PACKAGED DEVICE WITH CHIPLETS HAVING MEMORY RESOURCES - Patent application
By integrating memory chiplets with IC chips and utilizing non-volatile memory for processor state storage, the solution addresses data locality and power efficiency issues in IC chips, improving performance and adaptability across various applications.
Patent Information
- Application Number
- JP2020108363
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-09-27
- Filing Date
- 2020-06-24
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2040-06-24
AI Technical Summary
Existing integrated circuit (IC) chips face manufacturing limitations and inefficiencies in integrating cache memory with processor cores, leading to increased routing latencies and power consumption, particularly in large core count central processing units or graphical processing units, which affect data locality and energy efficiency.
The integration of memory chiplets with IC chips, where memory resources are accessible by processor cores through specialized hardware interfaces, allowing for improved data locality and reduced power consumption by storing processor core state in non-volatile memory (NVM) chiplets, facilitating faster sleep/wake cycles and modularity across different cache memory technologies.
This approach enhances data locality and power efficiency, particularly in mobile applications, by reducing energy overhead and enabling faster processor state transitions, while allowing compatibility with evolving cache memory technologies and versatility across different product types.
Smart Images

Figure 0007586658000001 
Figure 0007586658000002 
Figure 0007586658000003
Abstract
Description
[Technical field]
[0001] The present disclosure relates generally to integrated circuits, and more particularly, but not exclusively, to memory arrays integrated on a chip in a packaged device. [Background technology]
[0002] The term "data locality" refers to the physical proximity of a memory resource for storing data for computing circuitry that accesses the data to or from the memory resource. Data locality is important in many graphics and other computing applications. Typically, relatively close locality of data and computing circuitry is associated with improved processor execution speed and overall throughput. As a result, close data locality often contributes to energy efficiency for data load and / or data store operations by a processor.
[0003] A hierarchical cache system is one example of a technique that provides improved data locality. However, the integration of one or more processor cores and cache memory within a monolithic integrated circuit (IC) chip has various manufacturing and performance limitations. Such limitations include the difficulties associated with manufacturing large IC chips, inefficiencies in the associated circuit logic, and increased routing latency and / or power consumption, especially for central processing units or graphical processing units with large core counts.
[0004] As next-generation semiconductor manufacturing continues to scale in size, operating speed, and power efficiency, the demand for solutions to improve data locality for use by one or more processor cores is expected to increase. [Brief description of the drawings]
[0005] Various embodiments of the present invention are illustrated by way of example, and not by way of limitation, in the accompanying figures. [Figure 1A]FIG. 1A is a cross-sectional side view illustrating elements of a system for providing access to memory of a chiplet, according to one embodiment. [Figure 1B] FIG. 1B is a functional block diagram illustrating elements of a system for providing memory functionality in chiplets, according to one embodiment. [Diagram 2] FIG. 2 is a flow chart illustrating elements of a method for providing memory functionality in a chiplet, according to one embodiment. [Diagram 3] FIG. 3 is a functional block diagram illustrating elements of a multi-core system including memory chiplets, according to one embodiment. [Figure 4] FIG. 4 is a layout diagram illustrating elements of an integrated circuit chip configured to access memory chiplets, according to one embodiment. [Figure 5A] FIG. 5A is a layout diagram illustrating elements of an integrated circuit chip configured to access memory chiplets, according to one embodiment. [Figure 5B] FIG. 5B is a functional block diagram illustrating elements of a memory chiplet, according to one embodiment. [Figure 6A] FIG. 6A is a cross-sectional view of a respective packaged device according to a corresponding embodiment. [Figure 6B] FIG. 6B is a cross-sectional view of a respective packaged device according to a corresponding embodiment. [Figure 7A] FIG. 7A is a side cross-sectional view of each of the respective packaged devices according to a corresponding embodiment. [Figure 7B] FIG. 7B is a side cross-sectional view of each of the respective packaged devices according to a corresponding embodiment. [Figure 8] FIG. 8 is a functional block diagram illustrating a computing device according to one embodiment. [Figure 9] FIG. 9 is a functional block diagram illustrating an exemplary computer system, according to one embodiment. [Figure 10] FIG. 10 is a cross-sectional view of an interposer implementing one or more embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0006] In the following description, numerous details are set forth to provide a more detailed explanation of the embodiments of the present disclosure. However, it will be apparent to those skilled in the art that the embodiments of the present disclosure may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form, rather than in detail, in order to avoid obscuring the embodiments of the present disclosure.
[0007] It should be noted that in the corresponding drawings of the embodiments, signals are represented by lines. Some lines may be thicker to indicate more constituent signal paths and may have arrows at one or more ends to indicate the direction of information flow. Such designations are not intended to be limiting. Rather, lines are used in connection with one or more exemplary embodiments to facilitate understanding of a circuit or logic unit. Any signal represented as dictated by design necessity or choice may in fact include one or more signals that may go in either direction and be implemented with any suitable type of signaling.
[0008] In the specification and claims, the term "connected" means a direct connection, e.g., an electrical, mechanical, or magnetic connection, between the things that are connected, without any intermediate device. The term "coupled" means a direct or indirect connection, e.g., a direct electrical, magnetic, electromagnetic connection, between the things that are connected, or things that are indirectly connected through one or more passive or active intermediate devices. The term "circuit" or "module" refers to one or more passive and / or active components arranged to cooperate with each other to provide a desired function. The term "signal" refers to at least one current signal, voltage signal, magnetic signal, or data / clock signal. The meanings of "a", "an" and "the" include plural references. The meaning of "in" includes "in" and "on."
[0009] The term "device" can generally refer to an apparatus according to the context of the use of the term. For example, a device can refer to a stack of layers or structures, a single structure or layer, a connection of various structures, with active and / or passive elements, etc. In general, a device is a three-dimensional structure having a plane along the xy direction and a height along the z direction in an xyz Cartesian coordinate system. The plane of a device can also be the plane of an apparatus that constitutes the device.
[0010] The term "scaling" generally refers to converting a design (schematic and layout) from one process technology to another, where the layout area is reduced. The term "scaling" also generally refers to downsizing layouts and devices within the same technology node. The term "scaling" also refers to adjusting (e.g., slowing down or speeding up, i.e., scaling down or scaling up, respectively) signal frequency relative to other parameters, such as power supply levels.
[0011] The terms "substantially," "close," "approximately," "near," and "about," etc., generally mean within + / - 10% of a target value. For example, unless otherwise specified in the express context of their use, the terms "substantially equal," "about equal," and "approximately equal" mean that there is only incidental variation between those so described. In the art, such variation is typically no more than + / - 10% of a given target value.
[0012] The terms so used should be understood to be interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in other orientations than those shown or described herein, for example.
[0013] The use of ordinal adjectives "first," "second," "third," etc. to describe a common object, unless otherwise specified, merely indicates that different instances of similar objects are being referred to and does not imply that the objects so described must be in a given sequence, temporally, spatially, in ranking, or otherwise.
[0014] For purposes of this disclosure, the phrases "A and / or BE" and "A or BE" mean (A), (BE), or (A and BE). For purposes of this disclosure, the phrase "A, BE, and / or C" means (A), (BE), (C), (A and BE), (A and C), (BE and C), or (A, BE, and C).
[0015] In the specification and claims, the terms "left", "right", "front", "back", "top", "bottom", "up", "down" and the like, when present, are used for descriptive purposes and not necessarily to describe a permanent relative position. For example, as used herein, the terms "over", "under", "front side", "back side", "top", "bottom", "over", "under" and "on" refer to the relative position of one component, structure or material with respect to another component, structure or material within a device where such physical relationship is noted. These terms are used herein for descriptive purposes only and are primarily used within the context of the z-axis of the device and may therefore relate to the orientation of the device. Thus, if the device is oriented in reverse relative to the context of the figures provided, a first material that is "above" a second material in the context of the figures provided herein may be "below" the second material. In the context of materials, one material that is disposed over or under another material may be in direct contact or may have one or more intervening materials. Additionally, one material that is disposed between two materials may be in direct contact with the two layers or may have one or more intervening layers. In contrast, a first material "on" a second material is in direct contact with the second material. Similar distinctions are made in the context of component assemblies.
[0016] The term "between" can be used in the context of the z-axis, x-axis, or y-axis of a device. A material that is between two other materials can be in contact with one or both of those materials, or can be separated from both of the two other materials by one or more intervening materials. Thus, a material "between" two other materials can be in contact with either of the two other materials, or can be coupled to the two other materials via an intervening material. A device that is between two other devices can be directly connected to one or both of those devices, or can be separated from both of the other two devices by one or more intervening devices.
[0017] Throughout this specification and in the claims, lists of items joined by the terms "at least one" or "one or more" can mean any combination of the listed terms. For example, the phrase "at least one of A, BE, or C" can mean A, BE, C, A and BE, A and C, BE and C, or A, BE and C. It is noted that elements of a figure having the same reference numbers (or names) as elements of other figures can operate or function in a manner similar to that described, but is not limited to such.
