

[0101] In various embodiments, the bridge die 104 can be embedded in or surrounded by a dielectric 244 (e.g., silicon oxide, silicon nitride, mold compound). The dielectric 244 can have the same material as the dielectric 232 in some embodiments, while in other embodiments, the dielectric 244 can have a different material. A TDV 246 that penetrates the dielectric 244 can facilitate electrical coupling between the two sides of the third layer 210, for example, for power delivery and high speed signaling.

[0102] In various embodiments, layers 208 and 210 carry power to layer 206. Because layer 206 has no further layers on it, at least one surface of IP die 202 is available for cooling, e.g., for attaching heat sinks and other such heat transfer aids. Because IP cores 204 in multiple IP dies 202 may consume more power than base die 102 or bridge die 104, IP die 202 may include dedicated power connections through layers 208 and 210. Although data signals can move around within the layers, power is delivered straight from package substrate 212 through layers 210 and 208 to layer 206. At least some of TDVs 246, 234 and TSVs 242 and 224 facilitate such power connections in microelectronic assembly 100. Additionally, at least some of TDVs 246, 234 and TSVs 242 and 224 also facilitate signal (e.g., data) connections.

[0103] The third layer 210 may be coupled to the package substrate 212 using a DTPS interconnect 248. In the illustrated example embodiment, the DTPS interconnect 248 comprises a solder-based interconnect. Although not shown to avoid cluttering the drawing, one or more of the layers 206, 208, and 210 may have a redistribution layer (RDL) having a dielectric (e.g., polyimide, silicon oxide, silicon nitride) having electrical paths therein for routing or rerouting between an interconnect of a first pitch and an interconnect of a different second pitch. For example, such an RDL may facilitate routing at an interface between the second layer 208 and the third layer 210 (e.g., for routing between the TDVs 234 in the second layer 208 and the TDVs 246 in the third layer 210, or between the TDVs 234 in the second layer 208 and the bridge die 104 in the third layer 210).

[0104] In some embodiments, one or more of the base die 102, the bridge die 104, and the IP die 202 are 10 mm 2 In some other embodiments, one or more of the base die 102, bridge die 104, and IP die 202 may comprise semiconductor dies of any size. In yet other embodiments, one or more of the base die 102, bridge die 104, and IP die 202 may comprise other microelectronic assemblies, such as the microelectronic assembly 100, in a recursive (e.g., nested, hierarchical) arrangement. For example, the base die 102 may have substantially similar structure and components as the microelectronic assembly 100. In yet other embodiments, one or more of the base die 102, bridge die 104, and IP die 202 may comprise multiple semiconductor dies (e.g., IC dies) stacked together and electrically coupled with high density interconnects.

[0105] A quasi-monolithic hierarchical architecture as shown in FIG. 2B allows dies of different manufacturing technologies (e.g., technology nodes, or process nodes, or simply nodes) to be seamlessly bonded together in the microelectronic assembly 100. In a general sense, different processes often mean different circuit generations and architectures. The smaller (newer) the processing technology, the smaller the feature size, resulting in faster and more power efficient transistors. For example, the microelectronic assembly 100 may include an IP die 202 fabricated using a 10 nm process, a base die 102 fabricated using a 22 nm process, and a bridge die 104 fabricated using a 45 nm process.

[0106] In various embodiments, the selection of materials for dielectrics 232 and 244 may be appropriately based on the recursive repackaging and hierarchical combination of microelectronic assembly 100. In some embodiments, either of dielectrics 232 and 244 may comprise silicon oxide, silicon nitride, or other inorganic dielectric materials compatible with semiconductor manufacturing processes. In some other embodiments, either of dielectrics 232 and 244 may instead comprise an organic dielectric material, such as, for example, a polyimide material, a glass-reinforced epoxy matrix material, an organic material, such as, for example, a silica-filled epoxy, or a low-k or ultra-low-k dielectric (e.g., carbon-doped dielectrics, fluorine-doped dielectrics, porous dielectrics, organic polymer dielectrics, photosensitive dielectrics, and / or benzocyclobutene-based polymers).

[0107] Interconnects may also be described hierarchically, locally within a single die, intermediate between dies within a microelectronic assembly, and globally between hierarchical microelectronic assemblies. Such quasi-monolithic hierarchical integration architectures allow process optimization for each individual circuit block embodied in either the base die 102, the bridge die 104, and the IP die 202. Whereas previously such circuit blocks were incorporated into one large monolithic semiconductor die, embodiments of the present disclosure allow individual circuit blocks to be implemented in individual dies using processing techniques appropriate to their function and / or design, allowing for much better yield and manufacturability. Some embodiments of the present disclosure facilitate better reuse and reconfigurability of CPUs and other processors, providing greater granularity / customizability in process selection and interconnect routing.

[0108] This architecture is particularly useful for multi-core architectures, where multiple processing elements are formed using two levels of die (e.g., IP die 202 and base die 102) that can be combined together to form a larger computing structure. The larger computing structures may be further combined to form more processors, etc. One particular flexibility in this structure may be the ability to stack different dies one on top of the other to increase functionality. For example, memory dies may be stacked on top of each other to increase capacity. In another example, ALUs implemented on individual dies may be stacked on top of each other to increase throughput if the thermal solution can handle the increased power density of the stacked ALUs. The microelectronic assemblies described herein may help reduce costs and improve line utilization. The configurations disclosed in the various embodiments described herein may also enable interoperability with devices from other manufacturers or other accelerators.

[0109] 2C shows a detail 250 of the DTD interconnect 218 with hybrid bonds having silicon-level interconnect density. At the interface 226 between the layer 206 and the layer 208, the conductive contact 252 (e.g., belonging to the IP die 202) of the layer 206 can be bonded with the conductive contact 254 (e.g., belonging to the base die 102) of the layer 208, and similarly, the dielectric 256 (e.g., silicon oxide, silicon nitride, silicon oxynitride, etc.) in the layer 208 (e.g., belonging to the IP die 202) can be bonded with the dielectric 258 (e.g., silicon oxide, silicon nitride, silicon oxynitride, etc.) in the layer 208 (e.g., belonging to the base die 102). The bonded interconnect forms the DTD interconnect 218 with hybrid bonds, providing electrical and mechanical coupling between the layer 206 and the layer 208. It should be noted that the described structures may be applicable to any similar hybrid bond, for example, in some embodiments of DTD interconnect 230 at interface 228 between layer 208 and layer 210.

[0110] It should be noted that Figures 1 and 2A-2C are intended to show the relative placement of components within their assemblies, and that such assemblies may generally include other components not shown (e.g., various interface layers, or various other components related to optical functionality, electrical connections, or thermal mitigation). For example, in some further embodiments, the assemblies shown in Figures 1 and 2A-2B may include multiple dies and / or XPUs along with other electrical components.

[0111] Additionally, while some components of the assemblies are shown in FIGS. 1 and 2A-2B as being planar rectangular or formed from rectangular solids, this is merely for ease of illustration, and embodiments of these assemblies may be curved, rounded, or otherwise irregularly shaped as may result from the manufacturing processes used to produce the various components.

[0112] 3A and 3B are simplified block diagrams illustrating example configurations 300 and 310 of IP dies 202 on a base die 102. As shown in FIG. 3A, in some embodiments having the configuration 300, the IP dies 202 may be of any suitable shape and size suitable for the particular circuitry therein. Such differently sized and shaped IP dies 202 may be appropriately placed on the base die 102 and coupled to the base die 102 with DTD interconnects 218 (not shown) patterned accordingly. In such a configuration 300, one or more (or none) of the IP dies 202 may overhang the base die 102, i.e., extend beyond the boundaries of the base die 102. The separation between adjacent IP dies 202 may be on the order of tens of micrometers, comparable to that between two IP cores in a monolithic SOC. 3B, all IP dies 202 disposed on the base die 102 may be regularly shaped to fit neatly in an array within the boundaries of the base die 102. Such a regular arrangement may simplify routing on the NOC. In some embodiments, a server architecture with a microelectronic assembly 100 may have multiple base dies 102 and IP dies 202, generally in configuration 300, generally in configuration 310, or a mix of both.

[0113] FIG. 4 is a simplified block diagram illustrating a microelectronic assembly 100 for a processor (e.g., a cellular processor, a network processor, etc.) according to various embodiments. In the illustrated example embodiment, the microelectronic assembly 100 includes a first array of base die 102 and a second array of IP die 202 on the first array. The IP die 202 can be sized and shaped such that the second array of IP die 202 fits on the first array of base die 102. The microelectronic assembly 100 can include a NOC 402 that links various IP cores of the multiple IP die 202 of the first layer 206. A first portion of the IP die 202 (e.g., 202(1)) can include a general purpose processor circuit 404, and a second portion of the IP die 202 (e.g., 202(2)) can include an accelerator circuit 406. The base die 102 can include support circuitry to support the general purpose processor circuitry and the accelerator circuitry.

[0114] Cellular (e.g., 5G, 6G, etc.) processors and network processors (e.g., data centers, backbones, etc.) generally require very fast processing of specific workloads. They may include specific digital signal processing (DSP) functions or look-up table search functions. General purpose processors can perform many of these functions, but they are not fast, power efficient, or cost effective. Using custom application-specific ICs (e.g., ASICs) instead of general purpose processors can solve this problem, but this is generally an expensive solution and does not benefit from the economies of scale of general purpose processors. Furthermore, new ASICs require custom platforms and new software stacks that can be expensive and time consuming to develop and deploy.

[0115] Adding a monolithically integrated accelerator to a general-purpose processor can solve this problem. In a general sense, an accelerator is a separate architectural sub-structure (on the same IC die or on a different IC die) that has hardware engines for one or more specific tasks, such as offloading cryptographic functions, compression and decompression, regular expression (RegEx) processing, data storage functions, and networking operations. Accelerators are therefore configured (e.g., tuned, designed) to provide higher performance at lower cost, lower power, or less development effort than with general-purpose processors. Accelerators can have macro-architectures ranging from fixed-function dedicated chips to highly programmable engines configured for specific domain applications. Accelerators also tend to use specialized fixed-function hardware for frequent and regular calculations. Examples of accelerators include floating-point co-processors, GPUs for accelerating the rendering of vertex-based 3D models to two-dimensional (2D) display planes, and accelerators for the motion estimation step of video codecs. Accelerators are therefore added to a system to achieve greater functionality or performance.

[0116] Adding a monolithically integrated accelerator to a general-purpose processor reduces cost and time to market by benefiting from the existing ecosystem and platform for general-purpose processors. In addition, it benefits from high-speed on-die connections in the processor that enable high-speed communication between the accelerator and the processor. However, this approach is limited by the monolithic die size and requires high effort to integrate the accelerator with the general-purpose processor (e.g., tape-out and accelerator fitting within the processor die floorplan). Similarly, many accelerators can achieve very high power efficiency by targeting lower power silicon processes (e.g., by achieving high performance through high data-level parallelism), while general-purpose processors typically target higher power, high performance silicon processes. Also, any changes to the accelerator require a new tape-out of both the accelerator and the general-purpose processor.

[0117] In-package integration of accelerators with general-purpose processors is yet another option, for example through on-package fast routing or through advanced packaging technologies such as Intel's EMIB® or Foveros® technologies. This allows for a separation of accelerators and general-purpose processor dies for a more modular integration of accelerators that allows different accelerators to be connected to the same general-purpose processor or vice versa. However, this approach suffers from the higher power-limited bandwidth of traditional 2D / 2.xD on-package interconnects, which results in higher latency and slower, less power-efficient processing.

[0118] A common approach to integrating accelerators with server SOCs on-package involves using a general-purpose processor chip that is composed of several processing cores that communicate together and with other system components via a NOC. Networking / DSP operations are performed by sending data from the processing cores or on-chip memory controllers through the NOC over die-to-die connections on a bridge die, package traces, interposers, etc. The data is then received by the accelerator chiplet, processed, and sent back to the processor die. This allows for modular upgrades of each component (e.g., processor, accelerator) without the need to design and tape out new dies.

[0119] However, this approach suffers from energy efficiency and latency limitations, especially in high-core-count and / or non-consolidated processors where a significant portion of data (e.g., data that is later processed through cores or memory controllers that are very far from the accelerator) must travel a relatively long distance to reach the accelerator. This results in several problems, including: (1) the long on-die travel distance results in additional latency and more power consumption, and (2) on-die network congestion that can affect quality of service.

[0120] Some embodiments of the microelectronic assembly 100 may provide a possible solution to the above-mentioned problems by using a near-monolithic integration of custom network accelerators that may enable addressing more markets with the same platform. Moreover, it allows for mix-and-match of different accelerators and general-purpose processors with only negligible impact on performance. The microelectronic assembly 100 has a server architecture with finely de-concentrated general-purpose processors and accelerators. Such architectures are not efficient with currently available technologies because they introduce significant area, power, and latency overhead when splitting the system into several smaller chiplets. However, using the technologies described herein, including hybrid bonding with silicon-level interconnect densities exceeding 10,000 connections per square millimeter, and a quasi-monolithic architecture with at least three layers of stacking of IC dies, the microelectronic assembly 100 may have a much lower overhead that is close to that of a monolithic SOC.

[0121] In various embodiments, the NOC 402 is appropriately designed in the composite SOC architecture of the microelectronic assembly 100 so that IP core placement and other aspects of chip design are made easier and connection topology optimization is enabled. An example server architecture using the microelectronic assembly 100 may have two different types of IP cores: a general-purpose processor circuit 404 in the IP die 202(1) and an accelerator circuit 406 in the IP die 202(2). The IP die 202(2) with the accelerator circuit 406 is distributed (e.g., inter-distributed) between the IP die 202(1) with the general-purpose processor circuit 404, either as a coplanar chiplet, as a chiplet attached to the bottom surface of the IP die 202(1) similar to the bridge die 104, or in an interposer configuration, such as within the base die 102. The system-in-package embodied in the microelectronic assembly 100 may further include two parallel NOCs 402: a first NOC 402(1) for communication between the general purpose processor 404 and the accelerator 406, and a second NOC 402(2) for communication only between the general purpose processors 404.

