Modular peripheral chips for integrated circuit devices

By modularizing peripheral IP into independent chips and employing silicon interposers and direct communication, the problem of limited shoreline quantity in FPGA devices is solved, improving the scalability and performance of the devices and enabling flexible design and programming.

CN115525592BActive Publication Date: 2026-11-17ALTERA CORP
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202211031127.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2018-12-28
Filing Date
2019-11-28
Publication Date
2026-11-17
Estimated Expiration
2039-11-28

AI Technical Summary

Technical Problem

Modern integrated circuit devices such as FPGAs have limited shoreline quantity, which restricts the types and number of peripheral IPs, affecting device performance and scalability. Design compromises lead to reduced device performance and overall scalability of programmable devices.

Method used

By modularizing peripheral IP into independent chips, independent of the design and fabrication of programmable structure dies, and using silicon interposers and spliced ​​direct communication, modular interface connections between peripheral IP chips and programmable structure dies are achieved.

Benefits of technology

It improves the scalability and performance of integrated circuit devices, reduces tape-out preparation and positioning limitations, lowers design and fabrication costs, and enables flexible customization and efficient programming of different variants.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115525592B_ABST
    Figure CN115525592B_ABST
Patent Text Reader

Abstract

The subject of this invention is "Modular Peripheral Chips for Integrated Circuit Devices". The systems or methods disclosed herein can improve the scalability (e.g., component scalability, product variant scalability) of integrated circuit systems by decomposing peripheral intellectual property (IP) circuitry into modular peripheral IP chips that can be mounted as modules. Such an integrated circuit system may include: a first die including programmable configuration circuitry; and a second die including peripheral IP chips. The peripheral IP chips can be decomposed from the programmable configuration die and can be communicatively coupled to the first die via a modular interface.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] This application is a divisional application of patent application filed on November 28, 2019, with application number 201911192670.7 and invention title "Modular peripheral chip for integrated circuit device". Background Technology

[0002] This disclosure generally relates to integrated circuits such as processors and / or field-programmable gate arrays (FPGAs). More specifically, this disclosure relates to using modular periphery tiles to improve the scalability of integrated circuits (e.g., component scalability, product variant scalability).

[0003] This section aims to introduce the reader to various technical aspects that may relate to the various aspects of this disclosure described and / or claimed below. It is believed that this discussion will help provide the reader with background information to facilitate a better understanding of the various aspects of this disclosure. Therefore, it is to be understood that these statements will be read in this context and not as an admission of prior art.

[0004] Modern electronic devices (such as computers, portable devices, network routers, data centers, Internet-connected appliances, and many more) tend to include at least one integrated circuit device. Integrated circuit devices can take various forms; to name just a few examples, they include processors (e.g., central processing units (CPUs)), memory devices, and programmable devices (e.g., FPGAs). Programmable devices can in particular include programmable logic fabrics (e.g., configurable logic blocks (CLBs)) that can be programmed (e.g., configured) and reprogrammed (e.g., reconfigured) after manufacturing to provide a wide variety of functionalities based on circuit design.

[0005] To facilitate operation, programmable devices may include various peripheral intellectual property (IP) cores near and around the programmable architecture. For example, double data rate (DDR) IP may be placed on the shoreline of the programmable architecture to avoid consuming excessive routing circuitry. However, due to the limited amount of shoreline, design compromises may arise when determining the number and type of peripheral IP to be included around the shoreline. This may result in reduced device performance for certain product variants and reduced overall scalability of the programmable device (e.g., component scalability, product scalability). Attached Figure Description

[0006] Various aspects of this disclosure can be better understood by reading the following detailed description and by referring to the accompanying drawings, in which: Figure 1This is a block diagram of a data processing system including an integrated circuit device according to embodiments of the present disclosure; Figure 2 Embodiments according to this disclosure can promote Figure 1 A block diagram of a design workstation for programming integrated circuit devices; Figure 3 This is a schematic diagram of an integrated circuit device including a peripheral intellectual property core (IP) that has been disaggregated from the programmable structure of the integrated circuit device, according to embodiments of the present disclosure. Figure 4 This is a schematic diagram of a peripheral IP architecture when the peripheral IP has been modularized into peripheral IP fragments according to an embodiment of the present disclosure; Figure 5 This is a schematic diagram of an integrated circuit device including an architecture of programmable architecture of the integrated circuit device and an architecture of de-aggregated peripheral IP according to embodiments of the present disclosure; Figure 6 This is another schematic diagram of an integrated circuit device according to embodiments of the present disclosure, including an architecture of programmable architecture of the integrated circuit device and an architecture of de-aggregated peripheral IP; and Figure 7 This is another schematic diagram of an integrated circuit device including an architecture of programmable configuration of the integrated circuit device and an architecture of de-aggregated peripheral IP according to embodiments of the present disclosure. Detailed Implementation

[0007] One or more specific embodiments will be described below. For the purpose of providing a concise description of these embodiments, not all features of the actual implementation are described in the specification. It should be appreciated that, as in any engineering or design project, numerous implementation-specific decisions must be made to achieve the developer's specific goals, which may vary from implementation to implementation, such as compliance with system-related and business-related constraints. Furthermore, it should be appreciated that such development efforts may be complex and time-consuming, but will be nothing more than routine work of design, preparation, and manufacture for those skilled in the art who benefit from this disclosure.