[0018] Additionally, the various elements of combinatorial and sequential logic described in this disclosure may relate to both physical structures (such as AND gates, OR gates, or XOR gates) or synthesized or otherwise optimized collections of devices that implement logic structures that are the Boolean equivalent of the logic described.
[0019] It is pointed out that elements of a figure having the same reference number (or name) as elements of another figure may operate or function in a similar manner as described, but is not limited to such.
[0020] The embodiments described herein provide various techniques and mechanisms for a packaged device to include an IC chip and a chiplet coupled thereto, where the memory resources of the chiplet are accessible by one or more processor cores of the IC chip. In some embodiments, the hardware interface of such a packaged device includes one or more conductive contacts on one side of the chiplet, e.g., one or more other conductive contacts of the same hardware interface are electrically interconnected to the IC chip via respective paths that are independent of the chiplet. As used herein, a "host chip" refers to an IC chip that includes one or more processor cores, and a "chiplet" refers to a relatively small IC die structure (e.g.) that extends along only a portion of such a host chip.
[0021] In some embodiments, one or more chiplets are variously coupled to a host chip, with each of the one or more chiplets including a respective cache (LLC), including, for example, a last level cache (LLC), accessible to one or more cores of the host chip. Such embodiments variously facilitate relatively dense and closely localized data caching with respect to processor circuitry, as compared to conventional architectures. Additionally or alternatively, such embodiments enable the adaptation of legacy host chip technology for use with newer cache memory technology, even though the cache memory technology may undergo additional development before being fully integrated into one or more applications.
[0022] Some embodiments provide improved power efficiency for workloads, such as various mobile applications, where one or more cores are often placed in a low power state in the absence of user input. Previously, such low power states are achieved by storing some processor core states in an SRAM cache located on the host die, e.g., the host die's CPU cores are switched off while the SRAM remains powered. To further reduce power consumption in such applications, some embodiments variously store the processor core states in a chip's NVM packaged with the host chip. In providing such an NVM chiplet, some embodiments facilitate reduced energy overhead and / or enable faster sleep / wake cycles.
[0023] Additionally or alternatively, some embodiments variously promote modularity by allowing the use of one type of host die with any of various types of memory chiplets, e.g., allowing diversity between different stock keeping units (SKUs) without requiring tape from different masks.
[0024] 1A illustrates features of a system 100 for providing efficient access to memory resources, according to one embodiment. System 100 is an example of an embodiment in which a packaged device includes chiplets and an IC chip, each including one or more processor cores and memory accessible to the one or more processor cores. The hardware interface of one such embodiment includes conductive contacts on a side of the chiplet and additional conductive contacts variously offset from the side of the chiplet.
[0025] As shown in FIG. 1A, the system 100 includes a packaged device 110 and a substrate 134 connected to the packaged device 110 via a hardware interface 132. The host chip 112 includes one or more processor cores each acting as a consumer of memory resources, and the chiplets 124 include one or more memory arrays coupled such that each is accessible by a respective processor core of the host chip 112. In this particular context, "memory," "memory array," "memory resource," and related terms refer herein to either cache memory or non-cache memory (e.g., system memory), unless otherwise indicated. Similarly, a "memory controller," unless otherwise indicated, refers to a controller circuit that provides access to one of a cache memory or a non-cache memory. For example, a "cache controller" is used herein more specifically to relate to a memory controller that provides access to a cache memory.
[0026] Substrate 134 comprises one or more electrical interconnects for facilitating communication between packaged device 110 and one or more other devices (not shown) included in or coupled to system 100. In one embodiment, substrate 134 is (or includes) a package substrate, an interposer, or any of a variety of other structures suitable for communicating, for example, one or more signals and / or one or more voltages. In some alternative embodiments, system 100 omits substrate 134, e.g., the embodiment is variously implemented with only the structure of packaged device 110.
[0027] In the illustrated exemplary embodiment, host chip 112 is coupled to chiplet 124 via other hardware interface 120, with conductive contacts (e.g., contacts including copper pads, solder interconnects, etc.) of chiplet 124 disposed on a face 126 of chiplet 124 that extends along and overlaps a first surface region of host chip 112. Such first surface region is to be distinguished from, for example, one or more other surface regions (including, for example, the illustrated second surface region 118) that are not overlapped by face 126.
[0028] Hardware interface 132, in contrast, comprises at least some conductive contacts disposed on another surface 130 of chiplet 124, opposite surface 126. In some embodiments, hardware interface 132 extends beyond one or more edges of chiplet 124, e.g., other contacts of hardware interface 132 are variously disposed along one or more portions of packaged device 110 (e.g., including illustrated exemplary portion 122), each extending along a respective sidewall of chiplet 124 between surfaces 126, 130. For example, some or all of such other contacts overlap second surface region 118, e.g., second surface region 118 of host chip 112 is overlapped by only hardware interface 132 of chiplet 124 and hardware interface 132.
[0029] In various embodiments, a first minimum metallization feature pitch of hardware interface 120 is smaller than a second minimum metallization feature pitch of hardware interface 132. By way of example and not limitation, hardware interface 120 exhibits a first average minimum distance between nearest conductive contacts, e.g., said first average minimum distance is smaller (e.g., at least 10% smaller) than a second average minimum distance between nearest conductive contacts of hardware interface 132. Additionally or alternatively, in some embodiments, the first average cross-sectional area of hardware interface 120 is smaller (e.g., at least 10% smaller) than a second average cross-sectional area of hardware interface 132.
[0030] The host chip 112 includes a processor core 114 that acts as a consumer of memory resources. For example, the host chip 112 executes an operating system, a binary input / output system (BIOS), and / or any of a variety of other software processes. To facilitate the execution of such software, the chiplets 124 include one or more memory arrays (e.g., including the illustrated exemplary memory array 128) that are coupled to be accessible to the processor core 114 via the hardware interface 120 (and, for example, via an interconnect structure 116 coupled between the core 114 and the hardware interface 120). In one embodiment, the memory array 128 includes static random access memory cells or dynamic random access memory cells. Additionally or alternatively, the processor core 114 is coupled to cache data to the memory array 128, for example, the processor core 114 is coupled to access a last level cache (LLC) of the memory array 128. In various other embodiments, the memory array 128 includes non-volatile memory (NVM) cells.
[0031] In some embodiments, chiplet 124 further includes a memory controller (not shown) coupled between hardware interface 120 and memory array 128, which controls memory accesses on behalf of processes running on cores 114. However, in other embodiments, at least some memory controller logic for operating memory array 128 resides on host chip 112, e.g., in a device layer of host chip 112 that also includes cores 114.
[0032] In some embodiments, host chip 112 further includes one or more other processor cores (not shown), some or all of which also have respective access to respective portions of memory array 128. However, in other embodiments, host chip 112 omits such additional cores and / or such additional cores may access memory array 128 only via cores 114 (if any). By providing memory array 128 within chiplets 124 disposed between hardware interface 132 and host chip 112, some embodiments variously facilitate improved locality of data for use by one or more cores of packaged device 110 as compared to conventional techniques and architectures. This improved data locality enables access to memory resources that is relatively more space-, time-, and / or power-efficient as compared to requiring such access to be made via substrate 134, for example. Additionally or alternatively, some embodiments provide for variations between the memory types and / or processes of the host chip and the chiplets, for example, instead of using a monolithically integrated memory on the host chip, a separate SRAM or DRAM process is provided on the chiplet.
[0033] In some embodiments, at least one contact of the hardware interface 132, including, for example, a contact on the side surface 130, is coupled (or otherwise accommodates coupling) to facilitate communication of signals or voltages to a first device layer of the host chip 112 and / or a second device layer of the chip 124 via an interconnect of the chip 124. In one such embodiment, the first device layer includes active circuit elements including elements of the core 114 and the second device layer includes active circuit elements of the memory array 128, for example, one of the first device layer or the second device layer is coupled to be at least partially powered by a supply voltage provided via an interconnect (not shown) that extends at least partially through the chiplet 124 to the hardware interface 132. Although some embodiments are not limited in this respect, such an interconnect extends to, for example, each of the hardware interfaces 120, 132.
[0034] In some embodiments, one or more contacts of the hardware interface 132 are variously coupled (or adapted to be coupled) to each other to facilitate respective signal and / or voltage communication between the host chip 112 and the hardware interface 132, independent of the chiplet 124. In one such embodiment, one of the portions 122 has an interconnect structure (not shown) formed therein that extends to a contact of the hardware interface 132 offset from the face 130 of the chiplet 124. Such an interconnect structure is coupled to other conductive contacts (not shown), for example, at the surface region 118 of the host chip 112. By way of example and not limitation, in some embodiments, a given one of the portions 122 includes a copper pillar structure (not shown) that variously extends from the surface region 118 to the hardware interface 132, and a mold compound, interlayer dielectric, or other such insulating material of the portion 122 extends around (and, for example, between) some of said copper pillar structures. Alternatively or additionally, a given one of portions 122 has patterned metallization layers that are variously coupled by via structures to each other, to hardware interface 132, and to contacts (not shown) in surface region 118. Interlayer dielectric material, for example, provides electrical insulation between such patterned metallization layers. In some embodiments, molding compound, interlayer dielectric (ILD), and / or other such insulating materials of portions 122 should be distinguished from underfill (if any), for example, that is deposited under and / or around packaged device 110 and that facilitates bonding between packaged device 110 and substrate 134.