[0122] NOC 402 may have interconnections through one or more layers 206, 208, and 210, as well as interconnections within a single layer of layers 206, 208, and 210. Accelerator circuitry 406 of IP die 202(2) and / or their associated connection infrastructure support transparent pass-through of NOC packets generated by general-purpose processor circuitry 406. This allows general-purpose processor circuitry 406 to communicate with each other for other workloads without involving accelerator circuitry 406. Note that NOC 402, including 402(1) and 402(2), is shown diagrammatically in a simplified diagram and does not correspond to a particular NOC coupling. The parallel configuration of NOC 402 may allow for more customization and performance benefits than would be possible with only a single NOC. It also decouples the general-purpose NOC design of NOC 402(2) from the accelerator-specific NOC design of NOC 402(1). However, such decoupling comes at the cost of additional (eg, replicated) routing resources to support both NOCs 402(1) and 402(2).

[0123] Although the accelerator circuitry 406 and the general-purpose processor circuitry 404 in the IP die 202(2) and IP die 202(1), respectively, may be fabricated using different silicon processes, the accelerator circuitry 406 and the general-purpose processor circuitry 404 are designed together to be tiled, which allows customization for different workloads. Although the IP dies 202(1) and 202(2) having the general-purpose processor circuitry 404 and the accelerator circuitry 406 are shown in the figures as being similarly shaped and sized, they need not be the same size. Thus, in some embodiments, the IP die 202(1) having the general-purpose processor circuitry 404 may be of a similar size and shape to the IP die 202(2) having the accelerator circuitry 406, and in other embodiments, the IP die 202(1) having the general-purpose processor circuitry 404 may not be of a similar size and shape to the IP die 202(2) having the accelerator circuitry 406.

[0124] The IP die 202, however, may be shaped and / or sized according to the size and shape of the underlying base die 102, for example, to facilitate uniform and / or regular connections to the NOC 402 through the base die 102. For example, the IP die 202 may be of a first size and the base die 102 may be of a second size, the second size being approximately such that multiple IP dies 202 can be accommodated on a single base die 102 without any overhang (i.e., without extending beyond the boundaries of the base die 102). Thus, all DTD interconnects 218 (not shown) between the IP die 202 and the base die 102 may be contained entirely within the boundaries of the base die 102 (with or without a redistribution layer (RDL) between them). Also, the base die 102(1) supporting the IP die 202(1) with the general purpose processor circuitry 404 may be sized differently from the base die 102(2) supporting the IP die 202(2) with the accelerator circuitry 406. Thus, in some embodiments, a different number of IP dies 202(1) may be attached to the base die 102(1) compared to the number of IP dies 202(2) attached to the base die 102(2).

[0125] In various embodiments, the base die 102 includes support circuitry for the general purpose processor circuitry 404 and the accelerator circuitry 406. The support circuitry for the general purpose processor circuitry 404 may include memory controllers, caches, clock circuits, etc. The support circuitry for the accelerator circuitry 406 may include dedicated memory registers, data routers, and the like. In some embodiments, the base die 102(1) attached to the IP die 202(1) with the general purpose processor circuitry 404 may include support circuitry specifically configured for the general purpose processor circuitry 404, and the base die 102(2) attached to the IP die 202(2) with the accelerator circuitry 406 may include support circuitry specifically configured for the accelerator circuitry 406. In other embodiments, the base die 102(1) and 102(2) may include support circuitry for both the general purpose processor circuitry 404 and the accelerator circuitry 406, and thus may be interchangeable.

[0126] 5 is a simplified cross-sectional view of a microelectronic assembly 500 having two layers 502 and 504. Layer 502 has IC dies 506 and 508 with general purpose processor circuitry 404 and accelerator circuitry 408, respectively. Layer 504 has a bridge die 510 that provides electrical coupling between IC dies 506 and 508. In some embodiments, bridge die 510 may have active devices, and in other embodiments, bridge die 510 may not have active devices. In some embodiments, bridge die 510 may have an interposer. Bridge die 510 may be embedded in or surrounded by dielectric 512 with TDV 514. In some embodiments, dielectric 512 may have the same material as dielectric 232 and / or 244. Some embodiments of microelectronic assembly 500 may enable modularity through a common interface and tileable chiplet size (e.g., 10 square millimeters) for IC dies 506 and 508, as well as high performance through the use of hybrid bonding connections between layers 502 and 504 and an interposer architecture to minimize die-to-die connection overhead for area, power, and latency.

[0127] The NOC 402 in the microelectronic assembly 500 may include a first NOC 402(1) for communication between the general purpose processor circuitry 404 in the IC die 506 and the accelerator circuitry 406 in the IC die 506, and a second NOC 402(2) for communication between the general purpose processor circuitry 404 in the IC die 508. The sizes of the IC dies 506 and 508 may be scaled to the number of general purpose processor circuitry 404 and accelerator circuitry 406 in each of them, respectively. The DTD interconnects 516 between the bridge die 510 and the IC dies 506 and 508 have hybrid bonds (e.g., metal-on-metal and oxide-on-oxide bonds) to reduce latency and enable silicon-level interconnect density.

[0128] 6 is a simplified cross-sectional view of a microelectronic assembly 100 having three layers: a first layer 206, a second layer 208, and a third layer 210. As shown in this example embodiment, the IP die 202(1) can have a general purpose processor circuit 404, and the IP die 202(2) can have an accelerator circuit 406. In other embodiments, the IP die 202(1) can have more than one general purpose processor circuit 404, and similarly, the IP die 202(2) can have more than one accelerator circuit 406. The NOC 402 in such an embodiment can be provided through the base die 102 (e.g., 102(1) and 102(2)), and the interconnection between adjacent base dies 102 (e.g., 102(1) and 102(2)) is through the bridge die 104. In some embodiments, the base die 102 (e.g., 102(1) or 102(2)) may also include one or more of the general purpose processor circuitry 404 and / or the accelerator circuitry 406. In some embodiments, the base die 102(1) coupled to the IP die 202(1) with the general purpose processor circuitry 404 may be substantially the same as the base die 102(2) coupled to the IP die 202(2) with the accelerator circuitry 406. In other embodiments, the base die 102(1) coupled to the IP die 202(1) with the general purpose processor circuitry 404 may differ from the base die 102(2) coupled to the IP die 202(2) with the accelerator circuitry 406, e.g., to support faster feedthrough or for more advanced routing.

[0129] 7A and 7B are simplified block diagrams of a NOC 402 according to various embodiments. As shown in FIG. 7A, the NOC 402 may couple various IP cores 204 together in one or more networks within the microelectronic assembly 100. The NOC 402 may include a first NOC 402(1) and a parallel second NOC 402(2). For ease of illustration, the NOCs 402(1) and 402(2) are shown with different shading to distinguish one from the other. Certain IP cores 204, e.g., 204(1), may be interconnected on the NOC 402(1), while certain other IP cores 204, e.g., 204(2), may be interconnected on the NOC 402(2). Although not specifically shown, some IP cores 204 may be interconnected on more than one NOC. It should be noted that while only two parallel NOCs are shown, any number of NOCs feasible for the microelectronic assembly 100 may be provided within the microelectronic assembly 100 within the broad scope of the embodiments. In some embodiments, each IP core 204 may be provided on a separate IP die 202, in other embodiments, a single IP die 202 may have two or more IP cores 204, and in yet other embodiments, a single IP core 204 may be distributed across multiple IP dies 202.

[0130] Also, while a mesh topology is shown for NOC 402, any suitable topology, and combinations of different topologies, including torus, ring, star, etc., may be encompassed within the broad scope of embodiments. For example, in some embodiments, NOC 402(1) may have a mesh topology while NOC 402(2) may have a ring topology, in other embodiments, NOC 402(1) may have a torus topology while NOC 402(2) may have a mesh topology, etc. NOCs 402(1) and 402(2) may each have router circuitry 702, e.g., 702(1) and 702(2), respectively, and links 704, e.g., 704(1) and 704(2), respectively, that provide suitable communications channels. Note that "link" as used herein refers to a communications channel and not necessarily to a physical connection, e.g., metal traces. Certain sparse links 706 may sparsely connect NOCs 402(1) and 402(2) to enable communication between NOCs 402(1) and 402(2). These few connections included in sparse links 706 are used only when data traverses between the networks, and therefore early planning of connections may be important to ensure proper utilization and bandwidth distribution. Using sparse links 706 to couple NOCs 402(1) and 402(2) instead of dense connections may act to conserve resources within microelectronic assembly 100.

[0131] In various embodiments, the IP cores 204 are coupled to the router circuit 702 by links 704, each link 704 having multiple interconnects between the IP die 202 and the base die 102, and at least two links 704 are conductively coupled together at a node. The interconnects forming the links 704 may have multiple conductors. In some embodiments, a node forms a junction of multiple conductors that are orthogonal to each other.

[0132] Although not shown to avoid cluttering the figure, the NOC 402 may have further elements, such as network interfaces at the edge of each IP core 204, and on-chip interfaces such as high-definition multimedia interface (HDMI), I2C, USB, and universal asynchronous receiver-transmitter (UART). The network interfaces packetize data (e.g., digital signals) generated by the IP cores 204 and send the packets to the router circuit 702. The router circuit 702 buffers packets from the IP cores 204 or from other connected router circuits. The network interfaces facilitate the modular architecture of the microelectronic assembly 100 and ensure seamless communication between different IP cores 204 on one or more of the IP dies 202 with associated housekeeping operations, regardless of their communication protocols.

[0133] It should be noted that the NOC topology represents a physical organization of the network architecture and does not necessarily reflect the physical layout realized on one or more IC dies. In the illustrated mesh topology, each router circuit 702 is connected to one IP core 204 and four adjacent router circuits via links 704. With the mesh topology, a huge number of IP cores 204 in multiple IP dies 202 can be organized into a regularly shaped structure, allowing for scalability and path diversity. Communication between the various IP cores 204 can be achieved using suitable algorithms to route communication packets from source nodes to destination nodes on the NOC 402. In this context, suitable routing algorithms for efficient and accurate packet routing can be used in the router circuit 702, along with packet switching and circuit switching, or a combination of both.

[0134] In traditional planar die configurations, there is a competition between routing resources between general purpose processors and accelerators and the global interconnects required to distribute high bandwidth, low latency connections across the die. By using a modular interposer approach enabled by the quasi-monolithic packaging architecture disclosed herein, resources on the base die 102 can be utilized for global high bandwidth interconnects. Thus, according to various embodiments, the router circuitry 702 and many of the links 704 can be placed on one or more of the base die 102. With loosely connected parallel networks NOCs 402(1) and 402(2), bandwidth competition between inter-die communications and connected chiplet-to-chiplet global communications can be eliminated or significantly reduced.

[0135] 7B, NOC 402 may further include (among other things) several hierarchical parallel NOCs 402(1), 402(2), and 402(3). For example, NOC 402(1) may include a global network that spans the entire microelectronic assembly 100. Multiple base dies 102 may communicate with each other over NOC 402(1). The link and router circuitry of NOC 402(1) may be implemented using multiple base dies 102 and bridge dies 104.

[0136] NOC 402(2) may have an intermediate network spanning each of the separate base dies 102. Multiple IP dies 202 coupled to corresponding base dies 102 may communicate with each other over NOC 402(2). The links and router circuitry of NOC 402(2) may be implemented using multiple base dies 102 and IP dies 202, and in embodiments where two or more IP dies 202 are coupled to the same base die 102 via a bridge die 104, the links and router circuitry of NOC 402(2) may be implemented using such bridge die 104. NOC 402(2) may be coupled to NOC 402(1) using link 706 in some embodiments.

[0137] NOC 402(3) may have an intra-die network that spans each of the separate IP die 202. In embodiments in which IP die 202 has more than one IP core 204, NOC 402(3) may enable communication between separate IP cores 204 within the same IP die 202. Link and router circuitry in NOC 402(3) may be implemented using metal traces and vias within IP die 202. NOC 402(3) may be coupled to NOC 402(2) and / or NOC 402(1) using links 706, for example. By splitting networks along die boundaries as described herein, modular parallel hierarchical networks may be provided within microelectronic assembly 100.

[0138] 8A and 8B are simplified block diagrams of a portion of a NOC configuration of microelectronic assembly 100. According to various embodiments, most of the router circuitry 702 and links 704 of NOC 402 can be implemented in base die 102, and IP die 202 can be attached to individual NOCs, e.g., 402(1) or 402(2), with appropriate connections to base die 102 based on the size and bandwidth requirements of each IP die 202. In a physical (e.g., structural) sense, NOC 402 can be implemented in microelectronic assembly 100 with nodes 802 representing the junctions of various interconnects 804, 806, and 808 as shown in FIG. 8A. For example, consider node 802(1) belonging to NOC 402(1). Node 802(1) has three different interconnect junctions: interconnect 804(1), shown with a dashed line, representing an intra-layer coupling with another node in the same NOC 402(1); interconnect 806, shown with a dotted line, representing an intra-layer coupling with another node in a different NOC 402(2); and interconnect 808(1), shown with an arrow, representing an extra-layer coupling with an IP die 202 in the same NOC 402(1). One or more such nodes and interconnects within NOC 402(1) may suitably form link 704. Interconnect 806 may form a portion of link 706.

[0139] The interconnects 804 and 806 may be realized, for example, through metal paths (e.g., metal traces and vias) in the metallization stack 222 and TSVs 224 of the base die 102. The interconnects 808 may be realized by the DTD interconnects 218, such as hybrid bonds (e.g., metal-to-metal and oxide-to-oxide bonds) between the base die 102 and the IP die 202, as well as vias and metal traces within either die, as appropriate. In some embodiments, the interconnects 804, 806, and 808 may also be realized by the DTD interconnects 230 between the base die 102 and the bridge die 104, metal traces and vias in the metallization stack 238, and the TSVs 242 of the bridge die 104.

[0140] In various embodiments, the cache 810 may be located on the base die 102 and, in some embodiments, distributed across an array of base dies 102 and accessible by one or more IP dies 202 connected thereto. In some embodiments, the cache 810 on the base die 102 may be associated with a local network, for example, connected only to the caches of the IP dies 202 directly coupled thereto, reducing the size of the network cluster (e.g., from substantially all IP dies 202 in the microelectronic assembly 100), thereby reducing the latency required for cache lookup. In some embodiments, the cache 810 may be accessible by other IP dies 202 in the first tier 206 that are not directly coupled to the base die 102 that hosts the cache 810. In some embodiments, the cache 810 may be distributed across multiple base dies 102 such that it is accessible to some or all of the IP dies 202 in the microelectronic assembly 100. In some embodiments, NOC 402 may be split into separate networks (eg, NOC 402(1), NOC 402(2), etc.) at die boundaries and associated with separate socket IDs or combined into larger clusters.