[0008] When describing elements of various embodiments of this disclosure, the articles “a” and “the” are intended to refer to the presence of one or more of the elements. The terms “comprising,” “including,” and “having” are intended to be inclusive and mean that additional elements may be present in addition to those listed. Furthermore, it should be understood that references to “an embodiment” or “an embodiment” in this disclosure are not intended to be construed as excluding the existence of additional embodiments that also incorporate the described features.

[0009] Integrated circuits (such as field-programmable gate arrays (FPGAs)) may include programmable architectures (e.g., programmable architecture circuitry) that can be configured and, in some cases, reconfigured later to implement a circuit design. To facilitate operation of the integrated circuit, it may include one or more peripheral intellectual property cores (IPs) near the programmable architecture. In particular, for FPGAs, higher-bandwidth peripheral IPs (e.g., double data rate (DDR), low-power DDR (LPDDR), high-bandwidth memory (HBM), embedded static random access memory (eSRAM), universal interface bus (UIB), input / output (I / O), and the like) may be placed around (e.g., on top, bottom) (e.g., at the shoreline) the programmable architecture to improve resource (e.g., wiring) allocation throughout the FPGA. While this disclosure will primarily use examples of programmable devices (e.g., FPGAs), the systems and techniques of this disclosure can be applied to any suitable integrated circuit device, such as processors (e.g., central processing units (CPUs)), memories, and the like.

[0010] However, FPGAs can include a limited amount of shoreline around the programmable architecture, leading to compromises in the type and quantity of peripheral IP that can be included in the design. As an example, a single floorplan (e.g., an integrated circuit layout) can be used for multiple device variants, each with different specifications for peripheral IP and programmable architecture. Using a single floorplan can reduce design and fabrication costs. To improve the applicability of a single floorplan to multiple product variants, the floorplan can include designs that compromise on the number of UIBs, I / Os, eSRAMs, and other peripheral IPs included in the floorplan. Therefore, the limited amount of shoreline can lead to reduced device performance for each device variant and reduced overall scalability across multiple device variants.

[0011] FPGAs can be decomposed into smaller dies to enable peripheral IP customization for multiple device variants and generally improve silicon yield (e.g., the number of dies fabricated onto a silicon wafer with normal performance). However, because each type of peripheral IP can have a unique interface with the programmable assembly die, interface customization and redesign can occur for each device variant. Therefore, peripheral IP with undefined interface connections on decomposed smaller dies (e.g., non-standardized interface connections) can increase the number of silicon tape-in ​​preparations (e.g., IP integration), tape-in ​​costs (e.g., the result of a design process prior to sending the design for manufacturing), and the location limitations of the peripheral IP. Consequently, the scalability of FPGA components and the scalability of variant selection can be limited.

[0012] Therefore, and as further detailed below, embodiments of this disclosure generally relate to improving the scalability and performance of integrated circuit devices such as FPGAs by die de-assembling peripheral IP. In some embodiments, peripheral IP can be de-assembled from the main programmable architecture. In particular, peripheral IP can be modularized into peripheral IP chips (e.g., I / O chips, UIB chips, eSRAM chips, and the like) that can be designed and fabricated independently of the design and fabrication of the programmable architecture die. In this way, the cost and time consumed during tape-out preparation and tape-out of variant designs can be reduced. Furthermore, since the peripheral IP is no longer located on the main programmable architecture die, FPGA connectivity is not negatively affected by the peripheral IP.

[0013] Additionally or alternatively, in some embodiments, communication between programmable architecture dies and between programmable architecture dies and peripheral IP blocks may occur, for example, via silicon interposers (e.g., electrical signal conduits) and stitched (e.g., interconnected) direct communication (e.g., Advanced Interface Bus (AIB), Direct Interface Columns). Due to the direct communication links between programmable architecture dies and peripheral IP blocks, the programmable device can behave as a monolithic integrated circuit and be programmed as a monolithic integrated circuit (e.g., as if multiple dies were a single monolithic chip). Furthermore, modular interface connections, such as those via silicon interposers (e.g., silicon interfaces) and stitched direct communication, can reduce the cost / time of the design process and improve the scalability of the programmable device, as new peripheral IP can be included in the programmable device design without significant design overhead.