[0035] 1B illustrates features of a system 150 for efficiently accessing memory resources of a chiplet, according to one embodiment. System 150 illustrates an embodiment in which a packaged device includes a host chip and a chiplet coupled thereto, where the memory array of the chip is accessible by a processor core of the host chip. System 150 includes features of system 100, for example.
[0036] 1B, system 150 includes cache controller 164 and memory device 180 coupled thereto via interconnect 194. Cache controller 164 is further coupled to processor core 160 of system 150 via interconnect 192, which in some embodiments is a network-on-chip (NoC) interconnect. In the illustrated exemplary embodiment, execution unit 162 of core 160 comprises circuitry for executing an operating system, a binary input / output system (BIOS), and / or any of a variety of other software processes. To aid in such execution, core 160 further includes one or more local caches (e.g., including the illustrated exemplary L1 and L2 caches 161), circuitry for requesting cached data, and circuitry for pre-populating the cache with data predicted to be used in the future. The functionality of cache controller 164 may include, for example, one or more operations adapted from conventional data caching techniques, for example, to manage cache entries and / or support a cache coherence scheme.
[0037] Memory device 180 may include any of a variety of types of memory technologies, e.g., having multiple rows of memory cells where data is accessible via word lines or equivalent. In one embodiment, memory device 180 may include dynamic random access memory technology. Memory device 180 may include one or more memory arrays 186, e.g., including one or more logical and / or physical groupings of memory. An example of such a grouping of memory is illustrated by banks 187 of memory resources, e.g., each including a respective array of storage elements arranged in rows and columns. Memory device 180 may include access logic 184 to at least partially facilitate access to one or more memory arrays 186, e.g., where such access is provided to service one or more commands from cache controller 164. In one embodiment, access logic 184 may include or operate in conjunction with logic of memory device 180 that provides resource access in accordance with (e.g.) conventional techniques.
[0038] Cache controller 164 sends commands or instructions to memory device 180 on one or more buses, such as a command / address (CA) bus, not shown, of interconnect 194. Such commands are interpreted by memory device 180, including memory device 180 decoding command information to perform various access functions within the memory and / or decoding address information with column logic and / or row logic. For example, such logic may access a particular location in one of multiple banks 187 with a combination of a column address strobe or signal (CAS) and a row address strobe or signal (RAS). In some embodiments, the columns of memory are implemented according to known memory architectures or derivatives thereof. For example, a row of a given bank of multiple banks 187 includes one or more addressable columns of memory cells, as identified by a CAS generated by column logic of memory device 180. The rows are variously addressable via RAS, respectively, generated by row logic of memory device 180.
[0039] In one embodiment, access to one or more memory arrays 186 is for writing exchanged data and / or reading data to be exchanged via a data bus coupled to IO circuits 182 of memory device 180. For example, data bus signal lines (not shown) of interconnect 194 couple IO circuits 174 of cache controller 164 to IO circuits 182 and / or one or more other memory devices (not shown). In one embodiment, cache controller 164 includes command logic 172, e.g., including any of a variety of hardware logic and / or executing software logic, for sending commands via a CA bus of interconnect 194. Command logic 172 includes or is coupled to memory controller logic that performs operations to generate, transmit, or identify commands sent (in some embodiments) in accordance with one or more conventional techniques.
[0040] In various embodiments, some or all of system 150 is realized by a package device comprising one or more host chips and one or more chiplets each variously coupled to a respective one of said one or more host chips, one such host chip including core 160, the chiplet coupled thereto including one or more memory resources, e.g., memory resources including some or all of one or more memory arrays 186 (and, e.g., access logic 184). By way of example and not limitation, in some embodiments, core 160 corresponds functionally to core 114, e.g., some or all of bank 187 corresponds functionally to memory array 128. In one such embodiment, cache controller 164 is present on the chiplet, e.g., interconnect 192 includes or couples to one or more contacts of a hardware interface, such as hardware interface 120. In an alternative embodiment, at least some circuitry of cache controller 164 is present on the host chip, e.g., interconnect 194 includes or couples to one or more contacts of a hardware interface, such as hardware interface 120. By providing SRAM, DRAM, NVM, or other memory on a chip located between the packaged device's hardware interface (e.g., interface 132) and the packaged device's host chip, system 150 facilitates improved locality of data for use by one or more cores.
[0041] 2 illustrates features of a method 200 for providing access to a memory of a chiplet to a host chip, according to one embodiment. Method 200 may be implemented, for example, to provide functionality of system 100 or packaged device 150. In various embodiments, method 200 includes operations (e.g., including the illustrated example operation 205), for example, to manufacture a packaged device that includes features of packaged device 110. Additionally or alternatively, method 200 may include one or more other operations performed on such a packaged device.
[0042] For example, as shown in FIG. 2, operation 205 includes (at 210) forming a host chip, e.g., host chip 112, including a processor core. In some embodiments, the host chip includes multiple cores, e.g., a network node of the host chip includes the multiple cores and any of a variety of suitable arrangements of switches, routers, and / or bridges, e.g., facilitating communication between the cores. Operation 205 further includes forming (at 212) a chiplet comprising a memory array, e.g., including memory array 128. In an embodiment, the chiplet includes one or more arrays of volatile memory, e.g., including SRAM cells and / or DRAM cells. Alternatively, the chiplet includes one or more non-volatile memory arrays. In some implementations, forming at 210 and / or 212 includes operations adapted, e.g., from conventional semiconductor fabrication techniques.
[0043] Operation 205 further includes coupling (at 214) the host chip to the chiplet, including coupling the memory array to the processor core at a first side of the chiplet via a first hardware interface, e.g., hardware interface 120. For example, in some embodiments, the coupling at 214 includes forming a solder interconnect between the host chip and the chiplet. Alternatively or additionally, the coupling at 214 includes forming any of a variety of direct (e.g., copper-to-copper) interconnects between the host chip and the chiplet, such as by thermal compression bonding or hybrid bonding. In some embodiments, the processor core is coupled to cache data in (and / or access cached data from) the memory array via the first hardware interface. In one such embodiment, the memory array is coupled to provide a last level cache (LLC) for one or more processor cores of the host chip.
[0044] Operation 205 further includes forming (at 216) a second hardware interface having conductive contacts on a second side of the chiplet, where the second side is opposite the first side. The chiplet overlaps with the first surface region of the host chip, e.g., some contacts of the second hardware interface are on a side of the chiplet (e.g., side 130) that overlaps with the first surface region. In one such embodiment, other contacts of the second hardware interface are offset from the chiplet and overlap with the second surface region of the host chip (e.g., surface region 118), e.g., the second surface region of the host chip is overlapped by only the second hardware interface of the chiplet and the second hardware interface.
[0045] In some embodiments, forming at 216 includes depositing or otherwise constructing an interconnect structure (of portion 122) that extends variously from the host chip through one or more dielectric layers that surround or are adjacent to one or more sidewall structures of the chiplet. In one such embodiment, the chiplet is formed, at least in part, from a larger chiplet that undergoes one or more cutting, polishing, and / or other subtractive processes after being coupled to the host chip via a first hardware interface. Forming at 216 includes processes adapted from (for example) conventional mask, deposition, and etching techniques to form the conductive contacts of the hardware interface. In various embodiments, the first minimum metallization feature pitch of the first hardware interface is smaller than the second minimum metallization feature pitch of the second hardware interface.
[0046] In various embodiments, the host chip includes a network including a processor core, e.g., multiple nodes of the network are coupled to each other in an array structure having rows and columns. To facilitate communication to and / or from the network, the host chip further includes one or more physical layer (PHY) circuits, e.g., each coupled to the array structure via a respective end of one of the multiple rows or columns. In such an embodiment, the memory array of the chiplet is coupled to support at least a portion of the communication with the processor core and is independent of any PHY circuitry of the device layer that is coupled to the network via a respective end of one of the multiple rows or columns. For example, at least some other IO circuitry of the host chip is coupled between the contacts of the first hardware interface and the nodes of the array structure. In various embodiments, the host chip further includes a memory controller circuit, which is coupled between the processor core and the first hardware interface, e.g., the memory controller circuit provides the processor core with access to the memory of the chiplet.
[0047] In various embodiments, method 200 additionally or alternatively includes a process performed with the packaged device as resulting from act 205. For example, method 200 further includes (at 218) coupling a substrate, via a second substrate, to the packaged device including the host chip and the chiplets. For example, the substrate (e.g., substrate 134) is a silicon interposer, a package substrate, or a substrate of a circuit board in some embodiments. In one such embodiment, method 200 further includes (at 220) accessing the memory array with a processor core, the accessing being performed while the packaged device is coupled to the substrate.