[0141] As shown in FIG. 8B, one or more IP die 202 may be coupled to an appropriate NOC using interconnects 808. For example, IP die 202(1) may be coupled to NOC 402(2) by nodes 802(2) and interconnects 808(2). In the illustrated example embodiment, four nodes 802(2) are utilized to couple to NOC 402(2), allowing for a bandwidth that is four times greater than that available at a single node. The combination of the four nodes 802(2) and the four interconnects 808(2) may represent a single link 704 connecting IP cores 204 (not shown) included in IP die 202(1) with router circuitry (not shown) within base die 102. Although nodes 802(1) belonging to parallel NOC 402(1) on base die 102 are available for interconnection under IP die 202(1), such nodes may not be used for electrical coupling in this example embodiment because a particular IP core 204 in IP die 202(1) is not connected to NOC 402(1) based on design parameters and other considerations. Although hybrid bonds may be present at such locations for mechanical strength, such bonds may be electrically isolated and / or not used for coupling to NOC 402. On the other hand, if a particular IP core 204 is coupled to both NOCs 402(1) and 402(2), such available nodes may be preferably used for coupling.

[0142] IP die 202(2) and IP die 202(3) may similarly be coupled to NOC 402(1) by nodes 802(1) and interconnects 808(1). Nodes 802(2) belonging to NOC 402(2) may remain electrically uncoupled to IP dies 202(2) and 202(3). In the illustrated example embodiment, two of nodes 802(1) are utilized in each of IP dies 202(2) and 202(3) to couple to NOC 402(1), providing twice the bandwidth available with a single node and half the bandwidth available with four nodes (e.g., as in IP die 202(1)).

[0143] 8C is a simplified cross-sectional view of an example NOC 402 according to various embodiments. In the illustrated example embodiment, the microelectronic assembly includes IP cores 204(1), 204(2) in IP die 202(1), IP core 204(3) in IP die 202(2), and IP core 204(4) in IP die 202(3). IP dies 202(1) and 202(2) are attached to base die 102(1), and IP die 202(3) is attached to base die 102(2). IP dies 202(1), 202(2), and 202(3) are in a first layer 206. Base dies 102(1) and 102(2) are in a second layer 208. A bridge die 104 in a third layer 210 couples base dies 102(1) and 102(2). The NOC 402 couples the IP cores 204(1), 204(2), 204(3), and 204(4) in various ways, for example, in three parallel sub-networks NOC 402(1), NOC 402(2), and NOC 402(3). The NOC 402(1) couples the IP cores 204(1), 204(2), 204(3), and 204(4) through the base die 102(1) and 102(2) and the bridge die 104. The NOC 402(2) couples the IP cores 204(1) and 204(2) with the IP core 204(3) through the base die 102(1). The NOC 402(3) couples the IP cores 204(1) and 204(2) in the IP die 202(1).

[0144] A node 802 in base die 102 in second tier 208 belonging to NOC 402(1) may have three different interconnect connection points: interconnect 804 representing an intra-tier coupling with other nodes in the same NOC 402(1), interconnect 806 representing an intra-tier coupling with other nodes in a different NOC 402(2), and interconnect 808 representing an extra-tier coupling with IP die 202 in the same NOC 402(1). Interconnect 804 may traverse metal traces and vias in base dies 102(1) and 102(2) and metal traces and vias in bridge die 104, along with DTD interconnects 230 between tier 208 and tier 210. Similarly, interconnect 806 may traverse metal traces and vias in base die 102(1) and 102(2) and metal traces and vias in bridge die 104, as well as DTD interconnect 230 between layer 208 and layer 210. Interconnect 806 may traverse metal traces and vias in base die 102(1), IP die 202(1), and DTD interconnects at the interface between layer 206 and layer 208. Node 802 and interconnects 804, 806, and 808 may comprise portions of links 704 (and / or sparse 706).

[0145] Router circuitry 702 may be located in layer 210, such as in base die 102(1) and / or 102(2), to form a mesh across microelectronic assembly 100. In some embodiments, even an intra-die NOC, such as NOC 402(3), may use router circuitry 702 located in layer 210, such as in base die 102(1) and 102(2). Providing such network resources in layer 210 frees up die area in IP dies 202(1)-202(4).

[0146] Although only three networks are described, various other NOC structures and topologies are possible within the broad scope of the embodiments. For example, IP core 204(1) may be coupled to IP core 204(4) in a parallel network to NOC 402(1), IP cores 204(1) and 204(3) may be coupled to NOC 402(3) in a parallel network, etc. Thus, the networks may be subdivided based on die boundaries in some embodiments, while in other embodiments sub-networks may be formed at functional boundaries, e.g., all memory cores may be coupled via one sub-network while all cores may be coupled via another sub-network, etc. Any suitable network configuration may be used within the broad scope of the embodiments.

[0147] FIG. 9 is a simplified block diagram illustrating an example embodiment of a multi-layer microelectronic assembly 100 with a hybrid assembly of multiple dies, some fabricated using traditional CMOS processes and others fabricated using “Beyond CMOS” processes. As used herein, Beyond CMOS dies include dies fabricated using non-CMOS processes. Typically, Beyond CMOS dies use spin, phase, multipolar orientation, mechanical position, polarity, orbital symmetry, flux quantum, molecular configurations, and other quantum states instead of electronic charge to define computational states. An example of a Beyond CMOS die includes a spin transistor (e.g., spin FET, spin MOSFET transistor) that behaves as a transistor while functioning as a magnetoresistance device, with a ferromagnetic source and a ferromagnetic drain acting as a spin injector and detector, respectively. Another example of a Beyond CMOS die is a spin wave device with a type of magnetic logic device that utilizes collective spin oscillations to transmit and process information.

[0148] As used herein, the term "Beyond CMOS processes" refers to any process that includes any of a wide variety of materials, including ferromagnetic, ferroelectric, piezoelectric, piezoresistive, magnetostrictive, or equivalent, such as semi-metallic ferromagnetic materials, materials with strong spin-orbit interactions, such as indium gallium arsenide (InGaAs), indium arsenide (InAs), indium antimony (InSb), correlated electron materials, such as vanadium dioxide, lead magnesium niobate-lead titanate, nickel, highly anisotropic magnetic materials, carbon nanotubes, graphene, polymer-based materials, oxides, such as strained germanium, silicon germanium, compounds of oxygen with at least one of aluminum, hafnium, tantalum, titanium, tungsten, and zirconium, conductive metals, such as titanium, platinum, nickel, tungsten, conductive oxides, such as perovskites, polymeric materials, low density carbon, such as Pr 1-x Ca x Refers to any manufacturing process used in semiconductor technology that employs "Mott materials" such as MnO3, SmNiO3, NiO, Ca2RuO4, NbO2, and AM4X8 (A=Ga, Ge; M=V, Nb, Ta; X=S, Se).

[0149] As CMOS transistor ICs scale according to Moore's Law, the power dissipated within them grows beyond their ability to remove heat. Also, the demand for computing operations in data centers grows exponentially over time. These two trends highlight the need for more energy-efficient computing chips, which can be realized through Beyond CMOS dies. The unique differences in operating voltage, power, etc. between CMOS and Beyond CMOS dies require novel packaging methods to interconnect them at fine scales to enable seamless data exchange between the two different types of devices. Currently, Beyond CMOS dies (e.g., tunnel FETs, 2D material FETs, ferroelectric FETs, spintronic, piezoelectric, magnetoelectric, etc.) have only been demonstrated in small-scale circuits within unpackaged chips. For example, currently existing packaging methods based on standard bridge die or similar architectures are not suitable for Beyond CMOS circuits, which operate at lower voltages and slower clock frequencies than CMOS circuits. In particular, slower clock frequencies lead to more parallel cores for computing and wider buses for carrying data, both of which are not fully realized with currently existing packaging methods but may be made possible with the microelectronic assembly 100 described in this disclosure.

[0150] According to various embodiments, multiple IC dies, some fabricated using Beyond CMOS and others using CMOS, may be physically, conductively, and communicatively coupled to a relatively large base die 102 that provides collective support circuitry used by multiple conductively coupled IP cores. Examples of support circuitry may include, but are not limited to, voltage regulation circuitry, input / output circuitry, data storage circuitry, clock circuitry, power delivery network circuitry, and the like. Typical heterogeneous integration techniques, such as using bridge dies and silicon interposers, have relatively low vertical interconnect density. In contrast, hybrid bonding at multiple layers in a quasi-monolithic architecture can enable much higher silicon-level interconnect density that can support the formation of on-chip high bandwidth networks through multiple different layers. This is particularly important for IC dies formed using Beyond CMOS processes.

[0151] In an example embodiment, the microelectronic assembly 100 may have one or more IP dies 202 coupled to a base die 102, where the IP dies 202 and the base die 102 are fabricated using a traditional CMOS process. In an example embodiment, the IP die 202 may have an IP core 204 (sometimes referred to as a CMOS IP core 204 in this and subsequent figures to distinguish it from an IP core fabricated using a Beyond CMOS process) that functions as a processor circuit, and another IP die 202 may have another CMOS IP core 204 that functions as a memory. Various other types of CMOS IP cores 204 may be included in the microelectronic assembly 100 within the broad scope of the embodiments. In addition, the microelectronic assembly 100 may have one or more Beyond CMOS dies 902 with Beyond CMOS IP cores 904 that include devices, circuits, etc., fabricated using processes other than CMOS. Such a Beyond CMOS die 902 may include ferroelectric and ferromagnetic materials, including, by way of example and not limitation, nanomagnets, ferroelectric capacitors, and majority gates.

[0152] In some embodiments, the microelectronic assembly 100 may also include one or more super-low voltage CMOS (SLVC) dies 906 (also called advanced-very-low-voltage-CMOS (AVC) or advanced-ultra-low-voltage-CMOS (AUC)) coupled to the base die 102. Other embodiments of the microelectronic assembly 100 may not include any SLVC dies 906. The SLVC dies 906 may also be stacked on top of one another in some embodiments.

[0153] Because a characteristic feature of Beyond CMOS dies (and SLVC dies) is their lower operating voltage (e.g., 0.2V or 0.35V) compared to traditional CMOS devices (e.g., 0.6V, 0.8V), a hybrid system such as that shown in this example embodiment having both CMOS and Beyond CMOS dies may use conversion circuitry between these two different domains. For example, the base die 102 may have a conversion circuit 908 that functions as a level shifter (LS) (e.g., a circuit that converts a signal from one logic level or voltage domain to another, similar to a transformer in an analog circuit) and converts a signal from 0.6V to 0.2V. In some embodiments, the conversion circuit 908 for voltage conversion has a LS that converts a signal of a first voltage of 0.6V or 0.8V from the CMOS IP die 202 to a signal of a second voltage of 0.2V for the Beyond CMOS die 902. Similarly, the LS may convert a signal of a second voltage of 0.2V from the Beyond CMOS die 902 to a signal of a first voltage of 0.6V or 0.8V for the CMOS IP die 202. In some embodiments, the LS may convert a signal of a first voltage of 0.6V from the CMOS IP die 202 to a signal of a second voltage of 0.35V for the SLVC die 906. Similarly, the LS may convert a signal of a second voltage of 0.35V from the SLVC die 906 to a signal of a first voltage of 0.6V or 0.8V for the CMOS IP die 202. Furthermore, the conversion circuit 908 may also include a serializer / deserializer (SerDes) that enables conversion from a higher clock frequency of the CMOS circuit to a lower clock frequency of the Beyond CMOS circuit (e.g., for distributed data processing). The converted signal may then be transferred to the Beyond CMOS die 902. Another circuit block 910 having a voltage regulator may regulate low-voltage power to the Beyond CMOS die 902 and / or the SLVC die 906.

[0154] Note that while the conversion circuits 908 and 910 are shown as circuits within the base die 102, they may also be implemented in separate IC die coupled to the base die 102, or within the IP die 202, the Beyond CMOS die 902, and / or the SLVC die 906, as appropriate. For example, they may be incorporated into an IC die in an additional layer between the IP die 202 and the base die 102, with these layers coupled with interconnects having silicon-level interconnect density.

[0155] A set of interconnects 912 can facilitate signals between components, including between the two Beyond CMOS dies 902. In various embodiments, the interconnects 912 can be hybrid bonds (e.g., metal-to-metal and oxide-to-oxide bonds) or similar pitch and silicon level interconnect density (e.g., 10,000-50,000 connections / mm) that facilitate more connections for voltage step-down (or step-up as needed) and the lower clock frequency of the Beyond CMOS die 902. 2 ) may have other types of bonds. In some embodiments, the interconnect 912 between the CMOS die and the Beyond CMOS die may be realized in a separate bridge die 104 dedicated to providing a high bandwidth interconnect.

[0156] Also, because the Beyond CMOS die 902 operates at lower voltages and lower power, the heat generated therein can be lower than that of CMOS devices, allowing for vertical stacking of devices in a true 3D stack. Thus, multiple Beyond CMOS dies 902 (and / or SLVC dies 906) may be stacked on top of each other, each interconnected with interconnects 912 having hybrid bonds or equivalent with silicon-level interconnect density. The microelectronic assembly 100 described herein allows for the integration of CMOS and Beyond CMOS dies with high-density interconnects, allowing for the accommodation of different voltages, different clock frequencies, and bus widths (e.g., number of communication channels) between the CMOS die (e.g., 202) and the Beyond CMOS die (e.g., 902).

[0157] In particular, wide bus widths can be facilitated by interconnects 912 with silicon-level interconnect densities (e.g., over 10,000 connections per square millimeter). For example, the conversion circuit 908 includes a SerDes that changes the frequency of the signal between a first frequency and a lower second frequency, and a LS that changes the voltage of the signal between a first voltage and a lower second voltage. The CMOS IP die 202 operates with signals at a first frequency and a first voltage, and the Beyond CMOS die 902 operates with signals at a second frequency and a second voltage. Assuming that signals at a first frequency and a first voltage traverse a first link provided by a first number of interconnects (e.g., providing a required bus width), signals at a second frequency and a second voltage will necessarily need to traverse a wider bus (e.g., due to higher current), and the second link is essentially provided by a second number of interconnects that is greater than the first number in proportion to the difference in scale between the first and second frequencies. In traditional packages with low interconnect density, it may be infeasible to support a second number of interconnects on a conventionally sized die, but in embodiments of the present disclosure, the second number is facilitated by interconnect 912 with silicon-level interconnect density.