[0014] Considering the preceding text, Figure 1 This is a block diagram of a data processing system 100 including an integrated circuit device 102 according to embodiments of the present disclosure. The data processing system 100 may include more or fewer components than shown (e.g., electronic displays, user interface structures, application-specific integrated circuits (ASICs)). The data processing system 100 may include one or more host processors 104, such as Intel® Xeon® processors or reduced instruction processors (e.g., Reduced Instruction Set Computer (RISC), Advanced RISC Machine (ARM) processors), which can manage data processing requests for the data processing system 100 (e.g., to perform machine learning, video processing, speech recognition, image recognition, data compression, database search and ranking, bioinformatics analysis, cybersecurity pattern recognition, space navigation, or the like).

[0015] One or more host processors 104 may communicate with memory and / or storage device circuitry 106, which may be a tangible, non-transitory machine-readable medium, such as random access memory (RAM), read-only memory (ROM), one or more hard disk drives, flash memory, or any other suitable optical, magnetic, or solid-state storage medium. Memory and / or storage device circuitry 106 may hold data to be processed by data processing system 100, such as processor-executable control software, configuration software, system parameters, configuration data, etc. Data processing system 100 may also include a network interface 108 that allows data processing system 100 to communicate with other electronic devices. In some embodiments, data processing system 100 may be part of a data center that handles a variety of different requests. For example, data processing system 100 may receive data processing requests via network interface 108 to perform machine learning, video processing, speech recognition, image recognition, data compression, database search and sorting, bioinformatics analysis, network security pattern recognition, space navigation, or some other specialized task. The data processing system 100 may further include an integrated circuit device 102 that implements data processing requests.

[0016] According to embodiments of this disclosure, such as Figure 2 As shown, a designer can use design workstation 200 to develop a design that can configure integrated circuit device 102. In some embodiments, the designer can use design software 204 (e.g., Intel® Quartus®) to generate a design that can be used to program (e.g., configure) integrated circuit device 102. For example, the designer can program the modular integrated circuit device 102 to achieve specific functionalities, such as implementing a circuit design as if the integrated circuit device 102 were monolithic (e.g., a higher-level circuit design). Integrated circuit device 102 can be a programmable integrated circuit, such as a field-programmable gate array (FPGA), comprising one or more programmable configuration dies and one or more peripheral IP chip dies.

[0017] Accordingly, design software 204 can use compiler 206 to generate lower-level circuit design configurations for integrated circuit device 102. That is, compiler 206 can provide machine-readable instructions representing the functionality specified by the designer, for example, in the form of configuration bit stream 208, to integrated circuit device 102. One or more host processors 104 can coordinate the loading of bit stream 208 onto integrated circuit device 102 and the subsequent programming of programmable configurations.

[0018] To improve the scalability of integrated circuits (e.g., FPGAs), peripheral IP can be modularized into peripheral IP chips, allowing the design and fabrication of peripheral IP chips to occur independently of the design and fabrication of programmable architecture dies, and in a manner that enables modular interface connections between peripheral IP chips and programmable architecture dies (e.g., as defined by specifications to achieve a substantially uniform interface connection for different chips, which can be standardized). Figure 3 The illustration shows an example of an integrated circuit device 102 constructed using de-assembled peripheral IP chips and a programmable configuration die according to an embodiment of the present disclosure. Although the integrated circuit device 102 is shown as an FPGA, it should be appreciated that peripheral IP chips can be de-assembled for any suitable type of integrated circuit device 102.

[0019] As shown, one or more programmable configuration dies 302A, 302B, 302C, and 302D (collectively referred to as 302) may be included in the integrated circuit device 102 to facilitate FPGA programming. In particular, the programmable configuration die 302 may include programmable logic, such as sectors of programmable logic. The programmable logic may be arranged as a group sometimes referred to as configurable logic blocks (CLBs) or logic array blocks (LABs). The programmable logic may receive a configuration bit stream 208 from one or more host processors 104 and can be configured according to the circuit design embodied by the configuration bit stream 208. The programmable configuration die 302 may be configured or partially configured at runtime and / or may be reconfigured or partially reconfigured later, thereby enabling post-manufacturing design flexibility.

[0020] Furthermore, as shown, peripheral IP chip 304 can be placed near programmable architecture die 302. Peripheral IP chip 304 can be peripheral IP, such as I / O IP, UIB IP, memory IP, and the like. Peripheral IP chip 304 can be designed and / or fabricated independently of programmable architecture die 302 because it is not located on the same die as programmable architecture die 302. Therefore, peripheral IP no longer limits the buildable die size or reduces the programming flexibility of the FPGA.