[0048] 3 illustrates features of a system 300 for efficiently accessing memory resources of a chiplet, according to one embodiment. System 300 is an example embodiment in which a packaged device includes a host chip and one or more chiplets, each disposed under a respective region of the host chip, and in which multiple processor cores of the host chip are each configured to access memory resources of each chiplet. In various embodiments, system 300 includes features of system 100 or packaged device 150, e.g., the functionality of system 300 is provided in accordance with method 200.
[0049] As shown in FIG. 3, system 300 includes a host chip and chiplets coupled thereto (e.g., host chip 112 and chiplet 124, respectively), where the host chip includes at least some circuit resources 305. Circuit resources 305 illustrates an abstracted block-level view of selected components and functional blocks of a processing-capable chip, according to an exemplary embodiment. These components include multiple processor cores 310 that provide the primary processing operations of the host chip. While the illustrated exemplary embodiment illustrates at least three processor cores in circuit resources 305, it will be appreciated that circuit resources 305 may include any of a variety of numbers of processor cores, such as 1, 2, 4, 6, 8, 10, 12, etc. Cores 310 are variously coupled to one another (e.g., via the illustrated exemplary network-on-chip NoC 320). NoC 320 generally represents various circuits supporting communication between components within circuit resources 305, including buses, routers, and control logic, as applicable. Further details of such connections are not shown so as not to obscure the details of system 300.
[0050] In various embodiments, the host chip includes any of a variety of interfaces for communicating with components external to the host chip, such as disk drives and other input / output devices, network interfaces, BIOS / firmware, and peripheral devices, as well as other chips that may be coupled to the host chip via an interconnect between CPU sockets or other forms of interconnect used for chip-to-chip communication. For example, the interfaces for facilitating communication to external components include various physical layer (PHY) circuits, such as the illustrated exemplary PHY 322, each including a respective transmitter circuit and a respective receiver circuit. In one such embodiment, one or more of the PHYs 322 are coupled to respective PCIe interconnects, for example, PHY 322 includes one or more x16 PCIe interfaces (each having a link width of 16 lanes), one or more x8 PCIe interfaces (each having a link width of 8 lanes), etc. However, the link widths and number of such PCIe interfaces are merely exemplary and not limiting to certain embodiments. Such PCIe interfaces are used to interface with various peripheral and system components, such as, for example, PCIe expansion slots, video cards, video chips, etc.
[0051] System 300 further includes memory and memory logic resources, at least some of which reside on chiplets coupled to a host chip including circuit resources 305. In one such embodiment, such memory access resources support data caching and coherence functions, e.g., some or all of cores 310 include or have access to multiple levels of caches, respectively, with the cache closest to a given core having the smallest latency and smallest size, and caches further away having larger latency. For example, one typical configuration uses first and second level caches, commonly referred to as L1 and L2 caches. Another common configuration may further use a third level or L3 cache. In the context of packaged devices variously disclosed herein, the highest level cache is referred to as the Last Level Cache, or LLC. For example, the LLC of a given core may include an L3 type cache if L1 and L2 caches are also used, or an L2 type cache if the other cache is only an L1 cache. Of course, this can be extended to further levels of the cache, with the LLC corresponding to the last (ie, highest) level of the cache.
[0052] In an exemplary embodiment, system 300 includes first level (L1) and second level (L2) caches (represented by L1 / L2 blocks 312), each "private" to a respective one of each processor cores 310. The multiple cores 310 are variously coupled to one or more last level caches (LLCs) 330, each via a NoC 320 and a respective cache controller 332. Memory access resources further include memory controllers 334 and memory devices 336 that provide cores 310 with access to memory devices 336. Each of cache controllers 332 provides a coherence agent (or "home agent") function and / or a cache agent function, for example, managing data in a respective one of LLCs 330. In some embodiments, system 300 includes other PHY circuitry (not shown) to facilitate communication between various ones of one or more LLCs 330, cache controllers 332, memory controllers 334, and memory devices 336.
[0053] In some embodiments, memory device 336 provides system memory arranged as memory blocks that represent logical divisions of memory resources to be accessed, for example, via memory controller 334. In one such embodiment, memory device 336 comprises one or more memory modules, for example, memory controller 334 is coupled to memory device 336 via one or more DDR interfaces, such as a DDR3 interface.
[0054] In various embodiments, a packaged device of system 300 includes a host chip (including circuit resources 305) and one or more chiplets, each including respective memory access resources available to some or all of cores 310. By way of example and not limitation, one such chiplet of the packaged device includes some or all of one or more LLCs 330, e.g., the cache of such a chiplet includes SRAM cells. In one such embodiment, the same chiplet (or another chiplet of the packaged device) further includes a cache controller 332. Alternatively, the host chip includes the cache controller 332 in various embodiments.
[0055] Alternatively or additionally, the chiplets of the packaged device may include other types of memory, such as the illustrated exemplary non-volatile memory and one or more memory arrays of NVM controller 338. When providing NVM resources in the chiplets, some embodiments enable efficient storage of core state (e.g., for one or more cores 310) by the chiplets during very low power supplies, as compared to that used for traditional storage of such core state in off-package NVM. Additionally or alternatively, providing NVM resources in the chiplets enables low latency retrieval of state from the chiplets to various ones of the multiple cores 310.
[0056] 4 illustrates features of an IC chip 400 for accessing memory resources of a chiplet, according to one embodiment. IC chip 400 is an example of an embodiment in which a processor core in a network of host chips to be included in a packaged device has a path for accessing the chiplets of the packaged device. The paths extending to the cores and chiplets are independent of any PHY circuitry of the host chip coupled at the edge of the network. In various embodiments, IC chip 400 includes features of host chip 112 or circuit resources 305, e.g., IC chip 400 is coupled to operate with the memory of a chiplet in accordance with method 200.
[0057] 4, IC chip 400 includes a network of nodes 410, e.g., including at least some processor cores, and, e.g., any of various switches, bridges, or routers (see legend 405), where at least some of said nodes are coupled together in an array configuration including rows and columns. By way of example and not limitation, network 410 includes router nodes that are variously coupled together in an array of rows 430a, 430b, ..., 430m and columns 420a, 420b, ..., 420n. In one such embodiment, these routers facilitate a network-on-chip (e.g., including a mesh fabric) for switched communications between processor cores, memory controllers, caches, PCIE I / O circuitry, and / or other network agents.
[0058] In various embodiments, IC chip 400 further comprises one or more physical layer (PHY) circuits variously coupled to network 410 via a respective end of one of columns 430a, 430b, ..., 430m and / or at a respective end of one of columns 420a, 420b, ..., 420n, respectively. The one or more PHY circuits each include a respective transmitter circuit and / or receiver circuit to facilitate communication between network 410 and other circuitry (not shown) that is or is to be coupled to network 410. By way of example and not limitation, network 410 is formed within a region of IC chip 400, e.g., PHYs 450a, 450b, ..., 450n of IC chip 400 are disposed along a side of the region and variously coupled at the ends of columns 420a, 420b, ..., 420n (respectively). Alternatively or additionally, one or more PHYs (such as the illustrated example PHYs 440, 441) are each coupled at a respective end of one of the columns 430a, 430b, ..., 430m. In one embodiment, PHY circuitry 450a, 450b, ..., 450n and / or a given one of PHY circuits 440, 441 include a PHY for providing access to memory (for example), or a PHY for a high-speed IO port.
[0059] In some embodiments, IC chip 400 is coupled or adapted to enable at least some communication between processor cores of network 410 and memory arrays (not shown) of chiplets, although such communication is not necessarily all of the communication between the host chip and the chiplets, independent of any PHY circuitry of IC chip 400 that is coupled to network 410 via the respective ends of the rows or columns of the array structure. For example, one or more nodes of the array structure each have at least one additional port, in addition to those for linking to other processor cores or other routers of network 410, that may be coupled to chiplets (not shown) of a packaged device that also includes IC chip 400. By way of example and not limitation, router 461 in both row 430a and column 420a is coupled to cache controller circuitry 460 that facilitates access to memory resources of the chiplets. Additionally or alternatively, router 463 in both row 430m and column 420b is coupled to cache controller circuitry 462 that facilitates access to memory resources of the same chiplet (or alternatively, another chiplet of the packaged device). Additionally or alternatively, routers 465 in both row 430b and column 420n are coupled to cache controller circuits 464 that facilitate access to memory resources of the chiplets. In one such embodiment, micro-bumps, hybrid bond pads, and / or other vertical (z-axis) interconnect structures of IC chip 400 facilitate coupling of each of cache controller circuits 460, 462, 464 to a respective chiplet, such as chiplet 124. A given one of cache controller circuits 460, 462, 464 provides, for example, the functionality of cache controller 164 or the functionality of one of cache controllers 332.
[0060] In various embodiments, cache controller circuits 460, 462, 464 each reside on IC chip 400. However, in other embodiments, a given one of cache controller circuits 460, 462, 464 includes at least some circuitry that resides on a chiplet, and a corresponding one of routers 461, 463, 465 is coupled to such circuitry via a hardware interface, such as hardware interface 120. By allowing one or more cores of network 410 to access memory resources of a chiplet without having to communicate through the edge of network 410, some embodiments provide improved data locality, for example, for cores that are offset from the edge of a mesh (or other) multiprocessor network architecture. Additionally or alternatively, all or part of such embodiments may variously provide improved cache capacity, for example, when the chiplet includes a higher density memory type, such as stacked SRAM or DRAM.