[0158] 10 is a simplified cross-sectional view of a microelectronic assembly 100 having CMOS and Beyond CMOS circuits according to an embodiment of the present disclosure. In various embodiments, the microelectronic assembly 100 may embody a server architecture. The microelectronic assembly 100 has at least three layers 206, 208, and 210, and in some embodiments, another layer 920. A package substrate 212 may be electrically and mechanically coupled to the layer 210. The IP die 202 and the Beyond CMOS die 902 may be included in the layer 206 in the illustrated example embodiment. In some embodiments, the layer 206 may have multiple layers of Beyond CMOS die 902 stacked on top of each other. The layer 920 may have a bridge die 104 that provides an interconnect 912 to the Beyond CMOS die 902 in some embodiments. In other embodiments, layer 920 may include IC dies with conversion circuits 908 and 910 for voltage conversion, frequency shifting, and voltage regulation between the CMOS die and the Beyond CMOS die (and / or SLVC die). In some embodiments, layer 920 may not be present and the IP die 202 and the Beyond CMOS die 902 may directly connect to the base die 102 in layer 208. In some embodiments, layer 920 may be between layers 208 and 210. In other embodiments, additional layers with other IC dies may also be provided. Layer 208 may include multiple base dies 102 modularized into an array as described with reference to FIG. 1 and other figures of this disclosure. Layer 210 may include multiple bridge dies 104 coupling any base die 102 to other adjacent base dies 102. Layer 210 may be coupled to a package substrate 212, PCB, or other components as desired and based on specific needs.

[0159] The packaging scheme as shown has the following features: CMOS IP cores 204 (e.g., CMOS computing circuits) and Beyond CMOS IP cores 204 in the IP die 202 and Beyond CMOS die 902, respectively, in layers 206. A modular system of IP cores 904 (e.g., Beyond CMOS computing circuits); translation (e.g., voltage and frequency) through interconnects 912 provided by bridge die 104 in layer 920 between layer 206 (e.g., having IP die 202 and Beyond CMOS die 902) and layer 208 (e.g., having base die 102) for modularity and for interchangeably using CMOS and Beyond CMOS circuits; integrated translation functions (e.g., level shifters, SerDes, voltage regulators, etc.) inside the base die 102; interconnects 912 realized through hybrid bonds between bridge die 104 in layer 920 and IC dies in layer 206 (e.g., IP die 202 and Beyond CMOS die 902) or directly with the base die 102 (e.g., in the absence of layer 920). Some embodiments of the microelectronic assembly 100 may facilitate a variety of SOCs, such as, for example, a CMOS core and a Beyond CMOS accelerator, a Beyond CMOS processor and CMOS I / O, and Beyond CMOS and CMOS processors together with Beyond CMOS and CMOS memory registers.

[0160] 11 is a simplified cross-sectional view of a portion of a microelectronic assembly 100 having a CMOS die and a Beyond CMOS die within a NOC 402, according to various embodiments. In a hybrid system as shown in the figure, the microelectronic assembly 100 may have one or more IC dies (e.g., IP die 202, base die 102, bridge die 104) fabricated using a CMOS process and one or more Beyond CMOS dies 902 fabricated using a non-CMOS process. In such a 3D configuration, a large sized NOC 402 may be provided within the microelectronic assembly 100 having a smaller sized footprint than would be possible in a traditional 2D configuration.

[0161] The NOC 402 may have a 3D network spanning intra-layer links and router circuits as well as extra-layer links and router circuits, as will be further described. Whereas traditional solutions place IC dies on a two-dimensional circuit board, the systems and methods described herein stack IC dies in 3D space to reduce footprint, improve communication speed, and reduce power consumption. More specifically, the systems and methods disclosed herein place each IP core on a relatively small IC die.

[0162] Beyond CMOS dies can provide an energy-efficient alternative to heat-generating CMOS transistor ICs, the number of which scales according to Moore's Law and the number of computing operations required by today's data centers. Also, the demand for computing operations in data centers grows exponentially over time. As mentioned with respect to the preceding figures, novel packaging methods are needed to combine CMOS and Beyond CMOS circuits together in a system and to allow them to work together while seamlessly exchanging information. Beyond CMOS circuits are designed to operate at lower supply voltages than CMOS circuits. Additionally, Beyond CMOS logic can enable 3D stacking of dies.

[0163] The NOC 402 of the microelectronic assembly 100 described herein has CMOS and Beyond CMOS dies in different sub-networks utilizing a quasi-monolithic packaging architecture with hybrid bonding (or similar high density interconnect) and multiple layers of IC dies. The NOC 402 can allow multiple CMOS IP cores 204 to operate in parallel with small interconnect lengths between them, for example, to compensate for the slower clock speed in the Beyond CMOS IP cores 904. Furthermore, the NOC 402 can utilize the much lower power dissipation per unit area in the Beyond CMOS die 902, which makes heat removal via hybrid bonding sufficient for reliable operation. The NOC 402 can accommodate different configurations of multiple interconnected sub-networks to accommodate local and global workloads within each sub-network (also referred to as a "cluster"). Beyond CMOS dies 902 can be stacked in numbers significantly greater than one, thereby dramatically increasing the number of devices per unit area (following Moore's Law) and operations per second (TOPS) across the microelectronic assembly 100. In an example embodiment, the NOC 402 has a 3D mesh network configuration that takes advantage of the high interconnect density available between IC dies, both horizontally and vertically, in a quasi-monolithic packaging architecture.

[0164] Some embodiments of the NOC 402 allow for different arrangements of the die stacks and varying numbers of Beyond CMOS dies 902 in each stack. Multiple CMOS IP cores 204 may connect to the same base die 102. The functionality of the base die 102 may be provided by the Beyond CMOS die 902 in some embodiments. For example, Beyond CMOS dies 902(1) and 902(2) may function together as the base die 102(1). In such an embodiment, each IP core 204 has access to router circuitry provided by multiple stacked Beyond CMOS dies 902 to realize different NOC structures (e.g., assume that each base die in the die stack has one router circuit of the corresponding NOC accessible to each IP core 204). In some embodiments, TSVs may be provided in the Beyond CMOS die 902 and the CMOS base die 102 (e.g., 102(2)) to provide out-of-layer connections. Some of the Beyond CMOS die stacks may include caches (e.g., a distributed last level cache (LLC)) or IP cores 904 (e.g., 904(1)). In some embodiments, the NOC 402 may include hierarchical parallel sub-networks, for example, when the NOC needs to be adjusted to support connections.

[0165] In various embodiments, router circuitry 702 may be provided within base die 102 (e.g., as described with reference to FIG. 7 ) and / or within other IP die 202 with IP cores 204 (and / or Beyond CMOS die 902 with Beyond CMOS IP cores 904) within microelectronic assembly 100. In various embodiments, links 704 within NOC 402 that couple a Beyond CMOS die (e.g., 902(1)) with a CMOS IP die (e.g., 202(1)) are fabricated using a non-Beyond CMOS process and include circuitry (e.g., 908, 910, not shown) that facilitates voltage regulation (e.g., VR), voltage level shifting (e.g., LS), and frequency shifting (or variation) (e.g., SerDes). In some embodiments, the circuitry can be located within a CMOS IP die (e.g., 202(1)) that has a connected router circuit (e.g., 702(1)), while in other embodiments, the circuitry can be located within a separate CMOS IC die (e.g., bridge die 104(1)) that has interconnections or bridging between dissimilar IC dies (e.g., 202(1) and 902(1)).

[0166] In one example, IP die 202(1) may include IP cores 204(1) (e.g., compute cores) and 204(2) (e.g., memory) and router circuits 702(1) and 702(2). An intra-die NOC 402(1) may facilitate communication between IP cores 204(1) and 204(2), for example, through router circuit 702(2). An example Beyond CMOS die 902(1) may include IP cores 904(1) and router circuit 702(3). In some embodiments, IP cores 904(1) may belong to the same NOC 402(1) as IP cores 204(1). IP cores 204(1) in IP die 202(1) may communicate with IP cores 904(1) in Beyond CMOS die 902(1) through router circuit 702(1) in IP die 202(1). The router circuitry 702(1) in such an embodiment may include or be coupled to appropriate circuit blocks, such as voltage regulators, level shifters, and SerDes, to convert signals between the IP die 202(1) and the Beyond CMOS die 902(1). In another example, the IP die 202(2) may not include any router circuitry 702, instead using the router circuitry 702(4) provided in the base die 102(2).

[0167] Although only a few NOC structures are described by way of example, it can be understood that many other types of NOC structures are possible using combinations of CMOS and Beyond CMOS dies according to various embodiments of the present disclosure. For example, in some embodiments, the stacked Beyond CMOS die 902 may not function as the base die 102, and in other embodiments, some of the Beyond CMOS dies 902 in the stack may function as the base die 102 while some others may not function as the base die 102. In some embodiments, all Beyond CMOS IP cores 904 may be part of a sub-network while all CMOS IP cores 204 may be part of another sub-network, and in other embodiments, any one sub-network may span both the Beyond CMOS IP cores 904 and the CMOS IP cores 204.

[0168] 12 is a simplified cross-sectional view of a portion of microelectronic assembly 100 having a CMOS die and a Beyond CMOS die, according to various embodiments. NOC 402 may be implemented with a structure as shown in this example embodiment. Beyond CMOS die 902(1), Beyond CMOS die 902(2), and Beyond CMOS die 902(3) may be stacked on top of each other. Due to the low power consumption of the Beyond CMOS die, Beyond CMOS die 902(1), Beyond CMOS die 902(2), and Beyond CMOS die 902(3) can be stacked without encountering reliability issues due to heat dissipation. Beyond CMOS die 902(1), Beyond CMOS die 902(2), and Beyond CMOS die 902(3) may each have one or more Beyond CMOS IP cores 904 therein, such as 904(1), 904(2), 904(3), etc. Beyond CMOS die 902(1), Beyond CMOS die 902(2), and Beyond CMOS die 902(3) may also have router circuitry 702 for intra-die and inter-die routing, as desired. In some embodiments, more than one NOC may be provided in microelectronic assembly 100.

[0169] The base die 102 can act as a bridge, enabling coupling of the Beyond CMOS die 902(3) with a CMOS IP die 202 fabricated using a traditional CMOS process. The base die 102 can include appropriate circuit blocks, such as level shifters, SerDes, voltage regulators, etc., that enable signals to be seamlessly communicated between the Beyond CMOS die 902(1), the Beyond CMOS die 902(2), and the Beyond CMOS die 902(3) and the IP die 202.

[0170] In some embodiments, the Beyond CMOS die 902, such as Beyond CMOS die 902(1), Beyond CMOS die 902(2), and Beyond CMOS die 902(3), may be provided in one NOC (e.g., 402(1)), while other dies, such as CMOS IP die 202, may be provided in another parallel NOC (e.g., 402(2)). In other embodiments, the Beyond CMOS die 902, such as Beyond CMOS die 902(1), Beyond CMOS die 902(2), and Beyond CMOS die 902(3), and other dies, such as IP die 202, may be provided in the same NOC 402. Any suitable network topology may be used in microelectronic assembly 100 within the broad scope of the embodiments. The availability of hybrid interconnects (e.g., metal-to-metal and oxide-to-oxide bonds) or equivalent can enable coupling of the Beyond CMOS die 902 with other dies, such as the CMOS IP die 202, within a single NOC 402 in a compact, three-dimensional stacked configuration.

[0171] 13 is a simplified cross-sectional view of a portion of an example microelectronic package 100 having a CMOS die and a Beyond CMOS die, according to various embodiments. Various IC dies, such as 102, 104, may be embedded in a dielectric 1300 that spans two or more layers, such as layers 206, 208, 210, and 920. The dielectric 1300 may have the same material as the dielectric 232 or, in some embodiments, the dielectric 244. In other embodiments, the dielectric 1300 may have a different material than the dielectric 232 or 244. In yet other embodiments, each layer 206, 208, 210, and 920 may have a different material in the dielectric 1300. The layer 920 may have a bridge die 104 similar to the bridge die 104 in the layer 210. In the illustrated example embodiment, the IP core 204 may be distributed across the IP die 202(1) and the Beyond CMOS die 902 with the stack of Beyond CMOS dies. In one example embodiment, IP die 202(1) can include a processor circuit and Beyond CMOS die 902 can include a stack of mid-level caches (MLCs). IP die 202(2) can include another processor circuit and IP die 202(3) can include a controller.

[0172] In the illustrated example embodiment, the IP die 202(1) may be electrically coupled to the Beyond CMOS die 902 by an electrical path 1302. The electrical path 1302 may not reach the base die 102, but may instead be routed through the bridge die 104(1) in layer 920. Routing through the bridge die 104(1) that does not reach the base die 102 may facilitate a shorter path between portions of the IP cores 204 distributed across the IP die 202(1) and the Beyond CMOS die 902. Thus, in the illustrated example embodiment, the bridge die 104 in layer 920 may function as an electrical bridge across adjacent IP dies (e.g., 202, 902) located on a single base die 102 that have portions of the same IP cores 204.

[0173] The IP die 202(1) may be electrically coupled to the IP die 202(2) through the TDV (in layer 920) and the base die 102(1) (in layer 208) by another electrical pathway 1304. The IP die 202(1) may be electrically coupled to the package substrate 212 by yet another electrical pathway 1306. The electrical pathway 1306 may be routed through the TDV in layers 920 and 210 and the base die 102(1) in layer 208. The IP die 202(1) may be electrically coupled to the IP die 202(3) through yet another electrical pathway 1308 that is routed through the TDV in layer 920, the base die 102(1), the bridge die 104(2) in layer 210, and the base die 102(2) in layer 208. These few electrical pathways are shown and described merely for ease of explanation. Any number of such electrical pathways may be provided within the broad scope of embodiments in the microelectronic assembly 100. These electrical pathways may include suitable routing through metallization stacks in the IC die, TDVs in the dielectric, and / or TSVs in the IC die, as appropriate, based on particular needs.

[0174] In various embodiments, the electrical pathways as described may form part of one or more NOCs 402 in microelectronic assembly 100. For example, electrical pathway 1306 may be part of a global NOC 402(1), electrical pathway 1308 with IP die 202(3) may have part of another NOC 402(2), electrical pathway 1304 with IP die 202(2) on the same base die 102(1) may have yet another NOC 402(3), electrical pathway 1302 with beyond CMOS die 902 may have yet another NOC 402(4), and so on.

[0175] 1-12 may be combined with any other feature to form a package with one or more IC dies described herein, for example, to form a modified microelectronic assembly 100. Although some such combinations are described above, further combinations and modifications are possible in various embodiments.