[0021] Communication on the integrated circuit device 102 can occur via the silicon interposer 306 or via an embedded interface bridge (which is a signal conduction channel through the silicon substrate). In particular, despite modularizing the peripheral IP, intra-device communication within the programmable structure die (e.g., 302A) remains unchanged. Furthermore, inter-device communication (e.g., structure-to-device communication) can occur via the interface within the silicon interposer 306, just as communication can occur between the programmable structure die 302 and the peripheral IP chip 304. Because the communication interface (e.g., a modular interface) is defined by specifications and is direct between the programmable structure die 302 and the peripheral IP chip 304, the integrated circuit device 102 can be customized for different product variants and can be programmed via the design workstation 200 as if the integrated circuit device 102 were monolithic.

[0022] Figure 4 The illustration shows an example of a peripheral IP chip 304 that can be designed and fabricated independently of the programmable architecture die 302. In some embodiments, the peripheral IP chip 304 may be an I / O chip 304A and / or a UIB chip 304B. In short, the I / O chip 304A and UIB chip 304B can be used to facilitate communication between the programmable architecture die 302 and external devices. The I / O chip 304A and UIB chip 304B may be divided into separate chips due to design bandwidth and capacity limitations. As an example, some device applications can compromise on memory capacity while using higher memory bandwidth, or vice versa. Some memory IPs (such as High Bandwidth Memory (HBM) IPs or external ASICs) can compromise on memory capacity while supporting high memory bandwidth and can therefore be muxed to the UIB chip 304B. Other memory IPs (such as Double Data Rate (DDR) IPs) can compromise on memory bandwidth while supporting higher memory capacity and can therefore be muxed to the I / O chip 304A. By dividing the device interface into I / O blocks 304A and UIB blocks 304B, the connection between the die 302 and external devices can be tailored (e.g., via multiplexing) depending on the target product variant. In this way, the peripheral IP blocks 304 enable customization of product variants and future product scalability.

[0023] The I / O panel 304A can include any suitable number of I / O blocks and / or physical I / O connection lanes. Figure 4In the example, there are four IO96 blocks 404A, 404B, 404C, and 404D (collectively referred to as 404), each supporting 96 IO connection paths, but different blocks may support more or fewer IO connection paths. Each IO96 block 404 may be a baseline configurable I / O block that can be programmed to form general purpose input / output (GPIO) ports and / or programmed to interface with external devices, such as external memory (e.g., DDR). Each IO96 block 404 may include four microbumps providing die-to-die connections, and, as will be discussed in more detail below, each microbump may be associated with an Advanced Interface Bus-Direct (AIB-D) interface 406 for directly connecting the programmable configuration die 302 and peripheral IP chips 304 via DDR. In short, the AIB-D interface 406 may be an IP array within a silicon interposer 306 that houses a bidirectional I / O buffer, and may provide wire-to-wire connectivity and any suitable form of time-division multiplexing (TDM) support (e.g., 2:1 TDM, 3:1 TDM, 4:1 TDM, 5:1 TDM, 6:1 TDM, 7:1 TDM, 8:1 TDM, or higher). It should be further appreciated that TDM support can be of any kind.

[0024] Furthermore, I / O block 304A may include an interface for an on-chip fabrication network (FNOC) 408 within the silicon interposer 306, which can use a high-speed (e.g., high bandwidth of 0.25-0.5 megabytes per second per link per direction) NOC interface to bridge I / O block 304A to programmable fabrication die 302. Similarly, UIB block 304B may include an AIB-D interface 406 and an interface for FNOC 408 to facilitate connectivity between UIB block 304B and programmable fabrication die 302. Although I / O blocks 304A and UIB blocks 304B are not part of programmable fabrication die 302, blocks 304A and 304B may include FNOCs to facilitate communication. Furthermore, it should be understood that peripheral IP block 304 has been described using I / O block 304A and UIB block 304B. Peripheral IP block 304 can be any IP block, such as a DDR IP block.

[0025] Figure 5The de-aggregated FPGA device 500 further illustrates the connectivity and layout architecture of the modular peripheral IP chips 304 and programmable configuration dies 302 according to embodiments of the present disclosure. It should be appreciated that while the FPGA device 500 is shown as having a certain number of programmable configuration dies 302, connections, and peripheral IP chips 304 types, suitable variations of the FPGA device 500 are contemplated in this disclosure.