[0061] It should be understood by one of ordinary skill in the relevant art with the benefit of the disclosure herein that the particular arrangement of nodes in network 410, including, for example, the number of rows in the array structure, the number of columns in the array structure, the relative configuration of processors, switches and / or routers with respect to one another, etc., is merely exemplary and not limited to certain embodiments. For example, in various embodiments, the topology of network 410 is adapted from any of a variety of suitable ring, star, or mesh fabric (or other) multi-core network architectures, which are not limited to certain embodiments and are not detailed herein to avoid obscuring various particular features. Additionally, it should be understood that the particular number, arrangement and / or type of PHYs coupled to network 410 are also merely exemplary and not limited to certain embodiments.
[0062] 5A illustrates features of an IC chip 500 for accessing memory resources of a chiplet, according to another embodiment. The IC chip 500 is an example of an embodiment in which a host chip includes a network or one or more nodes including multiple cores, and the IO circuitry of the host chip is variously located at different locations within the network to accommodate respective communications between each core and respective memory resources of one or more chiplets. In various embodiments, the IC chip 500 includes features of the host chip 112 or circuit resources 305, e.g., the IC chip 500 is coupled to operate with the memory of the chiplet in accordance with method 200.
[0063] 5A, IC chip 500 comprises a network of nodes including routers and processor cores (see legend 505). Various ones of the nodes are coupled to each other in an array structure having multiple rows and multiple columns. By way of example and not limitation, IC chip 500 includes features of IC chip 400, e.g., rows 530a-530d of IC chip 500 functionally correspond to rows 430a, 430b,...,430m, and columns 520a-520d of IC chip 500 functionally correspond to columns 420a, 420b,...,420n. In one such embodiment, IC chip 500 further comprises IO circuitry 560, which is coupled to facilitate chiplet access by cores that are each variously linked to a respective router in one of columns 530a, 530b and one of columns 520a, 520b (e.g.). Alternatively or additionally, IO circuitry 562 of IC chip 500 is coupled to facilitate chiplet access by cores that are each variously linked to a respective router in one of columns 530a, 530b and one of columns 520c, 520d. Alternatively or additionally, IO circuitry 564 of IC chip 500 is coupled to facilitate chiplet access by cores that are each variously linked to a respective router in one of columns 530c, 530d and one of columns 520a, 520b. In various embodiments, a given one of IO circuits 560, 562, 564 further comprises a cache controller circuit, which provides, for example, the functionality of cache controller 164 or one of cache controller 332. In other embodiments, such functionality of a cache controller is instead implemented on a chiplet that is coupled to IC chip 500 via a given one of IO circuits 560, 562, 564.
[0064] 5B illustrates features of a chiplet 570 that is packaged with a host chip and provides memory resources for the host chip according to one embodiment. Chiplet 570 is an example of an embodiment in which multiple IO circuits are each operable to provide access to one or more respective memory banks, e.g., each IO circuit is coupled to a different respective network node of the host chip. In one such embodiment, chiplet 570 provides memory resource access to the host chip according to method 200, e.g., chiplet 570 includes features of chiplet 124, for example.
[0065] As shown in FIG. 5B, chiplet 570 includes various IO circuits 580-583, e.g., each providing a respective function of IO circuit 182. IO circuits 580-583 are each coupled to provide access to one or more memory banks of a respective one of chiplet 570. In the illustrated exemplary embodiment, bank 590 of chiplet 570 is accessible via IO circuit 580, e.g., bank 591 is accessible via IO circuit 581, bank 592 is accessible via IO circuit 582, and / or bank 593 is accessible via IO circuit 583. In some embodiments, chiplet 570 includes, e.g., 2N IO circuits (where N is a positive integer), each supporting communication with a different respective core of a host chip.
[0066] In an exemplary scenario according to one embodiment, chiplet 570 supports coupling to IC chip 500, e.g., IO circuitry 560 each includes IO circuitry configured to couple to a respective one of IO circuits 581-583 (for example) via a hardware interface, such as hardware interface 120. In other such embodiments, chiplet 570 supports coupling to IC chip 400, e.g., cache controller circuits 460, 462, 464 each include IO circuitry for coupling to a different respective one of IO circuits 581-583.
[0067] In the illustrated exemplary embodiment, IO circuits 580-583 are variously positioned around the area including memory banks 590-593, e.g., IO circuits 580-583 are not between any two of memory banks 590-593. In other embodiments, memory banks 590-593 are variously positioned around the area including IO circuits 580-583, e.g., memory banks 590-593 are not between any two of IO circuits 580-583. In various embodiments, conductive contacts coupling a given one of IO circuits 580-583 to a corresponding IO circuit of the host chip (e.g., IO circuit of IO circuit 560) are positioned over the corresponding one or more memory banks. In one such embodiment, the conductive contacts, including, for example, microbumps or hybrid bonding pads, accommodate the required number of electrical connections and / or the required physical pitch to access the one or more memory banks.
[0068] 6A illustrates features of a package device 600 for providing memory access to a host chip and chiplets, according to one embodiment. Package device 600 is an example of an embodiment in which chiplets disposed between a host chip and a hardware interface include memory resources accessible to one or more processor cores of the host chip. The functionality of package device 600 may be provided, for example, in accordance with method 200.
[0069] 6A , packaged device 600 includes a host chip 610 and a chiplet 630 coupled thereto via a hardware interface 620. Host chip 610 includes one or more processor cores, and memory resources of chiplets 630 are coupled to be variously accessible by some or all of such one or more processor cores. In such an embodiment, host chip 610 provides the functionality of one of chips 112, 400, 500, e.g., chiplet 630 provides the functionality of one of chiplets 124, 570, and hardware interface 620 corresponds functionally to hardware interface 120.
[0070] In the illustrated exemplary embodiment, the host chip 610 has a semiconductor substrate 614 and a device layer 612 including transistors and / or other circuit elements variously fabricated on the substrate 614. The device layer 612 includes one or more processor cores, such as, for example, cores 310, cores of network 410, or any of a variety of other suitable arrangements of cores. The host chip 610 further includes one or more metallization layers (including, for example, the illustrated exemplary metallization layer 616) that variously interconnect the circuit structures of the device layer 612 with one another, along with hardware interfaces 620 and / or hardware interfaces 650 of the packaged device 600. The hardware interfaces 650 may, for example, correspond functionally to the hardware interfaces 132.
[0071] Chiplet 630 similarly includes a semiconductor substrate 634 and a device layer 632 including transistors and / or other circuit elements variously fabricated on substrate 634. However, device layer 632 includes transistors and / or other circuit elements of one or more memory arrays, including, for example, banks 187, banks 590-593, or any of a variety of other suitable arrangements of one or more memory resources. Chiplet 630 further includes one or more metallization layers (including, for example, the illustrated exemplary metallization layer 636) that, together with hardware interface 620 and / or hardware interface 650, variously interconnect the circuit structures of device layer 632 to one another. In various embodiments, hardware interface 650 includes one or more first conductive contacts, including, for example, the illustrated exemplary contact 652, that are on one side of chiplet 630 (other than the side of chiplet 630 on which hardware interface 620 is disposed). In one such embodiment, hardware interface 650 further comprises one or more second conductive contacts (eg, including the illustrated example contact 654 ), which are offset from chiplet 630 .
[0072] Additionally or alternatively, some or all of the one or more first contacts are variously coupled (or adapted to be coupled) to facilitate respective signal and / or voltage communication between the hardware interface 650 and the metallization layer 636, e.g., such communication via the device layer 632 and the substrate 634. By way of example and not limitation, the chiplet 630 further comprises at least some interconnect structures (such as the illustrated interconnects 638) that extend to respective ones of the one or more first contacts of the hardware interface 650, through the substrate 634 to the device layer 632, and in some embodiments through the device layer 632. In one such embodiment, some or all of the interconnects 638 are each variously coupled to respective patterned conductors of the metallization layer 636 to facilitate signal communication or power delivery (via the hardware interface 620) with the circuitry of the host chip 610 and / or the device layer 632.
[0073] Additionally or alternatively, some or all of the one or more second contacts are variously coupled (or otherwise adapted to be coupled) to facilitate respective signal and / or voltage communication between the hardware interface 650 and the host chip 610, where such communication is independent of the chiplet 630. By way of example and not limitation, the packaged device 600 further comprises an insulator 640, including, for example, a molding compound or an interlayer dielectric (ILD) material, having at least some interconnect structures (such as the illustrated interconnects 642) formed therein, each of which variously extends to a respective one of the one or more second contacts of the hardware interface 650 and to a respective conductive contact at the surface of the metallization layer 616. For example, the interconnects 642 comprise any of a variety of other conductors and / or copper pillars suitable for communicating signals and / or voltages through the insulator 640. In some embodiments, the mold compound, ILD, and / or any other such materials, such as insulator 640, should be distinguished from, for example, underfill (if any), which is deposited under and / or around hardware interface 650.