[0176] Device and Component Examples The packages disclosed herein, such as any of the embodiments shown in Figures 1-13 or any further embodiments described herein, may be included in any suitable electronic component. Figures 14-16 illustrate various examples of packages, assemblies, and devices that may be used with or include any of the IC packages disclosed herein.

[0177] 14 is a side cross-sectional view of an example IC package 2200 that may include an IC package according to any of the embodiments disclosed herein. In some embodiments, the IC package 2200 may be a system-in-package (SiP).

[0178] As shown, the package substrate 2252 may be formed of an insulator (e.g., ceramic, build-up film, epoxy film with filler particles therein, etc.) and may have conductive paths extending through the insulator between the first side 2272 and the second side 2274, or between different locations on the first side 2272 and / or between different locations on the second side 2274. These conductive paths may take the form of any of an interconnect structure having lines and / or vias.

[0179] Package substrate 2252 may include conductive contacts 2263 coupled to conductive paths 2262 through package substrate 2252, allowing circuitry within die 2256 and / or interposer 2257 to be electrically coupled to various of the conductive contacts 2264 (or to other devices included in package substrate 2252, not shown).

[0180] IC package 2200 may include an interposer 2257 coupled to a package substrate 2252 via conductive contacts 2261 of the interposer 2257, a first level interconnect 2265, and conductive contacts 2263 of the package substrate 2252. The first level interconnect 2265 shown is a solder bump, although any suitable first level interconnect 2265 may be used, such as, for example, a solder bump, a solder post, or a bond wire.

[0181] The IC package 2200 may include one or more die 2256 coupled to an interposer 2257 via conductive contacts 2254 of the die 2256, first level interconnects 2258, and conductive contacts 2260 of the interposer 2257. The conductive contacts 2260 may be coupled to conductive paths (not shown) through the interposer 2257, allowing circuitry within the die 2256 to be electrically coupled to various ones of the conductive contacts 2261 (or to other devices included in the interposer 2257, not shown). The illustrated first level interconnects 2258 are solder bumps, although any suitable first level interconnects 2258 may be used, such as, for example, solder bumps, solder posts, or bond wires. As used herein, a "conductive contact" may refer to a portion of a conductive material (e.g., a metal) that acts as an interface between different components; a conductive contact may be recessed in, flush with, or extend away from a surface of a component, and may take any suitable form (e.g., a conductive pad or socket).

[0182] In some embodiments, underfill material 2266 can be disposed between package substrate 2252 and interposer 2257 around first level interconnect 2265, and molding 2268 can be disposed around die 2256 and interposer 2257 in contact with package substrate 2252. In some embodiments, underfill material 2266 can be the same as molding 2268. An example of a material that can be used for underfill material 2266 and molding 2268 is epoxy, if suitable. Second level interconnect 2270 can be coupled to conductive contact 2264. Second level interconnect 2270 shown in the figure is a solder ball (e.g., for a ball grid array (BGA) configuration), although any suitable second level interconnect 2270 can be used (e.g., a pin in a pin grid array configuration, or a land in a land grid array configuration). Second level interconnect 2270 may be used to couple IC package 2200 to another component, such as a circuit board (e.g., a motherboard), an interposer, or another IC package, as known in the art and as described below with reference to FIG. 15.

[0183] In various embodiments, any of the dies 2256 may be a microelectronic assembly 100 as described herein. In embodiments in which the IC package 2200 includes multiple dies 2256, the IC package 2200 may be referred to as a multi-chip package (MCP). The dies 2256 may include circuitry to perform any desired function. For example, in addition to one or more of the dies 2256 being a microelectronic assembly 100 as described herein, one or more of the dies 2256 may be a logic die (e.g., a silicon-based die), one or more of the dies 2256 may be a memory die (e.g., a high-bandwidth memory), and so forth. In some embodiments, any of the dies 2256 may be implemented as described with reference to any of the preceding figures. In some embodiments, at least some of the dies 2256 may not include an implementation as described herein.

[0184] Although the IC package 2200 shown is a flip chip package, other package architectures may be used. For example, the IC package 2200 may be a BGA package, such as an embedded wafer level ball grid array (eWLB) package. In another example, the IC package 2200 may be a wafer level chip scale package (WLCSP) or a panel fan-out (FO) package. Although two dies 2256 are shown in the IC package 2200, the IC package 2200 may include any desired number of dies 2256. The IC package 2200 may include additional passive components, such as surface mount resistors, capacitors, and inductors, disposed on the first side 2272 or second side 2274 of the package substrate 2252, or on either side of the interposer 2257. More generally, the IC package 2200 may include any other active or passive components known in the art.

[0185] In some embodiments, interposer 2257 may not be included in IC package 2200 , and instead, die 2256 may be directly coupled to conductive contacts 2263 on first surface 2272 by first level interconnect 2265 .

[0186] 15 is a side cross-sectional view of an IC device assembly 2300 that may include components having one or more microelectronic assemblies 200 according to any of the embodiments disclosed herein. The IC device assembly 2300 includes a plurality of components disposed on a circuit board 2302 (which may be, for example, a motherboard). The IC device assembly 2300 includes components disposed on a first side 2340 of the circuit board 2302 and an opposing second side 2342 of the circuit board 2302, and generally, components may be disposed on one or both of the sides 2340 and 2342. In particular, suitable components of any of the components of the IC device assembly 2300 may include any of one or more microelectronic assemblies 100 according to any of the embodiments disclosed herein, and for example, any of the IC packages described below with reference to the IC device assembly 2300 may take the form of any of the embodiments of the IC package 2200 described above with reference to FIG. 14.

[0187] In some embodiments, the circuit board 2302 may be a PCB including multiple metal layers separated from one another by layers of insulation and interconnected by conductive vias. Any one or more of the metal layers may be formed with a desired circuit pattern for routing electrical signals (optionally with other metal layers) between components coupled to the circuit board 2302. In other embodiments, the circuit board 2302 may be a non-PCB package substrate.

[0188] As shown, in some embodiments, IC device assembly 2300 may include a package-on-interposer structure 2336 coupled to a first surface 2340 of circuit board 2302 by coupling components 2316. Coupling components 2316 may electrically and mechanically couple package-on-interposer structure 2336 to circuit board 2302 and may include solder balls (as shown), male and female portions of a socket, adhesive, underfill material, and / or any other suitable electrical and / or mechanical coupling structure.

[0189] The package-on-interposer structure 2336 may include an IC package 2320 coupled to the interposer 2304 by a bonding component 2318. The bonding component 2318 may take any suitable form depending on the desired functionality, such as those described above with reference to the bonding component 2316. In some embodiments, the IC package 2320 may be or include an IC package 2200, such as those described above with reference to FIG. 14. In some embodiments, the IC package 2320 may include at least one microelectronic assembly 100 disclosed herein. To avoid cluttering the figure, the microelectronic assembly 100 is not specifically shown in this figure.

[0190] Although one IC package 2320 is shown in the figure, multiple IC packages may be coupled to the interposer 2304, and in fact additional interposers may be coupled to the interposer 2304. The interposer 2304 may provide an intermediate package substrate used to bridge the circuit board 2302 and the IC package 2320. In general, the interposer 2304 may rewire connections to a wider pitch or reroute connections to different connections. For example, the interposer 2304 may couple the IC package 2320 to a BGA of the coupling component 2316 for coupling to the circuit board 2302.

[0191] In the illustrated embodiment, the IC package 2320 and the circuit board 2302 are mounted on opposite sides of the interposer 2304. In other embodiments, the IC package 2320 and the circuit board 2302 may be mounted on the same side of the interposer 2304. In some embodiments, more than two components may be interconnected by the interposer 2304.

[0192] The interposer 2304 may be formed of epoxy, fiberglass reinforced epoxy, ceramic materials, or polymeric materials such as polyimide. In some implementations, the interposer 2304 may be formed of another rigid or flexible material, which may include the same materials mentioned above used for semiconductor substrates, such as silicon, germanium, other group IV materials, and group III-V materials. The interposer 2304 may include metal interconnects 2308 and vias 2310, including but not limited to TSVs 2306. The interposer 2304 may further include embedded devices 2314, including both passive and active devices. Such devices may include but are not limited to capacitors, decoupling capacitors, resistors, inductors, fuses, diodes, transformers, sensors, electrostatic discharge (ESD) devices, and memory devices. More complex devices, such as radio frequency (RF) devices, power amplifiers, power management devices, antennas, arrays, sensors, and microelectromechanical systems (MEMS) devices may also be formed on the interposer 2304. The package-on-interposer structure 2336 may take the form of any of the package-on-interposer structures known in the art.

[0193] In some embodiments, IC device assembly 2300 may include an IC package 2324 coupled to a first surface 2340 of circuit board 2302 by a coupling component 2322. Coupling component 2322 may take the form of any of the embodiments described above with reference to coupling component 2316, and IC package 2324 may take the form of any of the embodiments described above with reference to IC package 2320.

[0194] In some embodiments, IC device assembly 2300 may include a package-on-package structure 2334 coupled to second surface 2342 of circuit board 2302 by coupling component 2328. Package-on-package structure 2334 may include IC package 2326 and IC package 2332 coupled together by coupling component 2330 such that IC package 2326 is disposed between circuit board 2302 and IC package 2332. Coupling components 2328 and 2330 may take the form of any of the embodiments of coupling component 2316 described above, and IC packages 2326 and / or 2332 may take the form of any of the embodiments of IC package 2320 described above. Package-on-package structure 2334 may be configured according to any of the package-on-package structures known in the art.

[0195] 16 is a block diagram of an example computing device 2400 that may include one or more components having one or more IC packages according to any of the embodiments disclosed herein. For example, any suitable one of the components of computing device 2400 may include a microelectronic assembly (e.g., 100) according to any of the embodiments disclosed herein. In another example, any one or more of the components of computing device 2400 may include any of the embodiments of IC package 2200 (e.g., as shown in FIG. 14). In yet another example, any one or more of the components of computing device 2400 may include IC device assembly 2300 (e.g., as shown in FIG. 15).

[0196] Although several components are shown included in computing device 2400, any one or more of these components may be omitted or repeated as appropriate for an application. In some embodiments, some or all of the components included in computing device 2400 may be mounted on one or more motherboards. In some embodiments, some or all of these components are fabricated on a single SOC die.

[0197] Also, in various embodiments, the computing device 2400 may not include one or more of the components illustrated in the figure, but rather the computing device 2400 may include interface circuitry for coupling to one or more of the components. For example, the computing device 2400 may not include a display device 2406, but rather include display device interface circuitry (e.g., connectors and drive circuitry) to which the display device 2406 may be coupled. In another set of examples, the computing device 2400 may not include an audio input device 2418 or an audio output device 2408, but rather include audio input or output device interface circuitry (e.g., connectors and support circuitry) to which the audio input device 2418 or the audio output device 2408 may be coupled.

[0198] Computing device 2400 may include a processing device 2402 (e.g., one or more processing devices). As used herein, the term "processing device" or "processor" may refer to any device or portion of a device that processes electronic data from registers and / or memory and converts the electronic data into other electronic data that may be stored in registers and / or memory. Processing device 2402 may include one or more DSPs, ASICs, CPUs, GPUs, cryptographic processors (specialized processors that execute cryptographic algorithms in hardware), server processors, or other suitable processing devices. Computing device 2400 may include memory 2404, which may itself include one or more memory devices, such as, for example, volatile memory (e.g., dynamic random access memory (DRAM)), non-volatile memory (e.g., read only memory (ROM)), flash memory, solid state memory, and / or a hard drive. In some embodiments, memory 2404 may include memory that shares a die with processing device 2402. This memory can be used as cache memory and may include embedded dynamic random access memory (eDRAM) or spin-transfer torque magnetic random access memory (STT-MRAM).

[0199] In some embodiments, computing device 2400 may include a communications chip 2412 (e.g., one or more communications chips). For example, communications chip 2412 may be configured to manage wireless communications for transmission of data to and from computing device 2400. The term "wireless" and its variants may be used to describe circuits, devices, systems, methods, techniques, communications channels, etc. that may communicate data over a non-solid medium using modulated electromagnetic radiation. This term does not imply that the associated devices do not include any wires, although in some embodiments they may not include any wires.

[0200] The communications chip 2412 may implement any of a number of wireless standards or protocols, including, but not limited to, WiFi (IEEE 802.11 family), IEEE standards including the IEEE 802.16 standard (e.g., IEEE 802.16-2005 amendment), the Long Term Evolution (LTE) project and its amendments, updates and / or revisions (e.g., Advanced LTE project), the Ultra Mobile Broadband (UMB) project (also known as "3GPP2"), etc. IEEE 802.16-compliant broadband wireless access (BWA) networks are commonly referred to as WiMAX networks (WiMAX is an acronym that stands for Worldwide Interoperability for Microwave Access), which is the certification mark for products that have passed IEEE 802.16 standard conformance and interoperability testing. The communications chip 2412 may operate in accordance with a Global System for Mobile Communications (GSM), General Packet Radio Service (GPRS), Universal Mobile Telecommunications System (UMTS), High Speed ​​Packet Access (HSPA), Evolved HSPA (E-HSPA), or LTE network. The communications chip 2412 may operate in accordance with an Enhanced Data Rates for GSM Evolution (EDGE), GSM EDGE Radio Access Network (GERAN), Universal Terrestrial Radio Access Network (UTRAN), or Evolved UTRAN (E-UTRAN). The communications chip 2412 may operate according to Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Digital Enhanced Cordless Telecommunications (DECT), Evolution Data Optimized (EV-DO), and variations thereof, as well as other wireless protocols designated as 3G, 4G, 5G, and beyond. The communications chip 2412 may operate according to other wireless protocols in other embodiments.Computing device 2400 may include an antenna 2422 to facilitate wireless communications and / or to receive other wireless communications (eg, AM or FM radio transmissions).

[0201] In some embodiments, the communications chip 2412 may manage wired communications, such as, for example, electrical, optical, or other suitable communications protocols (e.g., Ethernet). As mentioned above, the communications chip 2412 may include multiple communications chips. For example, a first communications chip 2412 may be for shorter range wireless communications, such as, for example, Wi-Fi and / or Bluetooth, and a second communications chip 2412 may be for longer range wireless communications, such as, for example, Global Positioning System (GPS), EDGE, GPRS, CDMA, WiMAX, LTE, Ev-DO, or others. In some embodiments, the first communications chip 2412 may be for wireless communications and the second communications chip 2412 may be for wired communications.