[0026] The FPGA device 500 is shown as a top view 502 and side views 525, 575. The FPGA device 500 may include peripheral IP blocks 304, such as I / O blocks 304A and UIB blocks 304B. As previously described, the peripheral IP blocks 304 may include AIB-D interfaces 406 and FNOC 408. The number of AIB-D 406 included in the peripheral IP blocks 304 may be based on the number of available microbumps and may be defined by a specification that modularizes the interface between the programmable assembly dies (e.g., glue dies) 504A, 504B, 504C, 504D (collectively referred to as 504) included in the FPGA device 500 and the peripheral IP blocks 304. For example, I / O block 304A may include four IO96 blocks, each with four AIB-D 406s, because the width of the glued die 504 may be four sectors 508, and each sector 508 may include four AIB-D 506s. As another example, the width of the glued die 504 may be four sectors 508, and each sector 508 may include four AIB-D 506s matching the UIB interface and the HBM interface. The glued die 504 may include a security module (e.g., a Security Device Manager (SDM)) used to authenticate the configuration data of each glued die 504.

[0027] Furthermore, the AIB-D interface 406 in the silicon interposer 306 can provide a direct connection between the glued die 504A and the peripheral IP chip 304, allowing the FPGA device 500 to be programmed as if it were a monolithic device. Additionally, the 2:1 TDM capability of the AIB-D interface 406 can increase the number of direct connections. Specifically, for lower frequency interfaces such as between the glued die 504A and the peripheral IP chip 304, the 2:1 TDM can provide two connections per microbump due to the bidirectional buffer of the AIB-D interface 406. Thus, the limitation on the number of connections that may be caused by the limited number of microbumps can be overcome, and further reach (e.g., connections) is possible without having to push the AIB-D interface 406 column deeper into the glued die 504. On the other hand, structure-to-structure communication can bypass the 2:1 TDM functionality and can instead use wire-to-wire communication.

[0028] Furthermore, the AIB-D interface 406 can increase shoreline reach, allowing peripheral IP panel 304 to communicate with plywood die 504 without consuming excessive wiring circuitry. As an example, each sector 508 can have four columns of AIB-D interfaces 406. Four columns of AIB-D interfaces 406 can increase shoreline reach to four times the shoreline reach available when only horizontal connections of plywood die 504 are available for interface connections.

[0029] Additionally, components of the FPGA device 500 can communicate via NOCs (such as FNOC 408 or a source of an on-chip configuration network (CNOC)). As previously described, peripheral IP block 304 may include FNOC 408 for high-speed communication with NOC 510 of the glued die 504. In particular, an Advanced Interference Bus - Edge (AIB-E) (e.g., an edge interface) can facilitate high-speed DDR interface connections across FNOCs 408 and 510. For example, AIB-E can bridge NOC 510 of glued die 504, NOC 510 within glued die (e.g., 504A) together, and / or bridge NOC 510 of glued die 504 to FNOC 408 of peripheral IP block 304. Moreover, AIB-E can transmit packetized data from router 512 to router 512 until the packetized data reaches its destination.

[0030] The FPGA device 500 may also include an interface for bridging dies with low connectivity bandwidth to a general-purpose Advanced Interface Bus-Generic (AIB-G) (e.g., a generic interface) such as a transceiver block. Additionally, as shown in side views 525 and 572, the glued die 504 and peripheral IP block 304 may be placed on a silicon interposer (e.g., a passive interposer) 306. The silicon interposer 306 may be supported by a package substrate 514 that supports the electrical connections of the FPGA device 500. While the silicon interposer 306 is described as a passive interposer in many examples, it may also contain active components, and in some embodiments, may be an active interposer.

[0031] Figure 6The illustration shows additional examples of FPGA devices 600, 650 comprising depolymerized peripheral IP chips 304 and glued dies 504 according to embodiments of the present disclosure. Because the interface between the peripheral IP chips 304 and the glued die 504 is defined by the specification, and because the peripheral IP chips 304 are modular, various combinations of the peripheral IP chips 304 can be included within the FPGA die. For example, the FPGA device 600 may include a single UIB chip 304B and a plurality of I / O chips 304A. As another example, the FPGA device 650 may include only I / O chips. In some embodiments, the peripheral IP chips 304 at the bottom of the FPGA device 600, 650 may be a flipped version of the peripheral IP chips 304 at the top of the FPGA device 600, 650.