[0074] In the illustrated exemplary embodiment, hardware interface 620 comprises solder interconnects that variously couple respective contacts of host chip 610 and chiplets 630 to one another. For example, FIG. 6B shows a detailed view of packaged device 660, according to one embodiment such as packaged device 600. The host chip of packaged device 660 includes a device layer 672 and a substrate 674 (e.g., device layer 612 and substrate 614, respectively). The host chip is coupled via hardware interface 690 to the chiplets of packaged device 660, including a device layer 682 and a substrate 684 (e.g., device layer 632 and substrate 634, respectively).
[0075] For example, the host chip's metallization layers 676a-676e (e.g., metallization layer 616) have interconnect structures formed therein that variously electrically couple the device layer 672 to the hardware interface 690. In such an embodiment, the chiplet's metallization layers 686a-686e (e.g., metallization layer 636) have interconnect structures formed therein that variously electrically couple the device layer 682 to the hardware interface 690. The hardware interface 690 comprises conductive pads (e.g., including the illustrated exemplary pad 694) variously disposed on respective sides of the host chip or one of the chiplets. Solder interconnects (e.g., including the illustrated microbumps 692) provide various couplings between the host chip and the chiplets via the pads 694.
[0076] By way of example and not limitation, one or more interconnects of a chiplet may each extend from another hardware interface (not shown) of packaged device 660, such as hardware interface 132, through substrate 684 and device layer 682, and back to circuit elements of device layer 682 via interconnect structures in metallization layers 686a-686e. Additionally or alternatively, one or more interconnects of a chiplet may each extend from the other hardware interface through substrate 684, device layer 682, and metallization layers 686a-686e to hardware interface 690, for example to deliver power to device layer 672, via metallization layers 676a-676e.
[0077] In various embodiments, packaged device 600 comprises multiple device layers, each of which is a respective chiplet coupled to host chip 610, e.g., chiplet 630, which further includes one or more other device layers (not shown) that are vertically offset from one another and interconnected, e.g., by including through-hole vias, monolithic interlayer vias, etc. In other such embodiments, chiplet 630 is one of a stack of multiple chiplets coupled to host chip 610, e.g., the chips of the stack are variously coupled to one another by solder bumps or hybrid bond pads, and hardware interface 650 includes contacts at one such chiplet of the stack.
[0078] 7A illustrates features of a packaging device 700 for accessing memory resources of a chiplet, according to another embodiment. The packaging device 700 is an example of an embodiment in which a chiplet is bonded to a host chip to make memory resources available to the host chip, e.g., via any of a variety of wafer-level (or chip-level) direct bonding techniques, including thermo-compression bonding, hybrid bonding, etc. The functionality of the packaging device 700 may be provided, for example, in accordance with method 200.
[0079] As shown in FIG. 7A, packaged device 700 includes a host chip 710 and a chiplet 730, which may, for example, functionally correspond to host chip 610 and chiplet 630, respectively. Host chip 710 and chiplet 730 are coupled to each other via a hardware interface 720, and another hardware interface 750 facilitates coupling of packaged device 700 to an interposer, packaged substrate, circuit board, etc. Hardware interface 720, hardware interface 750 are on opposite respective sides of chiplet 730, for example, hardware interfaces 720, 750 functionally correspond to hardware interfaces 620, 650, respectively. In one such embodiment, hardware interface 750 includes first conductive contacts (e.g., including contact 752) on a bottom surface of chiplet 730. In contrast, second conductive contacts (e.g., including contact 754) of hardware interface 750 are offset from chiplet 730.
[0080] In the illustrated exemplary embodiment, host chip 710 includes a device layer 712, a semiconductor substrate 714, and a metallization layer 716, which correspond functionally to, for example, device layer 612, substrate 614, and metallization layer 616. Chiplet 730 includes a device layer 732, a semiconductor substrate 734, and a metallization layer 736, which correspond functionally to, for example, device layer 632, substrate 634, and metallization layer 636. A first conductive contact (e.g., contact 752) of hardware interface 750 variously facilitates communication of one or more signals and / or one or more voltages with each one of device layer 732 or device layer 712. For example, interconnect 738 (e.g., providing the functionality of interconnect 638) variously extends from the first contact through substrate 734 to at least device layer 732 (and in some embodiments, through device layer 732). In one such embodiment, the second conductive contact (e.g., contact 754) variously facilitates communication of one or more signals and / or one or more voltages via interconnect 742 extending through insulator 740, e.g., such communication is independent of chiplet 730.
[0081] In the illustrated exemplary embodiment, hardware interface 720 includes a direct-bond interconnect structure that couples host chip 710 and chiplets 730 to one another. For example, FIG. 7B illustrates a detailed view of packaged device 760 including one or more interface structures, such as those of packaged device 700, according to one embodiment. As shown in FIG. 7B, the host chip of packaged device 760 includes a device layer 772 and a substrate 774 (e.g., device layer 712 and substrate 714, respectively). The host chip is coupled via hardware interface 790 to the chiplets of packaged device 760 including a device layer 782 and a substrate 784 (e.g., device layer 732 and substrate 734, respectively).
[0082] The host chip metallization layers 776a-776e have interconnect structures formed therein that variously electrically couple the device layer 772 to a hardware interface 790. In such an embodiment, the chiplet metallization layers 786a-786e have interconnect structures formed therein that variously electrically couple the device layer 782 to a hardware interface 790. At the hardware interface 790, the host die conductive pads 792 are each bonded to a respective one of the chiplet conductive pads 794. In one such embodiment, one or more interconnects of the chiplet extend from other hardware interfaces (not shown) of the packaged device 760, through the substrate 784 and the device layer 782, respectively, and back to the circuit elements of the device layer 782 via the interconnect structures of the metallization layers 786a-786e. Additionally or alternatively, one or more interconnects of the chiplet may extend from each of the other hardware interfaces through substrate 784, device layer 782, and metallization layers 786a-786e to hardware interface 790, for example to deliver power to device layer 772 via metallization layers 776a-776e.
[0083] In various embodiments, packaged device 700 includes multiple device layers, each of which is a respective chiplet coupled to host chip 710, e.g., chiplet 730 further includes one or more other device layers (not shown), which are vertically offset from one another and interconnected by, e.g., including through-hole vias, monolithic interlayer vias, etc. In other such embodiments, chiplet 730 is one of a stack of chips coupled to host chip 710, e.g., the chips in the stack are variously coupled to one another by solder bumps or hybrid bond pads, and hardware interface 750 includes contacts in one chiplet of the stack.
[0084] 8 illustrates a computing device 800 according to one embodiment. The computing device 800 houses a board 802. The board 802 may include a number of components, including but not limited to a processor 804 and at least one communications chip 806. The processor 804 is physically and electrically coupled to the board 802. In some implementations, the at least one communications chip 806 is also physically and electrically coupled to the board 802. In a further implementation, the communications chip 806 is part of the processor 804.
[0085] Depending on its application, computing device 800 may include other components that may or may not be physically and electrically coupled to board 802. These other components include, but are not limited to, volatile memory (e.g., DRAM), non-volatile memory (e.g., ROM), flash memory, graphics processors, digital signal processors, cryptographic processors, chipsets, antennas, displays, touch screen displays, touch screen controllers, batteries, audio codecs, video codecs, power amplifiers, Global Positioning System (GPS) devices, compasses, accelerometers, gyroscopes, speakers, cameras, and mass storage devices (such as hard disk drives, compact discs (CDs), digital versatile discs (DVDs), etc.).
[0086] The communications chip 806 enables wireless communication to transfer data to and from the computing device 800. The term "wireless" and its derivatives may be used to describe circuits, devices, systems, methods, techniques, communications channels, etc. that can communicate data through the use of modulated electromagnetic radiation over a non-solid medium. The term does not imply that the associated devices do not include wires, although in some embodiments the associated devices may not include wires. The communications chip 806 may implement any of the wireless standards or protocols, including but not limited to Wi-Fi (IEEE 802.11 family), WiMAX (IEEE 802.16 family), IEEE 802.20, Long Term Evolution (LTE), Ev-DO, HSPA+, HSDPA+, HSUPA+, EDGE, GSM, GPRS, CDMA, TDMA, DECT, Bluetooth, derivatives thereof, and any other wireless protocols designated as 3G, 4G, 5G and beyond. Computing device 800 may include multiple communications chips 806. For example, a first communications chip 806 may be dedicated to shorter range wireless communications such as Wi-Fi and Bluetooth, and a second communications chip 806 may be dedicated to longer range wireless communications such as GPS, EDGE, GPRS, CDMA, WiMAX, LTE, Ev-DO, etc.
[0087] The processor 804 of the computing device 800 includes an integrated circuit die packaged within the processor 804. The term "processor" may refer to a device or part of a device that processes electronic data from registers and / or memory and converts the electronic data into other electronic data that can be stored in registers and / or memory. The communications chip 806 also includes an integrated circuit die packaged within the communications chip 806.