[0202] Computing device 2400 may include battery / power circuitry 2414. Battery / power circuitry 2414 may include one or more energy storage devices (e.g., batteries or capacitors) and / or circuitry for coupling components of computing device 2400 to an energy source separate from computing device 2400 (e.g., AC line power).

[0203] Computing device 2400 may include a display device 2406 (or corresponding interface circuitry, as described above), which may include any visual indicator, such as a heads-up display, a computer monitor, a projector, a touch screen display, a liquid crystal display (LCD), a light emitting diode display, or a flat panel display.

[0204] The computing device 2400 may include an audio output device 2408 (or corresponding interface circuitry, as described above). The audio output device 2408 may include any device that generates an audible indicator, such as, for example, a speaker, headphones, or earphones.

[0205] Computing device 2400 may include audio input device(s) 2418 (or corresponding interface circuitry, as described above), which may include any device that generates a signal representation of a sound, such as a microphone, a microphone array, or a digital musical instrument (e.g., an instrument with a musical instrument digital interface (MIDI) output).

[0206] Computing device 2400 may include a GPS device 2416 (or corresponding interface circuitry, as described above), which is in communication with a satellite-based system and may receive the location of computing device 2400, as is known in the art.

[0207] The computing device 2400 may include other output devices 2410 (or corresponding interface circuitry, as described above). Examples of other output devices 2410 may include audio codecs, video codecs, printers, wired or wireless transmitters for providing information to other devices, or additional storage devices.

[0208] The computing device 2400 may include other input devices 2420 (or corresponding interface circuitry, as described above). Examples of other input devices 2420 may include an accelerometer, a gyroscope, a compass, an image capture device, a keyboard, a cursor control device such as a mouse, a stylus, a touch pad, a barcode reader, a quick response (QR) code reader, any sensor, or a radio frequency identification (RFID) reader.

[0209] Computing device 2400 may have any desired form factor, such as, for example, a handheld or mobile computing device (e.g., a mobile phone, a smartphone, a mobile Internet device, a music player, a tablet computer, a laptop computer, a netbook computer, an ultrabook computer, a personal digital assistant (PDA), an ultra-mobile personal computer, etc.), a desktop computing device, a server or other networked computing component, a printer, a scanner, a monitor, a set-top box, an entertainment control unit, a vehicle control unit, a digital camera, a digital video recorder, or a wearable computing device. In some embodiments, computing device 2400 may be any other computing device that processes data.

[0210] Selection example The following paragraphs provide various examples of the embodiments disclosed herein.

[0211] Example 1 provides a microelectronic assembly (e.g., 100, Figures 1-2B) having a first plurality of IC dies (e.g., 202) on a first layer (e.g., 206), a second plurality of IC dies (e.g., 102) on a second layer (e.g., 208) between the first layer and a third layer (e.g., 210), and a third plurality of IC dies (e.g., 104) on the third layer, where the second plurality of IC dies have IC dies in an array of rows and columns (e.g., Figure 1), each IC die of the second plurality of IC dies is coupled to two or more IC dies of the first plurality of IC dies, and the third plurality of IC dies provides electrical coupling between adjacent IC dies of the second plurality of IC dies (e.g., Figure 1).

[0212] Example 2 provides the microelectronic assembly of Example 1, wherein the first plurality of IC die includes a first IC die having a first circuit, the second plurality of IC die includes a second IC die having a second circuit and a third IC die having a third circuit, the first circuit having more electrical connections with the second circuit than with the third circuit, and the first IC die is coupled to the second IC die with an interconnect having a pitch of less than 10 micrometers between adjacent ones of the interconnects.

[0213] Example 3 provides the microelectronic assembly of Example 1, wherein the first plurality of IC die includes a first IC die having a first circuit, the first circuit being electrically coupled to a second circuit external to the first IC die, the second plurality of IC die includes a second IC die proximate to the second circuit and a third IC die farther from the second circuit than the second IC die, and the first IC die is coupled to the second IC die by an electrical path from the first circuit to the second circuit, the electrical path being shorter than another electrical path from the third IC die to the second circuit.

[0214] Example 4 provides the microelectronic assembly of Example 1, wherein the first plurality of IC dies includes a first IC die having a first core and a second IC die having a second IP core, the first IP core being conductively coupled to the second IP core, and the second plurality of IC dies includes a third IC die, the first IC die and the second IC die being coupled to the third IC die with an interconnect having a pitch of less than 10 micrometers between adjacent ones of the interconnects.

[0215] Example 5 provides the microelectronic assembly of any one of Examples 1-4, wherein the first plurality of IC dies has smaller transistors than transistors of the second plurality of IC dies.

[0216] Example 6 provides the microelectronic assembly of any one of Examples 1 to 5, wherein a first IC die of the first plurality of IC dies has smaller transistors than transistors in a second IC die of the first plurality of IC dies.

[0217] Example 7 provides the microelectronic assembly of example 6, wherein the first IC die does not have a transistor with a ferromagnetic material and the second IC die has a transistor with a ferromagnetic material.

[0218] Example 8 provides the microelectronic assembly of example 6, wherein the first IC die has transistors configured to operate at a lower voltage than transistors in the second IC die.

[0219] Example 9 provides the microelectronic assembly of any one of Examples 1 to 8, wherein the IC dies of the second plurality of IC dies are substantially identical to one another in size and circuitry, and at least some of the IC dies of the first plurality of IC dies are not identical to one another in at least one of size and circuitry.

[0220] Example 10 provides the microelectronic assembly of any one of Examples 1 to 9, wherein a first IC die has a first circuit having a first function, a second IC die has a second circuit having a second function, a first microprocessor having the first function and the second function includes the first IC die and the second IC die of the first plurality of IC dies, and a second microprocessor having the first function but not the second function includes the first IC die but not the second IC die of the first plurality of IC dies.

[0221] Example 11 provides the microelectronic assembly of example 10, wherein the second plurality of IC dies are substantially identical in size, number, and circuitry between the first microprocessor and the second microprocessor.

[0222] Example 12 provides the microelectronic assembly of example 11, wherein the third plurality of IC dies are identical between the first microprocessor and the second microprocessor.

[0223] Example 13 provides the microelectronic assembly of any of Examples 1 to 12, wherein the interconnects at the interface between the first layer and the second layer are hybrid bonds distributed at an interconnect density greater than 10,000 interconnects per square millimeter.

[0224] Example 14 provides the microelectronic assembly of Example 13, wherein the interconnects at another interface between the second layer and the third layer are hybrid bonds distributed at an interconnect density greater than 10,000 interconnects per square millimeter.

[0225] Example 15 provides the microelectronic assembly of any one of Examples 1 to 14, wherein the microelectronic assembly further includes a network-on-chip (NOC) (e.g., FIG. 7 ), the NOC including a plurality of router circuits conductively coupled to IP cores in the first plurality of IC dies by a plurality of conductors, each IP core of the first plurality of IC dies being conductively coupled to at least one router circuit by one or more conductors in the plurality of conductors, and the plurality of router circuits being of the second plurality of IC dies.

[0226] Example 16 provides the microelectronic assembly of Example 15, wherein the plurality of router circuits comprises a first plurality of router circuits, the plurality of conductors comprises a first plurality of conductors, and the NOC further comprises a second plurality of router circuits of the second plurality of IC dies, each IC die of the second plurality of IC dies comprising at least one router circuit of the second plurality of router circuits.

[0227] Example 17 provides the microelectronic assembly of any one of Examples 1-16, wherein the IC dies of the first plurality of IC dies are of various shapes and sizes (eg, FIGS. 3A-3B).

[0228] Example 18 provides the microelectronic assembly of any one of Examples 1-16, wherein the IC dies of the first plurality of IC dies are similarly sized (eg, FIG. 4).

[0229] Example 19 provides the microelectronic assembly of any one of Examples 1 to 18, wherein one of the IP cores comprises a central processing unit (CPU) core and another of the IP cores comprises a memory.

[0230] Example 20 provides the microelectronic assembly of any one of Examples 1 to 19, wherein one of the IP cores has a general-purpose processor circuit and another of the IP cores has an accelerator (e.g., Figures 4-6).

[0231] Example 21 provides the microelectronic assembly of Example 20, wherein the array has a first array, the first plurality of IC dies are in a second array above the first array, and the IC dies in the second array are sized and shaped to fit within the boundaries of the first array (e.g., FIG. 4).

[0232] Example 22 provides the microelectronic assembly of Example 21, wherein a particular IC die of the first plurality of IC dies is in a subarray on a single IC die of the second plurality of IC dies, and the particular IC die in the subarray is sized and shaped to fit within a boundary of the single IC die of the second plurality of IC dies.

[0233] Example 23 provides the microelectronic assembly of any one of Examples 1 to 22, wherein the second plurality of IC dies includes a voltage regulator circuit (e.g., FIG. 9).

[0234] Example 24 provides the microelectronic assembly of any one of Examples 1 to 23, wherein the second plurality of IC dies includes a serial-deserializer (SerDes) circuit and a level shifter (LS) circuit (e.g., FIG. 9).

[0235] Example 25 provides the microelectronic assembly of any one of Examples 1 to 24, wherein the microelectronic assembly further includes a fourth plurality of IC dies in a fourth layer (e.g., 920, FIG. 10), the fourth layer being between the first layer and the second layer.

[0236] Example 26 provides the microelectronic assembly of any one of Examples 1 to 25, wherein at least one IC die of the first plurality of IC dies comprises a stack of IC dies having transistors with ferromagnetic material (e.g., FIG. 10).

[0237] Example 27 provides the microelectronic assembly of any one of Examples 1 to 26, further comprising a fourth plurality of IC dies (e.g., 104) in a fourth layer (e.g., 920) between the first layer and the second layer (e.g., FIG. 13), the fourth plurality of IC dies having conductive paths between adjacent IC dies of the first plurality of IC dies.

[0238] Example 28 provides the microelectronic assembly of any one of Examples 1 to 27, wherein at least one IP core is distributed across multiple IC dies of the first plurality of IC dies.

[0239] Example 29 provides an IC package (e.g., FIG. 2B ) having a first IC die (e.g., 202) on a first layer (e.g., 206), a second IC die (e.g., 102) and a third IC die (e.g., 102) on a second layer (e.g., 208), a fourth IC die (e.g., 104) on a third layer (e.g., 210), and a package substrate (e.g., 212) coupled to the third layer, where the second layer is between the first layer and the third layer, the first IC die is attached to the second IC die, the third IC die is adjacent to the second IC die, and the fourth IC die has a conductive path between the second IC die and the third IC die.

[0240] Example 30 provides the IC package of example 29, wherein the second IC die has circuitry conductively coupled to one or more IP cores of the first plurality of IC dies.

[0241] Example 31 provides the IC package of example 30, wherein at least one of the IP cores comprises a processor circuit, the circuit comprising a voltage regulator circuit.

[0242] Example 32 provides the IC package of any one of Examples 29 to 31, wherein the second IC die is the same in size and circuitry as the third IC die.

[0243] Example 33 provides the IC package of any one of examples 29-32, wherein the first IC die has smaller transistors than transistors in the second IC die.

[0244] Example 34 provides the IC package of any one of Examples 29 to 33 (e.g., FIG. 11 ), wherein the second IC die includes a plurality of stacked IC dies, each IC die of the stacked IC dies including a transistor having a ferromagnetic material.

[0245] Example 35 provides the IC package of any one of Examples 29-34 (e.g., FIGS. 9-10), wherein the first IC die includes a plurality of stacked IC dies, each IC die of the plurality of stacked IC dies including a transistor having a ferromagnetic material.

[0246] Example 36 provides the IC package of any one of Examples 29 to 35, wherein the fourth IC die does not have any diode or transistor.

[0247] Example 37 provides the IC package of any one of Examples 29 to 35, wherein the fourth IC die includes at least one of a diode and a transistor.

[0248] Example 38 provides the IC package of any one of Examples 29 to 37, wherein the second IC die, the third IC die, and the fourth IC die have through-substrate vias (TSVs).

[0249] Example 39 provides the IC package of example 38, wherein the TSVs carry power to the first IC die.

[0250] Example 40 provides the IC package of any one of Examples 29 to 39, wherein the fourth IC die is embedded in a dielectric material.

[0251] Example 41 provides the IC package of example 40, wherein the dielectric material has a through dielectric via (TDV).

[0252] Example 42 provides the IC package of any one of Examples 29 to 41, wherein the second IC die and the third IC die are in a dielectric material.

[0253] Example 43 provides an arrangement of IC dies, the arrangement comprising: a first plurality of IC dies having an IC die with an IP core; a second plurality of IC dies having an IC die, each of the IC dies of the second plurality of IC dies having at least one circuit conductively coupled to one or more of the IP cores; and a third plurality of IC dies having an IC die, each of the IC dies of the third plurality of IC dies having at least one conductive path between the second plurality of IC dies with an interconnect having a pitch of less than 10 micrometers, wherein the first plurality of IC dies are on a first layer, the second plurality of IC dies are on a second layer, the second layer is between the first layer and a third layer, and the third plurality of IC dies are on the third layer.

[0254] Example 44 provides the arrangement of Example 43, wherein the arrangement is a first arrangement of IC dies, the first plurality of IC dies includes a first IC die having a first IP core at a position within the first layer, and a second arrangement of IC dies is substantially similar in structure to the first arrangement of IC dies, except that a second IC die having a second IP core different from the first IP core is at the position within the first layer.

[0255] Example 45 provides the arrangement of any one of examples 43-44, wherein the second plurality of IC dies includes a shared cache (e.g., 810) accessible by the first plurality of IC dies.

[0256] Example 46 provides the arrangement of any one of Examples 43-45, wherein the first plurality of IC dies are coupled to the second plurality of IC dies with interconnects having a pitch of less than 10 micrometers between adjacent ones of the interconnects.

[0257] Example 47 provides the arrangement of any one of examples 43 to 46, wherein the second plurality of IC dies are in a regular array of rows and columns.

[0258] Example 48 provides the arrangement of example 47, wherein the second plurality of IC dies are identical to one another in shape and circuitry.

[0259] Example 49 provides the arrangement of any one of Examples 43 to 48, wherein the IP cores are conductively coupled to router circuits disposed on the second plurality of IC dies, the router circuits routing electrical signals between the IP cores.