[0032] Figure 7 The illustration shows an additional example of a depolymerized FPGA device 700 comprising depolymerized peripheral IP chips 304 and glued dies 712A, 712B (collectively referred to as 712) according to embodiments of the present disclosure. In particular, the peripheral IP chips 304 and glued dies 712 may include an AIB-E interface 702 for bridging the NOCs and / or FNOCs of one or more glued dies 712 and the peripheral IP chips 304 to each other. The peripheral IP chips 304 and glued dies 712 may include an AIB-D interface 704, which provides a direct interface between the peripheral IP chips 304 and the glued dies 712 without requiring on-chip network-based interface connections. In some embodiments, the AIB-D interface 704 of the peripheral IP chips 304 may include wider columns to facilitate direct communication. Furthermore, the AIB-D interface 704 of the glued dies 712 may include five columns of AIB-D interfaces 704. It should be recognized that any number and any width of AIB-D interface 704 columns can be included in the peripheral IP block 304 and / or glued core 712.

[0033] In some embodiments, the SDM 706 can be depolymerized from the glued die 712 and instead can be included in the peripheral IP block 304. This can further remove barriers from the FPGA wiring circuitry within the programmable architecture. Since the SDM 706 has been included in the peripheral IP block 304 instead of the glued die 712, a die subsystem manager (DSSM) can be included in the glued die 712 to act as a proxy for the SDM. Additionally, a hard processor system (HPS) 708 can be included in the peripheral IP block 304. As shown, the glued die 712 can include different instances of core architectures. For example, a glued die (e.g., glued die 1) 712A can include a higher amount of digital signal processing (DSP) blocks than other glued dies 712. As another example, a glued die (e.g., glued die 2) 712B can include more memory blocks than other glued dies 712.

[0034] This system and technology relate to embodiments for improving the scalability of an integrated circuit device 102 by modularizing peripheral IP. This embodiment enables the customization of the integrated circuit device 102 for different variants using a single tape in a scalable manner. Furthermore, this embodiment can reduce die size and improve silicon yield. In addition, the embodiment allows design software to view and program the deassembled peripheral devices and programmable architecture as if the integrated circuit device 102 were monolithic. Furthermore, parallel connectivity can be improved due to the reduction in package traces used for wiring connections and the reduction in obstacles caused by peripheral IP within the wiring of the programmable architecture.

[0035] While the embodiments set forth in this disclosure may be readily modified and alternatively formulated, specific embodiments have been illustrated by way of example in the accompanying drawings and described in detail herein. However, it should be understood that this disclosure is not intended to be limited to the particular forms disclosed. This disclosure will cover all modifications, equivalents, and alternatives falling within the spirit and scope of this disclosure as defined by the following appended claims.

[0036] The techniques proposed and claimed herein are referenced and applied to physical and specific examples that can arguably improve the practical nature of the art, and are therefore not abstract, intangible, or purely theoretical. Furthermore, if any claim appended to this specification contains one or more elements designated as “component for [performing]...[function]” or “step for [performing]...[function]”, it is intended that such elements be interpreted according to 35 USC 112(f). However, for any claim containing elements designated in any other way, it is intended that such elements not be interpreted according to 35 USC 112(f).

[0037] The following examples are provided: Example 1. An integrated circuit system, comprising: A first die, the first die including a first programmable architecture circuit; and The second die includes a first modular peripheral intellectual property (IP) chip, wherein the first modular peripheral IP chip includes a first circuit configured to perform a first function associated with a programmable configuration circuit, wherein the second die is communicatively coupled to the first die via a modular interface configured to enable the following communication: When the first die is coupled to the second die via the modular interface, communication occurs between the first die and the second die; When the first die is coupled to the third die instead of the second die via the modular interface, communication occurs between the first die and the third die, wherein the third die includes a second modular peripheral intellectual property (IP) chip, wherein the second modular peripheral IP chip includes second circuitry configured to perform a second function associated with the first programmable architecture circuitry; and When the second die is coupled to the fourth die instead of the first die via the modular interface, communication occurs between the second die and the fourth die, wherein the fourth die includes a second programmable architecture circuit, and wherein the first circuit of the first modular peripheral IP block is configured to perform the first function associated with the second programmable architecture circuit.

[0038] Example 2. An integrated circuit system as described in Example 1, wherein the programmable configuration circuitry includes a field-programmable gate array (FPGA) circuitry.

[0039] Example 3. An integrated circuit system as described in Example 1, wherein the first modular peripheral IP chip or the second modular peripheral IP chip comprises a double data rate (DDR) chip, a low power DDR (LPDDR) chip, a high bandwidth memory (HBM) chip, an embedded static random access memory (eSRAM) chip, a universal interface bus (UIB) chip, or an input / output (I / O) chip, or any combination thereof.

[0040] Example 4. An integrated circuit system as described in Example 3, wherein the modular interface includes circuitry defined by a specification, the circuitry enabling communication via a silicon interposer with the Double Data Rate (DDR) chip, the Low Power DDR (LPDDR) chip, the High Bandwidth Memory (HBM) chip, the Embedded Static Random Access Memory (eSRAM) chip, the Universal Interface Bus (UIB) chip, the Input / Output (I / O) chip, or any combination thereof, to transfer data to and from the programmable architecture circuitry.

