Universal substrate and server

By electrically connecting open acceleration modules and high-speed connectors on independent carriers, and utilizing low-loss cables and conductive frames, the problems of low high-speed signal interaction rate and complex wiring in artificial intelligence servers are solved, achieving more efficient signal transmission and cost reduction.

CN120949897APending Publication Date: 2025-11-14XFUSION DIGITAL TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202510902743.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-30
Publication Date
2025-11-14

AI Technical Summary

Technical Problem

In artificial intelligence servers, the high-speed signal interaction rate between the open acceleration module and the high-speed connector needs to be improved. In the existing technology, the signal attenuation is large and the wiring complexity is high, which leads to increased cost and power consumption.

Method used

By setting open acceleration modules and high-speed connectors on independent carrier components and using a first cable for electrical connection, the dependence on the motherboard is reduced or eliminated, the signal transmission path is simplified, and low-loss high-speed cables and conductive frames are used for electrical connection to adapt to different power requirements.

Benefits of technology

It improves the high-speed signal interaction rate between the open acceleration module and the high-speed connector, reduces signal attenuation, simplifies wiring difficulty and complexity, reduces the number of retimers, and lowers cost and power consumption.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120949897A_ABST
    Figure CN120949897A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a universal substrate and a server, and relates to the technical field of servers. The universal substrate comprises a first bearing part, a second bearing part, an open acceleration module, a high-speed connector and a first cable. The open acceleration module is arranged on the first bearing part, the high-speed connector is arranged on the second bearing part, the first cable is located outside the first bearing part and the second bearing part, the high-speed connector is electrically connected with the open acceleration module through the first cable, and high-speed signals between the open acceleration module and the corresponding high-speed connector can have a high interaction rate.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of server technology, and in particular to a general-purpose substrate and server. Background Technology

[0002] Servers requiring high performance, such as artificial intelligence (AI) servers, often incorporate a universal baseboard (UBB) that conforms to the Open Compute Project (OCP) specification. This UBB includes multiple Open Accelerator Modules (OAMs) based on the OCP standard, corresponding high-speed connectors for each OAM, and a motherboard. In these technologies, both the OAMs and high-speed connectors are located on and electrically connected to the motherboard. The OAMs and their corresponding high-speed connectors are interconnected via the motherboard to enable high-speed signal exchange between them. However, the high-speed signal exchange rate between the OAMs and their corresponding high-speed connectors still needs further improvement in these technologies. Summary of the Invention

[0003] This application provides a general-purpose substrate and server that enables high-speed signals between the open acceleration module and the corresponding high-speed connector to have a high interaction rate.

[0004] A first aspect of this application provides a universal substrate, which includes a first carrier, a second carrier, an open acceleration module, a high-speed connector, and a first cable. The open acceleration module is disposed on the first carrier, the high-speed connector is disposed on the second carrier, and the first cable is located outside the first and second carriers. The high-speed connector and the open acceleration module are electrically connected via the first cable.

[0005] The universal substrate provided in this application embodiment allows high-speed signals to be transmitted between the high-speed connector and the open acceleration module via a first cable. The attenuation of the high-speed signal transmitted through the first cable is minimal, resulting in less attenuation of the high-speed signals interacting between the open acceleration module and the high-speed connector. This reduces the number of retimers required between the open acceleration module and the high-speed connector, or even eliminates the need for retimers at all. This simplifies the high-speed signal transmission link between the open acceleration module and the high-speed connector, reduces the latency of high-speed signals between them, and thus improves the interaction rate of high-speed signals. Reducing the number of retimers also helps to lower the cost of the universal substrate and reduce the power consumption of the server.

[0006] Since the high-speed connector and the corresponding open acceleration module are electrically connected via a first cable, the motherboard of the universal substrate does not need to have traces for high-speed signal transmission between the high-speed connector and the corresponding open acceleration module. This reduces the wiring difficulty and complexity of the motherboard of the universal substrate, thereby reducing the number of layers and the size of the motherboard, and ultimately reducing the cost of the universal substrate.

[0007] The open acceleration module and the high-speed connector, which are electrically connected by the first cable, are respectively located on the first carrier and the second carrier, which are independent of each other. At least one of the open acceleration module and the high-speed connector is not located on the motherboard of the general-purpose substrate. This helps to reduce the difficulty and complexity of the device and wiring layout of the motherboard of the general-purpose substrate, reduces the number of layers of the motherboard of the general-purpose substrate, reduces the size of the motherboard of the general-purpose substrate, and reduces the cost of the general-purpose substrate.

[0008] The open acceleration module and its corresponding high-speed connector are electrically connected via a first cable. This allows the first and second carrier components to be independent of their circuit connection function, enabling greater flexibility in material selection for both components and reducing the cost of general-purpose substrates. Furthermore, since the component supporting the high-speed connector does not need to perform the circuit connection function, the high-speed connector does not require a bend in the connector for electrical connection, further enhancing the flexibility in high-speed connector selection.