[0088] In various implementations, computing device 800 may be a laptop, netbook, notebook, ultrabook, smartphone, tablet, personal digital assistant (PDA), ultra-mobile PC, mobile phone, desktop computer, server, printer, scanner, monitor, set-top box, entertainment control unit, digital camera, portable music player, or digital video recorder, In further implementations, computing device 800 may be any other electronic device that processes data.
[0089] Some embodiments can be provided as a computer program product or software, including a machine-readable medium having stored thereon instructions that can be used to program a computer system (or other electronic device) to perform a process according to the embodiments. A machine-readable medium includes any mechanism for storing or transmitting information in a form readable by a machine (e.g., a computer). For example, a machine-readable (e.g., computer-readable) medium includes a machine (e.g., computer) readable storage medium (e.g., read-only memory (ROM), random access memory (RAM), magnetic disk storage media, optical storage media, flash memory disks, etc.), a machine (e.g., computer) readable transmission medium (electronic, optical, acoustic, or other form of propagated signal (e.g., linear signal, digital signal, etc.)), etc.
[0090] FIG. 9 illustrates a schematic diagram of an exemplary form of computer system 900 in which a set of instructions may be executed to cause the machine to perform any one or more of the methods described herein. In other embodiments, the machine may be connected (e.g., networked) to other machines in a local area network (LAN), an intranet, an extranet, or the Internet. The machine may operate in the capacity of a server or client machine in a client-server network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The machine may be a personal computer, a tablet PC, a set-top box (STB), a personal digital assistant (PDA), a mobile phone, a web appliance, a server, a network router, switch or bridge, or any machine capable of executing a set of instructions (chronologically or otherwise) that specify actions to be taken by the machine. Moreover, although only a single machine is shown, the term "machine" is intended to include any collection of machines (e.g., computers) that individually or jointly execute a set (or sets) of instructions to perform any one or more of the methods described herein.
[0091] The exemplary computer system 900 includes a processor 902, a main memory 904 (e.g., dynamic random access memory such as read only memory (ROM), flash memory, synchronous DRAM (SDRAM) or Rambus DRAM (RDRAM)), a static memory 906 (e.g., flash memory, static random access memory (SRAM), etc.), and a secondary memory 918 (e.g., a data storage device), which communicate with each other via a bus 930.
[0092] Processor 902 represents one or more general-purpose processing devices, such as a microprocessor, a central processing unit, etc. More specifically, processor 902 may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. Processor 902 may also be one or more special purpose processing devices, such as an application specific integrated circuit, a field programmable gate array (FPGA), a digital signal processor (DSP), a network processor, etc. Processor 902 is configured to execute processing logic 926 to perform the operations described herein.
[0093] The computer system 900 may further include a network interface device 908. The computer system 900 may also include a video display unit 910 (e.g., a liquid crystal display (LCD), a light emitting diode display (LED), or a cathode ray tube (CRT)), an alphanumeric input device 912 (e.g., a keyboard), a cursor control device 914 (e.g., a mouse), and a signal generation device 916 (e.g., a speaker).
[0094] The secondary memory 918 may include a machine-accessible storage medium (or, more specifically, a computer-readable storage medium) 932 on which one or more sets of instructions (e.g., software 922) that implement any one or more of the methods or functions described herein are stored. The software 922 may also reside, completely or at least partially, within the main memory 904 and / or within the processor 902 during execution by the computer system 900, the main memory 904 and / or the processor 902, which also constitute machine-readable storage media. The software 922 may further be transmitted or received over the network 920 via the network interface device 908.
[0095] While the machine-accessible storage medium 932 is shown to be a single medium in the exemplary embodiment, the term "machine-readable storage medium" should be interpreted to include a single medium or multiple media (e.g., a centralized or distributed database and / or associated caches and servers) that store one or more sets of instructions. The term "machine-readable storage medium" should also be interpreted to include a medium that can store or encode a set of instructions for execution by a machine and cause the machine to perform any of the one or more embodiments. Thus, the term "machine-readable storage medium" should be interpreted to include, but is not limited to, solid-state memories and optical and magnetic media.
[0096] FIG. 10 illustrates an interposer 1000 that includes one or more embodiments. The interposer 1000 is an intervening substrate used to bridge a first substrate 1002 to a second substrate 1004. The first substrate 1002 can be, for example, an integrated circuit die. The second substrate 1004 can be, for example, a memory module, a computer motherboard, or another integrated circuit die. In general, the purpose of the interposer 1000 is to spread connections to a wider pitch or to reroute connections to different connections. For example, the interposer 1000 can bond an integrated circuit die to a ball grid array (BGA) 1006, which can then be bonded to the second substrate 1004. In some embodiments, the first and second substrates 1002, 1004 are attached to opposite sides of the interposer 1000. In other embodiments, the first and second substrates 1002, 1004 are attached to the same side of the interposer 1000. In yet other embodiments, three or more substrates are interconnected via the interposer 1000.
[0097] The interposer 1000 may be formed from a polymeric material such as epoxy, fiberglass reinforced epoxy, ceramic material, or polyimide, etc. In further embodiments, the interposer may be formed from alternative rigid or flexible materials, which may include the same materials mentioned above, for use with semiconductor substrates such as silicon, germanium, and other III-V and IV materials.
[0098] The interposer includes metal interconnects 1008 and vias 1010, including, but not limited to, through silicon vias 1012. The interposer 1000 may further include embedded devices 1014, including both passive and active devices. Such devices include, but are not limited to, capacitors, decoupling capacitors, resistors, inductors, fuses, diodes, transformers, sensors, and electrostatic discharge (ESD) devices. More complex devices, such as radio frequency (RF) devices, power amplifiers, power management devices, antennas, arrays, sensors, and MEMS devices, may also be formed on the interposer 1000. According to some embodiments, the apparatus or processes disclosed herein may be used in the manufacture of the interposer 1000.
[0099] Techniques and architectures for providing improved data locality in packaged devices are described herein. In the above description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of certain embodiments. However, it will be apparent to one skilled in the art that certain embodiments may be practiced without these specific details. In other instances, structures and devices are shown in block diagram form in order to avoid obscuring the description.
[0100] References in the specification to "one embodiment" or "an embodiment" mean that at least one embodiment of the invention includes a particular feature, structure, or characteristic described in connection with that embodiment. The appearances of the phrase "in one embodiment" in various places in the specification do not necessarily refer to the same embodiment.
[0101] Some portions of the detailed descriptions herein are presented in terms of algorithms and symbolic representations of operations on data bits within a computer memory. These algorithmic descriptions and representations are the means used by those skilled in the computer arts to most effectively convey the substance of their work to others skilled in the art. An algorithm is here, and generally, conceived to be a self-consistent sequence of steps leading to a desired result. The steps are those requiring the physical manipulation of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic signals that can be stored, transferred, combined, compared, and otherwise manipulated. It has proven convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, or the like.
[0102] However, all of these and similar terms should be associated with the appropriate physical quantities and should be borne in mind as being primarily convenient labels to apply to these quantities. As will be apparent from the description herein, and unless otherwise indicated, it will be understood throughout the description that. Discussions using terms such as "processing" or "computing" or "calculating" or "determining" or "displaying" refer to the processing and operation of a computer system or similar electronic computing device that manipulates data represented as physical (electronic) quantities in the computer system's registers and memory and converts it into other data also represented as physical quantities in the computer system's memory or registers or other such information storage, conversion or display device.
[0103] Certain embodiments also relate to apparatus for performing the operations herein, which may be specially constructed for the required purposes or may comprise a general purpose computer selectively activated or reconfigured by a computer program stored in the computer, which may be stored on a computer readable storage medium, such as, but not limited to, a floppy disk, an optical disk, a CD-ROM, any type of disk including a magneto-optical disk, a read only memory (ROM), a dynamic RAM (DRAM), a random access memory (RAM) such as an EPROM, an EEPROM, a magnetic or optical card, or any type of media suitable for storing electronic instructions, and coupled to a computer system bus, but not limited to.
[0104] The algorithms and displays presented herein are not inherently related to any particular computer or other apparatus. Various general-purpose systems may be used with programs in accordance with the teachings herein, or it may prove convenient to construct a more specialized apparatus to perform the required method steps. The required structure for a variety of these systems will appear from the description herein. Moreover, particular embodiments have not been described with reference to any particular programming language. It will be understood that a variety of programming languages may be used to implement the teachings of the embodiments as described herein.
[0105] In addition to those described herein, various modifications can be made to the disclosed embodiments and their embodiments without departing from the scope thereof. Accordingly, the descriptions and examples herein should be construed in an illustrative rather than a limiting sense. The scope of the present invention should be evaluated solely by reference to the claims that follow.
Claims
1. a host chip having a processor core; a chiplet having a memory coupled to communicate with the processor core via a first hardware interface in a first face of the chiplet; a second hardware interface having conductive contacts on a second side of the chiplet, the second side being opposite the first side, a first minimum metallization feature pitch of the first hardware interface being smaller than a second minimum metallization feature pitch of the second hardware interface, the chiplet overlapping a first surface area of the host chip, and a second surface area of the host chip being overlapped by only the second hardware interface of the chiplet and the second hardware interface; having a first device layer of the host chip having a switch network including the processor core; the nodes of the switch network are interconnected in an array configuration having a plurality of rows and a plurality of columns; the memory is coupled to communicate with the processor core via a respective end of one of the rows or one of the columns, independent of any PHY circuitry of a device layer coupled to the switch network; Package device.