[0041] Example 5. An integrated circuit system as described in Example 1, wherein the second die is configurable to bridge an on-chip configuration network (CNOC) connection from the third die, which includes additional programmable configuration circuitry, to the second die, wherein the third die includes a configuration source.

[0042] Example 6. An integrated circuit system as described in Example 1, wherein the modular interface is configured to perform time division multiplexing (TDM).

[0043] Example 7. An integrated circuit system as described in Example 6, wherein the modular interface is configured to facilitate direct communication between the first die and the second die, such that the integrated circuit appears as a monolithic integrated circuit to the design software.

[0044] Example 8. An integrated circuit system as described in Example 6, wherein the modular interface is configured to support 2:1 time division multiplexing (TDM) between the peripheral IP chips and the programmable building blocks.

[0045] Example 9. An integrated circuit system as described in Example 1, wherein the first die includes a first on-chip fabrication network (FNOC), and wherein the second die includes a second on-chip fabrication network (FNOC).

[0046] Example 10. An integrated circuit system as described in Example 9, wherein the modular interface is configurable to bridge the first FNOC to the second FNOC.

[0047] Example 11. An integrated circuit system as described in Example 9, wherein the first FNOC and the second FNOC are configured to facilitate high-speed, high-bandwidth communication between the peripheral IP chip and the programmable architecture circuit, between multiple programmable architecture circuits, or any combination thereof.

[0048] Example 12. A method for programming a de-aggregated integrated circuit system, comprising: Higher-level circuit designs are generated using design software for implementation on the de-aggregated integrated circuit system; At least based on the higher-level circuit design, the lower-level circuit design configuration is generated using the design software; The design software is used to load the lower-level circuit design onto the depolymerized integrated circuit system; and The de-aggregated integrated circuit system is configured using the design software based on the lower-level circuit design, wherein the de-aggregated integrated circuit system is configured as if the de-aggregated integrated circuit system were a monolithic integrated circuit device.

[0049] Example 13. The method as described in Example 12, wherein a modular interface between the depolymerized dies of the depolymerized integrated circuit system facilitates the configuration of the depolymerized integrated circuit system as if the depolymerized integrated circuit system were the monolithic integrated circuit device.

[0050] Example 14. The method as described in Example 13, wherein the modular interface includes a direct interface in a silicon interface, wherein the direct interface is configurable to facilitate direct communication between the de-galvanized dies of the de-galvanized integrated circuit system, wherein the direct interface includes a time-division multiplexing (TDM) interface.

[0051] Example 15. An integrated circuit system, comprising: Programmable configuration die, the programmable configuration die comprising: A first on-chip network of configuration (FNOC) configured to facilitate high-bandwidth communication, wherein the FNOC includes an on-chip network of configuration (NOC) embedded within the programmable configuration die; and A first direct interface column, configured to facilitate direct communication without requiring an interface connection via the first FNOC, wherein the first direct interface column includes a time division multiplexing (TDM) interface; Modular peripheral intellectual property (IP) chips, the modular peripheral intellectual property (IP) chips being communicatively coupled to the programmable architecture die, wherein the modular peripheral IP chips include: A second FNOC, configured to facilitate high-bandwidth communication with the programmable die via an interface connection to the second FNOC; and A second direct interface column, configured to facilitate direct communication with the programmable configuration die, wherein the direct communication with the programmable configuration die is performed via an interface connection with the first direct interface column without via an interface connection through the second FNOC, wherein the second interface column includes the TDM interface.

[0052] Example 16. An integrated circuit system as described in Example 15, wherein the second FNOC is configured to be interfaced with the second FNOC via an edge interface column, facilitating communication with the programmable configuration die.

[0053] Example 17. An integrated circuit system as described in Example 15, wherein the modular peripheral IP chips include a plurality of modular I / O blocks, each I / O block including a direct interface column.

[0054] Example 18. An integrated circuit system as described in Example 15, wherein the modular peripheral IP chip includes a Universal Interface Bus (UIB) chip configured to interface with High Bandwidth Memory (HBM) IP.

[0055] Example 19. An integrated circuit system as described in Example 18, wherein the UIB chip facilitates bridging from an additional programmable configuration die to an on-chip configuration network (CNOC) connection of the programmable configuration die, wherein the additional programmable configuration die includes a configuration source.

[0056] Example 20. An integrated circuit system as described in Example 15 includes at least one additional programmable configuration die, wherein the at least one additional programmable configuration die is communicatively coupled to the programmable configuration die via at least one direct interface column and the FNOC.