[0260] Example 50 provides the arrangement of any one of Examples 43 to 49, wherein a first portion of the first plurality of IC dies has general purpose processor circuitry and a second portion of the first plurality of IC dies has accelerator circuitry, the first portion and the second portion being in an array of rows and columns.

[0261] Example 51 provides the arrangement of any one of examples 43-50, wherein the first plurality of IC dies have transistors that are different sizes than transistors in the second plurality of IC dies.

[0262] Example 52 provides the arrangement of example 51, wherein the first plurality of IC dies and the second plurality of IC dies have transistors that are different sizes than transistors in the third plurality of IC dies.

[0263] Example 53 provides the arrangement of any one of examples 43 to 52, wherein the third plurality of IC dies does not include active circuit elements.

[0264] Example 54 includes a first plurality of IC dies (e.g., 202) in a first layer (e.g., 206), a second plurality of IC dies (e.g., 102) in a second layer (e.g., 208), and a third plurality of IC dies (e.g., 104) in a third layer (e.g., 210), the second layer being between the first layer and the third layer, a first portion of the first plurality of IC dies (e.g., 202(1)) having a general purpose processor circuit (e.g., 404), and a second portion of the first plurality of IC dies (e.g., 202(2)) having a general purpose processor circuit (e.g., 404). ) has accelerator circuitry (e.g., 406), the first portion and the second portion are physically separate from one another, the second plurality of IC die has IC dies including circuitry conductively coupled to the general purpose processor circuitry and the accelerator circuitry (e.g., FIG. 4), and the third plurality of IC dies has conductive paths providing electrical coupling between adjacent ones of the second plurality of IC dies (e.g., FIG. 6), providing a microelectronic assembly (e.g., 100, FIGS. 4-6).

[0265] Example 55 provides the microelectronic assembly of example 54, wherein the first plurality of IC dies are in a first array of rows and columns and the second plurality of IC dies are in a second array of rows and columns.

[0266] Example 56 provides the microelectronic assembly of any one of Examples 54-55, wherein the first plurality of IC dies is smaller than the second plurality of IC dies.

[0267] Example 57 provides the microelectronic assembly of any one of Examples 54 to 56, wherein each IC die in the first portion includes one or more general-purpose processor circuits and each IC die in the second portion includes one or more accelerator circuits.

[0268] Example 58 provides the microelectronic assembly of any one of Examples 54 to 57, wherein the first portion of the first plurality of IC dies is bonded to a third portion of the second plurality of IC dies and the second portion of the first plurality of IC dies is bonded to a fourth portion of the second plurality of IC dies.

[0269] Example 59 provides the microelectronic assembly of example 58, wherein the third portion and the fourth portion are physically independent of each other.

[0270] Example 60 provides the microelectronic assembly of any one of Examples 58 to 59, wherein a plurality of the IC dies in the first portion are included on and within the footprint of a single IC die in the third portion.

[0271] Example 61 provides the microelectronic assembly of Example 60, wherein a plurality of the IC dies of the second portion are included on and within the footprint of a single IC die of the IC dies of the fourth portion.

[0272] Example 62 provides the microelectronic assembly of Example 61, wherein the single IC die of the third portion of IC dies is sized differently than the single IC die of the fourth portion of IC dies.

[0273] Example 63 provides the microelectronic assembly of any one of Examples 54 to 62, wherein the IC dies of the second portion are interspersed among the IC dies of the first portion.

[0274] Example 64 provides an IC package having a first IC die having general purpose processor circuitry, a second IC die having accelerator circuitry, a third IC die having circuitry conductively coupled to the general purpose processor circuitry, a fourth IC die having circuitry conductively coupled to the accelerator circuitry, and a fifth IC die having a conductive path between the third IC die and the fourth IC die, wherein the first IC die and the second IC die are on a first layer, the third IC die and the fourth IC die are on a second layer, the second layer is between the first layer and a third layer, the fifth IC die is on the third layer, the first IC die is electrically and mechanically coupled to the third IC die, and the second IC die is electrically and mechanically coupled to the fourth IC die.

[0275] Example 65 provides the IC package of Example 64, further comprising a plurality of first IC dies electrically and mechanically coupled to the third IC die with interconnects having a pitch of less than 10 micrometers between adjacent ones of the interconnects.

[0276] Example 66 provides the IC package of any one of Examples 64 to 65, further comprising a plurality of second IC dies electrically and mechanically coupled to the fourth IC die with interconnects having a pitch of less than 10 micrometers between adjacent ones of the interconnects.

[0277] Example 67 provides the IC package of any one of examples 64-66, wherein the first IC die has transistors that are different sizes than transistors in the second IC die.

[0278] Example 68 provides the IC package of example 67, wherein the third IC die has transistors that are different sizes than transistors in the fourth IC die.

[0279] Example 69 provides an arrangement of IC dies having a general purpose processor circuit in a first plurality of IC dies, an accelerator circuit in a second plurality of IC dies, support circuitry in a third plurality of IC dies conductively coupled to the general purpose processor circuit and the accelerator circuit, and a NOC having a router circuit conductively coupled to the general purpose processor circuit and the accelerator circuit, the router circuit being in the third plurality of IC dies, the accelerator circuit being distributed among the general purpose processor circuits, and in some embodiments, a conductive path between the general purpose processor circuits may not pass through the accelerator circuit.

[0280] Example 70 provides the arrangement of example 69, including a fourth plurality of IC dies having conductive paths between adjacent ones of the third plurality of base dies.

[0281] Example 71 provides the arrangement of any one of Examples 69 to 70, wherein the NOC includes a first NOC and a second NOC, the first NOC including a first portion of the router circuitry and the second NOC including a second portion of the router circuitry, the first portion of the router circuitry configured to conductively couple the general purpose processor circuitry and the accelerator circuitry, and the second portion of the router circuitry configured to conductively couple the general purpose processor circuitry therebetween.

[0282] Example 72 provides the arrangement of any one of Examples 69-71, wherein the first plurality of IC dies and the second plurality of IC dies are coplanar within a first layer, and the third plurality of IC dies are in a second layer, the second layer is not coplanar with the first layer, and the first layer is electrically and mechanically coupled to the second layer with interconnects having a pitch of less than 10 micrometers between adjacent ones of the interconnects.

[0283] Example 73 provides the arrangement of any one of Examples 69 to 72, wherein the support circuitry includes a first set of support circuitry conductively coupled to the general purpose processor circuitry and a second set of support circuitry conductively coupled to the accelerator circuitry.

[0284] Example 74 provides a microelectronic assembly having a first plurality of IC die (e.g., 202) on a first tier (e.g., 206), a second plurality of IC die (e.g., 102) on a second tier (e.g., 208), a third plurality of IC die (e.g., 104) on a third tier (e.g., 210), and a network-on-chip (NOC) spanning the microelectronic assembly, the NOC having a plurality of conductively coupled router circuits, the second tier being between the first tier and the third tier, the first plurality of IC die having IP cores conductively coupled through the router circuits (e.g., FIG. 7A), the router circuits being of the second plurality of IC die (e.g., FIG. 8A), and the third plurality of IC die having conductive paths between adjacent ones of the second plurality of IC die (e.g., FIGS. 7A-8B, 11, 12).

[0285] Example 75 provides the microelectronic assembly of Example 74, wherein the IP cores of the first plurality of IC dies are coupled to the router circuit by links, each link having a plurality of interconnects between the first plurality of IC dies and the second plurality of IC dies, a first IC die having an IP core using a higher bandwidth than another IP core in the second IC die has a greater number of interconnects per link than the second IC die and is greater than a fourth IC die of the first plurality of IC dies, and a third IC die of the first plurality of IC dies has a greater number of interconnects per link than the fourth IC die.

[0286] Example 76 provides the microelectronic assembly of any one of Examples 74-75, wherein the NOC includes a first NOC, a second NOC, and a third NOC, the first NOC having a first plurality of router circuits conductively coupled to an IC die of the second plurality of IC dies, the second NOC having a second plurality of router circuits conductively coupled to an IC die of the first plurality of IC dies, and the third NOC having a third plurality of router circuits conductively coupled to IP cores in an IC die of the first plurality of IC dies (e.g., FIG. 7B ).

[0287] Example 77 provides the microelectronic assembly of example 76, wherein each IC die of the second plurality of IC dies has a distinct one of the second NOCs.

[0288] Example 78 provides the microelectronic assembly of any one of Examples 76-77, wherein each IC die of the first plurality of IC dies has a distinct one of the third NOCs.

[0289] Example 79 provides the microelectronic assembly of any one of Examples 74 to 78, wherein the NOC includes a first NOC having a first plurality of router circuits and a parallel second NOC having a second plurality of router circuits (e.g., FIG. 7A).

[0290] Example 80 illustrates a system in which IP cores of the first plurality of IC dies are coupled to the router circuit by links (e.g., 704), each link having a plurality of interconnects between the first plurality of IC dies and the second plurality of IC dies, at least two links being conductively coupled together at a node, a first node of the NOC is coupled to at least one of a first interconnect (e.g., 804), a second interconnect (e.g., 806), and a third interconnect (e.g., 808), the first interconnect being coupled to the first node. 80. The microelectronic assembly of any one of Examples 74-79, wherein the first node is on the second tier and the fourth node is on the first tier, the second interconnect is configured to conductively couple the first node to a second node of the NOC, the second interconnect is configured to conductively couple the first node to a third node of another NOC, and the third interconnect is configured to conductively couple the first node to a fourth node of the NOC, the first node, the second node, and the third node are on the second tier, and the fourth node is on the first tier (e.g., FIG. 8A).

[0291] Example 81 provides the microelectronic assembly of example 80, wherein the first interconnect and the third interconnect form part of a link (eg, 704) within the NOC.

[0292] Example 82 provides the microelectronic assembly of example 80, wherein the second interconnect forms part of a link (eg, 706) connecting the NOC with the other NOC.

[0293] Example 83 provides the microelectronic assembly of any one of Examples 80 to 82, wherein the first interconnect and the second interconnect have conductive paths within at least one of the second layer and the third layer, and the third interconnect has one or more die-to-die (DTD) interconnects between the first layer and the second layer (e.g., FIG. 8C ).

[0294] Example 84 provides the microelectronic assembly of any one of Examples 80 to 83, wherein the first node is on at least one of the second plurality of IC dies and the third plurality of IC dies, and the fourth node is on the first plurality of IC dies.

[0295] Example 85 provides the microelectronic assembly of any one of Examples 74 to 84, wherein a cache is distributed among the second plurality of IC dies, and the cache is conductively coupled to the IP cores by node, link, and router circuitry of the NOC.

[0296] Example 86 provides the microelectronic assembly of any one of Examples 74 to 85, wherein at least one of the first plurality of IC dies and the second plurality of IC dies comprises a stack of IC dies having a transistor having a ferromagnetic material.

[0297] Example 87 provides the microelectronic assembly of example 86, wherein a link in the NOC couples an IC die having a transistor with ferromagnetic material to another IC die having voltage regulation circuitry, voltage level shifting circuitry, and frequency conversion circuitry.

[0298] Example 88 includes a first IC die, a second IC die, and a third IC die in a first layer, a first IP core in the first IC die, a second IP core in the second IC die, a third IP core in the third IC die, and a fourth IP core in the third IC die, a fourth IC die and a fifth IC die in a second layer, a sixth IC die in a third layer, a package substrate coupled to the third layer, and a package substrate coupled to the first IP core, the second IP core, the third IP core, and the fourth IP core. and a network-on-chip (NOC) coupling the cores to the first and second IC die, the second layer being between the first and third layers, the NOC having a plurality of conductively coupled router circuits, the first IC die and the second IC die being attached to the fourth IC die, the third IC die being attached to the fifth IC die, and the sixth IC die having a conductive path between the fourth IC die and the fifth IC die (e.g., FIG. 8C).

[0299] Example 89 provides the IC package of Example 88, wherein the NOC has a first NOC and a second NOC, the first NOC has a first plurality of router circuits conductively coupling the first IP core and the second IP core to the third IP core via the fourth IC die and the fifth IC die, and the second NOC has a second plurality of router circuits conductively coupling the first IP core to the second IP core via the fourth IC die.

[0300] Example 90 provides the IC package of Example 89, wherein the IC package further includes a fourth IP core in the third IC die, and the NOC further includes a third NOC having a third plurality of router circuits conductively coupled to the third IP core and the fourth IP core.

[0301] Example 91 provides the IC package of any one of Examples 88-90, wherein the plurality of router circuits are on the second layer and are conductively coupled together in a mesh network (e.g., FIG. 8A).

[0302] Example 92 provides the IC package of Example 91, wherein a plurality of conductors couple the plurality of router circuits, each conductive coupling between any two router circuits being a link in the mesh network, each link having conductive joints and interconnects passing through the first layer, the second layer, and the third layer.

[0303] Example 93 provides the IC package of any one of examples 88-92, wherein the third IC die comprises a stack of IC dies having transistors with ferromagnetic material.

[0304] Example 94 provides the IC package of Example 93, wherein the third IP core is within one of the IC dies in the stack of IC dies and the fourth IP core is within another of the IC dies in the stack of IC dies.

[0305] Example 95 provides the IC package of example 94, wherein the third IP core comprises a cache and the fourth IP core comprises a processor circuit.

[0306] Example 96 provides the IC package of any one of Examples 88 to 95, wherein the fourth IP die comprises a stack of IC dies having transistors with ferromagnetic material coupled with DTD interconnects having a pitch of less than 10 micrometers.

[0307] Example 97 provides the IC package of any one of Examples 88 to 96, wherein a cache is distributed across the fourth IC die and the fifth IC die of the second layer, and the cache is conductively coupled to the first IP core, the second IP core, and the third IP core.

[0308] Example 98 provides an arrangement of IC dies having IP cores in a first plurality of IC dies in a first layer and router circuits in a second plurality of IC dies in a second layer, the second layer not coplanar with the first layer, and each IP core conductively coupled to one of the router circuits by one or more conductors over a NOC.

[0309] Example 99 provides the arrangement of Example 98, wherein the NOC has a first NOC coupling a first portion of the IP cores by a first plurality of router circuits, and the NOC has a second NOC coupling a second portion of the IP cores by a second plurality of router circuits.

[0310] Example 100 provides the arrangement of any one of Examples 98-99, wherein at least one of the first plurality of IC dies and the second plurality of IC dies comprises a stack of IC dies having transistors having ferromagnetic material.

[0311] Example 101 provides the arrangement of any one of Examples 98-100, wherein the router circuitry (eg, 702) is distributed across the second plurality of IC dies.