2. A host chip having a processor core; a chiplet having a memory coupled to communicate with the processor core via a first hardware interface in a first face of the chiplet; a second hardware interface having conductive contacts on a second side of the chiplet, the second side being opposite the first side, a first minimum metallization feature pitch of the first hardware interface being smaller than a second minimum metallization feature pitch of the second hardware interface, the chiplet overlapping a first surface area of the host chip, and a second surface area of the host chip being overlapped by only the second hardware interface of the chiplet and the second hardware interface; having a first device layer of the host chip including the processor core and a memory controller circuit coupled between the processor core and the memory; the memory controller circuit provides access to the memory to the processor core; Package device.
3. a first device layer of the host chip having the processor core; a second device layer of the chiplet comprising the memory; either the first device layer or the second device layer is coupled to receive power via an interconnect that extends at least partially through the chiplet to the second hardware interface.
3. The packaged device according to claim 1 or 2.
4. the interconnect extends to the first hardware interface; 4. The packaged device of claim 3.
5. the memory comprises one of a static random access memory or a dynamic random access memory; 3. The packaged device according to claim 1 or 2.
6. The memory comprises a non-volatile memory.
3. The packaged device according to claim 1 or 2.
7. the processor core is coupled to cache data in the memory; 3. The packaged device according to claim 1 or 2.
8. the processor core is coupled to access a last level cache of the memory; The packaged device of claim 7.
9. the processor core is a first processor core; the host chip further comprises a second processor core; the memory includes a first memory bank and a second memory bank; the chiplet further comprises a first input / output circuit (IO circuit) coupled between the first processor core and the first memory bank, and a second IO circuit coupled between the second processor core and the second memory bank; a first cache controller coupled between the first memory bank and the first processor core, and a second cache controller coupled between the second memory bank and the second processor core; 3. The packaged device according to claim 1 or 2.
10. the chiplet includes the first cache controller and the second cache controller; 10. The packaged device of claim 9.
11. forming a host chip having a processor core; forming a chiplet having a memory array; coupling the host chip to the chiplet, the coupling including coupling the memory array to the processor core via a first hardware interface at a first side of the chiplet; forming a second hardware interface having conductive contacts on a second side of the chiplet; A method comprising: the second surface is opposite the first surface; a first minimum metallization feature pitch of the first hardware interface is smaller than a second minimum metallization feature pitch of the second hardware interface; the chiplet overlaps a first surface region of the host chip; a second surface area of the host chip is overlapped by only the second hardware interface of the chiplet and the second hardware interface; a first device layer of the host chip having the processor core; a second device layer of the chiplet comprising a memory; either the first device layer or the second device layer is coupled to receive power via an interconnect that extends at least partially through the chiplet to the second hardware interface. method.
12. The method of claim 1, further comprising the steps of: forming a host chip having a processor core; forming a chiplet having a memory array; coupling the host chip to the chiplet, the coupling including coupling the memory array to the processor core via a first hardware interface at a first side of the chiplet; forming a second hardware interface having conductive contacts on a second side of the chiplet; A method comprising: the second surface is opposite the first surface; a first minimum metallization feature pitch of the first hardware interface is smaller than a second minimum metallization feature pitch of the second hardware interface; the chiplet overlaps a first surface region of the host chip; a second surface area of the host chip is overlapped by only the second hardware interface of the chiplet and the second hardware interface; The first device layer of the host chip comprises: a switch network including the processor core; the nodes of the switch network are interconnected in an array configuration having a plurality of rows and a plurality of columns; a memory is coupled to communicate with the processor cores independent of any PHY circuitry of a device layer that is coupled to the switch network via a respective end of one of the rows or one of the columns; method.
13. The method of claim 1, further comprising the steps of: forming a host chip having a processor core; forming a chiplet having a memory array; coupling the host chip to the chiplet, the coupling including coupling the memory array to the processor core via a first hardware interface at a first side of the chiplet; forming a second hardware interface having conductive contacts on a second side of the chiplet; A method comprising: the second surface is opposite the first surface; a first minimum metallization feature pitch of the first hardware interface is smaller than a second minimum metallization feature pitch of the second hardware interface; the chiplet overlaps a first surface region of the host chip; a second surface area of the host chip is overlapped by only the second hardware interface of the chiplet and the second hardware interface; a first device layer of the host chip including the processor core and a memory controller circuit coupled between the processor core and a memory; the memory controller circuit provides access to the memory to the processor core; method.
14. The method of claim 1, further comprising the steps of: forming a host chip having a processor core; forming a chiplet having a memory array; coupling the host chip to the chiplet, the coupling including coupling the memory array to the processor core via a first hardware interface at a first side of the chiplet; forming a second hardware interface having conductive contacts on a second side of the chiplet; A method comprising: the second surface is opposite the first surface; a first minimum metallization feature pitch of the first hardware interface is smaller than a second minimum metallization feature pitch of the second hardware interface; the chiplet overlaps a first surface region of the host chip; a second surface area of the host chip is overlapped by only the second hardware interface of the chiplet and the second hardware interface; the processor core is a first processor core; the host chip further comprises a second processor core; the memory includes a first memory bank and a second memory bank; forming the chiplet further includes forming a first input / output circuit (IO circuit) coupled between the first processor core and the first memory bank and a second IO circuit coupled between the second processor core and the second memory bank; a first cache controller coupled between the first memory bank and the first processor core, and a second cache controller coupled between the second memory bank and the second processor core; method.
15. the processor core is coupled to cache data in the memory; 15. The method according to any one of claims 11 to 14.
16. 1. A system comprising a packaged device, the packaged device comprising: a host chip having a processor core; a chiplet having a memory coupled to communicate with the processor core via a first hardware interface in a first face of the chiplet; a second hardware interface having conductive contacts on a second side of the chiplet, the second side being opposite the first side, a first minimum metallization feature pitch of the first hardware interface being smaller than a second minimum metallization feature pitch of the second hardware interface, the chiplet overlapping a first surface area of the host chip, and a second surface area of the host chip being overlapped by only the second hardware interface of the chiplet and the second hardware interface; having The system further comprises a display device coupled to the package device, the display device configured to display an image based on signals between the processor core and the memory; a first device layer of the host chip including the processor core and a memory controller circuit coupled between the processor core and the memory; the memory controller circuit provides access to the memory to the processor core; system.
17. a first device layer of the host chip having the processor core; a second device layer of the chiplet comprising the memory; either the first device layer or the second device layer is coupled to receive power via an interconnect that extends at least partially through the chiplet to the second hardware interface.
17. The system of claim 16.
18. A system comprising a package device, the package device comprising: a host chip having a processor core; a chiplet having a memory coupled to communicate with the processor core via a first hardware interface in a first face of the chiplet; a second hardware interface having conductive contacts on a second side of the chiplet, the second side being opposite the first side, a first minimum metallization feature pitch of the first hardware interface being smaller than a second minimum metallization feature pitch of the second hardware interface, the chiplet overlapping a first surface area of the host chip, and a second surface area of the host chip being overlapped by only the second hardware interface of the chiplet and the second hardware interface; having The system further comprises a display device coupled to the package device, the display device configured to display an image based on signals between the processor core and the memory; a first device layer of the host chip having the processor core; a second device layer of the chiplet comprising the memory; Either the first device layer or the second device layer is coupled to receive power via an interconnect that extends at least partially through the chiplet to the second hardware interface; the first device layer of the host chip includes a switch network including the processor core; the nodes of the switch network are interconnected in an array configuration having a plurality of rows and a plurality of columns; the memory is coupled to communicate with the processor core independently of any PHY circuitry of a device layer that is coupled to the switch network via a respective end of one of the rows or one of the columns; system.
19. the interconnect extends to the first hardware interface; 19. A system according to claim 17 or 18.
20. the memory comprises one of a static random access memory or a dynamic random access memory; 19. A system according to claim 16 or 18.
21. the processor core is a first processor core; the host chip further comprises a second processor core; the memory includes a first memory bank and a second memory bank; the chiplet further comprises a first input / output circuit (IO circuit) coupled between the first processor core and the first memory bank and a second IO circuit coupled between the second processor core and the second memory bank; a first cache controller coupled between the first memory bank and the first processor core, and a second cache controller coupled between the second memory bank and the second processor core; 19. A system according to claim 16 or 18.
22. the chiplet includes the first cache controller and the second cache controller; 22. The system of claim 21.
Citation Information
Patent Citations
Asynchronous control transfer
JP2008165793A
3D integrated circuit package with window interposer
JP2015507843A
Scalable Package Architecture and Related Techniques and Structures
JP2016526306A
Semiconductor device
JP2018074065A
Ring-in-ring configurable-capacitance stiffeners and methods of assembling the same
JP2019068052A