Claims

1. A multi-die package, comprising: Substrate; A plurality of first dies mounted on the substrate, the plurality of first dies including a plurality of first wiring devices for facilitating external communication, wherein a respective die among the plurality of first dies performs processing; and An input / output (I / O) die mounted on the substrate, the I / O die being configurable to support an external memory interface, the I / O die including a second plurality of routers, the second plurality of routers being configured to facilitate external communication with the first die via an interface connection to the first plurality of routers. The I / O die facilitates out-of-package communication between the plurality of first dies and one or more out-of-package devices via the substrate.

2. The multi-die package as described in claim 1, wherein, The I / O die can be configured to support a double data rate DDR external memory interface.

3. The multi-die package as described in claim 2, wherein, The plurality of first dies and the I / O dies are communicatively coupled via the substrate through one or more channels formed in the substrate.

4. The multi-die package of claim 2, comprising a packet transmission system spanning the plurality of first dies and the I / O dies, wherein, The packet transmission system includes a first plurality of routers and a second plurality of routers to transmit data packets between the plurality of first dies and the I / O dies.

5. The multi-die package as described in claim 1, wherein, The respective die among the plurality of first dies has a first size and the I / O die has a second size different from the first size.

6. The multi-die package as described in claim 1, wherein: The plurality of first dies are manufactured using a first process technology; The I / O die is manufactured using a second process technology; as well as The first process technology is different from the second process technology.

7. The multi-die package as described in claim 1, wherein, The die in the plurality of first dies includes a security module for authenticating the data of the die in the plurality of first dies.

8. The multi-die package as described in claim 1, wherein, The corresponding die among the plurality of first dies includes the corresponding router among the plurality of routers.

9. The multi-die package as described in claim 8, wherein, The corresponding router facilitates the transfer of packetized data to the destination processor.

10. The multi-die package as described in claim 1, wherein, The one or more external devices include random access memory (RAM), read-only memory (ROM), flash memory, one or more hard disk drives, network interface, one or more host processors, or any combination thereof.

11. An integrated circuit package, comprising: A plurality of first dies, the plurality of first dies including a first packet transmission circuit for facilitating communication, wherein a respective die among the plurality of first dies performs processing; and An input / output (I / O) die is used to support an external memory interface, wherein the I / O die includes a second packet transmission circuit, which facilitates communication with the plurality of first dies by connecting to the first packet transmission circuit via an interface. The I / O die facilitates out-of-package communication between the plurality of first dies and one or more out-of-package devices via a substrate. Among them, the corresponding die in the plurality of first dies has a first size and the I / O die has a second size different from the first size.

12. The integrated circuit package as described in claim 11, wherein, The I / O die can be configured to support a double data rate DDR external memory interface.

13. The integrated circuit package as described in claim 11, wherein, The die in the plurality of first dies includes a security module for authenticating the data of the die in the plurality of first dies.

14. The integrated circuit package as described in claim 11, wherein, The one or more external devices include random access memory (RAM), read-only memory (ROM), flash memory, one or more hard disk drives, network interface, one or more host processors, or any combination thereof.

15. An integrated circuit system, comprising: One or more external devices; Multi-die package, the multi-die package comprising: Substrate; A plurality of first dies mounted on the substrate, wherein a corresponding die among the plurality of first dies performs a process, wherein the plurality of first dies includes a plurality of first wiring devices for facilitating communication; and Input / output (I / O) dies mounted on the substrate, the I / O dies being used to support external memory interfaces, wherein the I / O dies: Facilitating out-of-package communication between the plurality of first dies and the one or more out-of-package devices via the substrate; and It includes a second plurality of routers, which are used to facilitate communication with the plurality of first dies by connecting to the first plurality of routers via an interface.

16. The integrated circuit system of claim 15, wherein, The one or more external devices include random access memory (RAM), read-only memory (ROM), flash memory, one or more hard disk drives, network interface, one or more host processors, or any combination thereof.

17. The integrated circuit system of claim 15, wherein, The die in the plurality of first dies includes a security module for authenticating the data of the die in the plurality of first dies.

18. The integrated circuit system of claim 16, further comprising the network interface configured to communicate with one or more devices outside the package.

19. The integrated circuit system of claim 15, wherein, The plurality of first dies and the I / O die are communicatively coupled via the substrate through one or more channels formed in the substrate; and wherein a respective die among the plurality of first dies has a first size and the I / O die has a second size different from the first size.

20. The integrated circuit system of claim 15, wherein: The plurality of first dies are manufactured using a first process technology; The I / O die is manufactured using a second process technology; and The first process technology is different from the second process technology.

Citation Information

Patent Citations

  • Compact low cost plastic MCM to PCB

    US20040190274A1