[0312] Example 102 provides the arrangement of any one of Examples 98 to 101, wherein the arrangement further includes a cache (e.g., 810) distributed across the second plurality of IC dies, the cache being conductively coupled on the NOC by the IP cores of the first plurality of IC dies.

[0313] Example 103 provides the arrangement of any one of Examples 98 to 102, wherein the NOC has a mesh topology having a first plurality of conductors and a second plurality of conductors, each of the second plurality of conductors intersecting with at least one of the first plurality of conductors to form a plurality of network nodes, each of the network nodes being at an intersection of one of the first plurality of conductors with one of the second plurality of conductors, and each of the first plurality of conductors being disposed orthogonally to each of the second plurality of conductors.

[0314] Example 104 provides a microelectronic assembly having a first layer and a second layer of IC dies, the first layer and the second layer being non-coplanar, the first layer and the second layer being electrically and mechanically coupled with interconnects having a pitch of less than 10 micrometers between adjacent ones of the interconnects, and a first IC die of the plurality of IC dies does not have a transistor having a ferromagnetic material and a second IC die of the plurality of IC dies has a transistor having a ferromagnetic material (e.g., FIG. 9).

[0315] Example 105 provides the microelectronic assembly of example 104, wherein the first IC die is coupled to the second IC die, the first IC die having circuitry for voltage conversion, frequency shifting, and voltage regulation.

[0316] Example 106 provides the microelectronic assembly of Example 105, wherein the circuit for voltage conversion is configured to shift a voltage between a first voltage level and a second voltage level, the first IC die is configured to operate at the first voltage level, and the second IC die is configured to operate at the second voltage level, the first voltage level being higher than the second voltage level.

[0317] Example 107 provides the microelectronic assembly of any one of Examples 105 to 106, wherein the circuitry for frequency shifting includes circuitry for varying a frequency of a signal between a first frequency and a second frequency, the first IC die configured to operate with a signal at the first frequency, and the second IC die configured to operate with a signal at the second frequency, the first frequency being higher than the second frequency.

[0318] Example 108 provides the microelectronic assembly of Example 107, wherein a first number of interconnects are for signals at the first frequency and a second number of interconnects are for signals at the second frequency, the first number being smaller than the second number, and the second number of interconnects having a pitch of less than 10 micrometers between adjacent ones of the second number of interconnects.

[0319] Example 109 provides the microelectronic assembly of any one of Examples 105 to 108, wherein the first IC die is on the first layer and the second IC die is on the second layer.

[0320] Example 110 provides the microelectronic assembly of example 109, wherein the first IC die has conductive paths between the second IC die and other IC dies of the plurality of IC dies.

[0321] Example 111 provides the microelectronic assembly of any one of Examples 109-110, wherein the first IC die has circuitry conductively coupled to an IP core (eg, 904) in the second IC die.

[0322] Example 112 provides the microelectronic assembly of any one of Examples 104 to 111, wherein the microelectronic assembly further comprises a third IC die of the plurality of IC dies, the third IC die having circuitry configured to operate at a voltage of 0.35 V or less.

[0323] Example 113 provides the microelectronic assembly of any one of Examples 104 to 112, wherein the first IC die comprises a general purpose processor circuit and the second IC die comprises an accelerator.

[0324] Example 114 provides the microelectronic assembly of any one of Examples 104 to 113, further comprising a stack of IC dies including transistors having ferromagnetic material within the plurality of IC dies.

[0325] Example 115 provides the microelectronic assembly of example 114, wherein the stack of IC dies comprises circuitry conductively coupled to IP cores of the first IC die and the second IC die.

[0326] Example 116 provides the microelectronic assembly of any one of Examples 104 to 115, wherein the IC die of the second layer are in an array of rows and columns, the IC die of the second layer have circuitry conductively coupled to the IC die of the first layer, the IC die of the first layer have IP cores, and the IC die of the second layer have router circuitry of a NOC facilitating communication between the IP cores.

[0327] Example 117 provides the microelectronic assembly of example 116, wherein the first IC die is on the first layer and the second IC die is on the second layer.

[0328] Example 118 provides the microelectronic assembly of example 116, wherein the second IC die is on the first layer and the first IC die is on the second layer.

[0329] Example 119 provides the microelectronic assembly of any one of Examples 116-118, wherein the IC die of the second tier includes a distributed cache conductively coupled to the IC die of the first tier.

[0330] Example 120 provides an IC package having a first IC die on a first layer, a second IC die on a second layer, a third IC die on a third layer, and a package substrate coupled to the third layer, the second layer being between the first layer and the third layer, interconnects between the first layer and the second layer having an interconnect with a pitch of less than 10 micrometers between adjacent ones of the interconnects, one of the first IC die and the second IC die having a transistor with a ferromagnetic material, and the other of the first IC die and the second IC die having a first circuit for voltage conversion, a second circuit for frequency shifting, and a third circuit for voltage regulation.

[0331] Example 121 provides the IC package of example 120, wherein the second IC die has circuitry conductively coupled to one or more IP cores in the first IC die.

[0332] Example 122 provides the IC package of any one of Examples 120-121, wherein the third circuit for voltage regulation provides power at a voltage level at which the one of the first IC die and the second IC die having a transistor with a ferromagnetic material is configured to operate.

[0333] Example 123 provides the IC package of any one of Examples 120 to 122, wherein the first circuit for voltage conversion is configured to shift a voltage between a first voltage level and a second voltage level, the first voltage level being higher than the second voltage level, the one of the first IC die and the second IC die having a transistor with a ferromagnetic material configured to operate at the second voltage level, and the other of the first IC die and the second IC die is configured to operate at the first voltage level.

[0334] Example 124 provides the IC package of any one of Examples 120 to 123, wherein the second circuit for frequency shifting includes circuitry for varying a frequency of a signal between a first frequency and a second frequency, the first frequency being greater than the second frequency, and wherein the one of the first IC die and the second IC die having transistors with ferromagnetic material is configured to operate with signals at the second frequency, and the other of the first IC die and the second IC die is configured to operate with signals at the first frequency.

[0335] Example 125 provides the IC package of any one of Examples 120-124, wherein the third IC die has a conductive path between the second IC die and an adjacent IC die in the second layer.

[0336] Example 126 provides the IC package of any one of Examples 120-125, wherein the one of the first IC die having a transistor with a ferromagnetic material and the second IC die is a stack of IC dies.

[0337] Example 127 provides the IC package of any one of Examples 120 to 126, wherein the one of the first IC die and the second IC die having a transistor with a ferromagnetic material generates less heat than the other of the first IC die and the second IC die.

[0338] Example 128 provides an arrangement of IC dies comprising: a plurality of IP cores distributed across at least a first layer and a second layer; a router circuit conductively coupled to the plurality of IP cores; and a circuit distributed in an array of IC dies and conductively coupled to the plurality of IP cores and the router circuit, wherein the first layer and the second layer are not coplanar, at least one IP core of the plurality of IP cores comprises a transistor having a ferromagnetic material, and the router circuit is distributed in the array of IC dies.

[0339] Example 129 provides the arrangement of example 128, wherein at least one of the plurality of IP cores includes a general purpose processor circuit and at least another of the plurality of IP cores includes an accelerator circuit.

[0340] Example 130 provides the arrangement of any one of Examples 128-129, wherein the at least one IP core having transistors with ferromagnetic material comprises at least one of a memory, an accelerator circuit, and a general purpose processor circuit.

[0341] Example 131 provides the arrangement of any one of Examples 128 to 130, wherein at least one of the IC dies in the array includes a transistor having a ferromagnetic material.

[0342] Example 132 provides the arrangement of any one of Examples 128 to 131, wherein one of the IC dies in the array is coupled to the at least one IP core having a transistor with a ferromagnetic material and has circuitry for voltage conversion, frequency shifting, and voltage regulation.

[0343] Example 133 provides the arrangement of any one of Examples 128-132, wherein a cache conductively coupled to the plurality of IP cores is distributed among the array of IC dies.

[0344] The above description of illustrated implementations of the disclosure, including what is described in the Abstract, is not intended to be exhaustive or to limit the disclosure to the precise forms disclosed herein. While specific implementations and examples of the disclosure are described herein for illustrative purposes, various equivalent modifications are possible within the scope of the disclosure, as those skilled in the art will recognize.

Claims

1. a first plurality of integrated circuit (IC) dies in a first layer; a second plurality of IC dies in a second layer between the first layer and a third layer; a third plurality of IC dies in the third layer; having the second plurality of IC dies includes an array of IC dies in rows and columns; each IC die of the second plurality of IC dies is coupled to two or more IC dies of the first plurality of IC dies; the third plurality of IC dies providing electrical coupling between adjacent IC dies of the second plurality of IC dies; Microelectronics assembly.

2. the first plurality of IC dies includes a first IC die having a first circuit; the second plurality of IC dies includes a second IC die having a second circuit and a third IC die having a third circuit; the first circuit has more electrical connections with the second circuit than it has with the third circuit; the first IC die is coupled to the second IC die with interconnects having a pitch of less than 10 micrometers between adjacent ones of the interconnects; The microelectronic assembly of claim 1 .

3. the first plurality of IC dies includes a first IC die having a first circuit, the first circuit being electrically coupled to a second circuit external to the first IC die; the second plurality of IC dies includes a second IC die proximate to the second circuit and a third IC die farther from the second circuit than the second IC die; the first IC die is coupled to the second IC die by an electrical path from the first circuit to the second circuit, the electrical path being shorter than another electrical path from the third IC die to the second circuit; The microelectronic assembly of claim 1 .

4. the first plurality of IC dies includes a first IC die having a first intellectual property (IP) core and a second IC die having a second IP core; the first IP core is conductively coupled to the second IP core; the second plurality of IC dies includes a third IC die; the first IC die and the second IC die are coupled to the third IC die with interconnects having a pitch of less than 10 micrometers between adjacent ones of the interconnects. The microelectronic assembly of claim 1 .

5. 5. The microelectronic assembly of claim 1, wherein the first plurality of IC dies has smaller transistors than transistors in the second plurality of IC dies.

6. 5. The microelectronic assembly of claim 1, wherein a first IC die of the first plurality of IC dies has smaller transistors than transistors in a second IC die of the first plurality of IC dies.

7. 7. The microelectronic assembly of claim 6, wherein the first IC die does not have transistors with ferromagnetic material and the second IC die has transistors with ferromagnetic material.

8. The microelectronic assembly of claim 6 , wherein the first IC die has transistors configured to operate at a lower voltage than transistors in the second IC die.

9. the first IC die has a first circuit having a first function; the second IC die has a second circuit having a second function; a first microprocessor having the first function and the second function includes the first IC die and the second IC die in the first plurality of IC dies; a second microprocessor having the first functionality but not the second functionality includes the first IC die but not the second IC die in the first plurality of IC dies; 5. A microelectronic assembly according to any one of the preceding claims.

10. The microelectronic assembly further comprises a network on chip (NOC), the NOC comprising: a plurality of router circuits conductively coupled to IP cores in the first plurality of IC dies by a plurality of conductors; having each IP core in the first plurality of IC dies is conductively coupled to at least one router circuit by one or more conductors in the plurality of conductors; the plurality of router circuits are in the second plurality of IC dies. The microelectronic assembly of claim 1 .

11. the plurality of router circuits comprises a first plurality of router circuits; the plurality of conductors comprises a first plurality of conductors, the NOC further comprising a second plurality of router circuits within the second plurality of IC dies, each IC die of the second plurality of IC dies comprising at least one router circuit within the second plurality of router circuits; The microelectronic assembly of claim 10.

12. a first IC die in a first layer; a second IC die and a third IC die in a second layer; a fourth IC die in a third layer; and a package substrate coupled to the third layer; having the second layer is between the first layer and the third layer; the first IC die is attached to the second IC die; the third IC die is adjacent to the second IC die; the fourth IC die having a conductive path between the second IC die and the third IC die; IC package.

13. The IC package of claim 12 , wherein the second IC die has circuitry conductively coupled to one or more IP cores in the first plurality of IC dies.

14. The IC package of claim 13 , wherein at least one of said IP cores comprises a processor circuit, said circuit comprising a voltage regulator circuit.

15. 15. The IC package of claim 12, wherein the second IC die is identical in size and circuitry to the third IC die.

16. 15. The IC package of claim 12, wherein the first IC die has smaller transistors than transistors in the second IC die.

17. 15. The IC package of claim 12, wherein the second IC die comprises a plurality of stacked IC dies, each IC die of the plurality of stacked IC dies comprising a transistor having a ferromagnetic material.

18. 15. The IC package of claim 12, wherein the fourth IC die does not include any diodes or transistors.

19. The IC package of claim 12 , wherein the second IC die, the third IC die, and the fourth IC die have through-substrate vias (TSVs).

20. 15. The IC package of claim 12, wherein the fourth IC die is embedded in a dielectric material.

21. 21. The IC package of claim 20, wherein there is a through-dielectric via (TDV) in the dielectric material.

22. 1. An arrangement of IC dies, comprising: a first plurality of IC dies having IC dies with IP cores; a second plurality of IC dies having IC dies, each of the IC dies of the second plurality of IC dies having at least one circuit conductively coupled to one or more of the IP cores; a third plurality of IC dies having an IC die, each of the IC dies of the third plurality of IC dies having at least one conductive path between the second plurality of IC dies with an interconnect having a pitch of less than 10 micrometers; having the first plurality of IC dies are in a first layer; the second plurality of IC dies are in a second layer; the second layer is between the first layer and the third layer; the third plurality of IC dies are in the third layer; arrangement body.

23. 23. The arrangement of claim 22, wherein the second plurality of IC dies has a shared cache accessible by the first plurality of IC dies.

24. 24. The arrangement of claim 22 or 23, wherein the first plurality of IC dies are coupled to the second plurality of IC dies with interconnects having a pitch of less than 10 micrometers between adjacent ones of the interconnects.

25. 24. The arrangement of claim 22 or 23, wherein the second plurality of IC dies are in a regular array of rows and columns.

Citation Information

Patent Citations

  • Spin transistor, integrated circuit, and magnetic memory

    JP2010080536A

  • Small Electronic Assemblies

    JP2021526309A

  • Multi-Dimensional Integrated Circuit Structures and Methods of Forming the Same

    US20130187292A1

  • Direct-bonded native interconnects and active base die

    US20180102251A1

  • Distributed semiconductor die and package architecture

    US20190221556A1