[0009] In some possible implementations, the open acceleration module includes a first circuit board, a processor, and a first connector. The first circuit board is disposed on a first carrier, and the processor and the first connector are disposed on the first circuit board, with the processor electrically connected to the first connector. A first cable is located outside the first circuit board and the processor, and is connected to the first connector. A high-speed connector is electrically connected to the first connector via the first cable. This facilitates the installation and removal of the first cable from the open acceleration module.

[0010] In some possible implementations, the universal substrate further includes a second circuit board. The first carrier includes a first part and a second part, with the open acceleration module disposed in the first part and the second circuit board disposed in the second part. The first cable is located outside the second circuit board. This further reduces the difficulty and complexity of device and wiring layout on the motherboard of the universal substrate, which is beneficial for further reducing the number of layers and the size of the motherboard, and further reducing the cost of the universal substrate. In addition, since both the second circuit board and the open acceleration module are disposed on the first carrier, the second circuit board, the open acceleration module, and the first carrier can be pre-assembled as a whole before assembly, making the assembly and disassembly of the universal substrate easier.

[0011] In some possible implementations, the open acceleration module is positioned between the second circuit board and the high-speed connector. In this case, the smaller distance between the open acceleration module and the high-speed connector facilitates shortening the length of the first cable. Furthermore, the second circuit board and the components mounted thereon are less likely to interfere with the routing of the first cable, simplifying its arrangement and reducing its length. The shortened length of the first cable results in a shorter transmission path for high-speed signals between the open acceleration module and the high-speed connector, leading to less attenuation and higher speed in high-speed signal interaction between them.

[0012] In some possible implementations, the universal substrate further includes a power connector and a conductive element. The power connector is located on the second carrier, and the conductive element is located outside the first and second carriers. The power connector is electrically connected to the open acceleration module through the conductive element.

[0013] In this way, the conductive components located outside the first and second carriers can have a larger current-carrying area, facilitating the flow of larger currents and enabling the power supply of higher-power open-circuit acceleration modules. Furthermore, replacing the conductive components outside the first and second carriers is relatively easy, allowing for flexible adjustments based on the power of the open-circuit acceleration module to adapt to different modules. Additionally, the conductive components outside the first and second carriers can directly exchange heat with the external environment, resulting in higher heat dissipation efficiency and improving the stability and reliability of the general-purpose substrate.

[0014] Since the power connector supplies power to the open acceleration module via conductive components, the motherboard of the universal substrate does not require routing for powering the open acceleration module. This reduces the wiring difficulty and complexity of the motherboard, thereby reducing the number of layers and size of the motherboard and ultimately lowering the cost of the universal substrate.

[0015] The open acceleration module and the power connector are respectively located on the first carrier and the second carrier, which are independent of each other. At this time, at least one of the open acceleration module and the power connector is not located on the motherboard of the general-purpose substrate. This helps to reduce the difficulty and complexity of the device and wiring layout of the motherboard of the general-purpose substrate, reduce the number of layers of the motherboard of the general-purpose substrate, reduce the size of the motherboard of the general-purpose substrate, and reduce the cost of the general-purpose substrate.

[0016] The open acceleration module and the power connector are electrically connected via conductive components, allowing the first and second carrier components to be exempt from circuit connection functions. This enables greater flexibility in the selection of materials for the first and second carrier components, helping to reduce the cost of general-purpose substrates. Furthermore, since the component carrying the power connector does not need to perform the circuit connection function, the power connector does not need to be a bendable connector for electrical connection with the component carrying the power connector, making the selection of the power connector more flexible.

[0017] In some possible implementations, the conductive element is a conductive frame, which is fixed and electrically connected to both the power connector and the open acceleration module. This makes it convenient to connect multiple open acceleration modules to the power connector after the relative positions of the open acceleration module and the power connector are fixed. Furthermore, the conductive frame is a rigid structure, reducing the risk of accidental contact with other conductive structures due to erratic movement. Additionally, because the conductive frame is less prone to accidental contact with other conductive structures, an insulating layer is not required on its outer side, allowing for higher heat dissipation performance. Moreover, the absence of an insulating layer on the outer side of the conductive frame allows for a larger current-carrying area, facilitating the supply of power to higher-power open acceleration modules.

[0018] In some possible implementations, the open acceleration module includes a second connector disposed on the first circuit board of the open acceleration module and electrically connected to the processor of the open acceleration module. A conductive element is located outside the first circuit board and the processor, and is connected to the second connector. A power connector is electrically connected to the second connector via the conductive element. This facilitates the installation and removal of the conductive element from the open acceleration module.

[0019] In some possible implementations, the open acceleration module also includes an open-loop power supply, which is located on the first circuit board and electrically connected to a second connector. The processor is also electrically connected to the open-loop power supply, and the processor is connected to the second connector via the open-loop power supply. This open-loop power supply facilitates the conversion of current from the power connector into the current required for the operation of various devices in the open acceleration module, such as the processor. This allows the power connector to directly supply power to the various devices in the open acceleration module via conductive components. When the power supply is open-loop, its structure is relatively simple, it occupies less space, and it is easy to arrange on a space-constrained first circuit board. Furthermore, because the open-loop power supply has a fixed conversion ratio, its operating efficiency is high, resulting in high efficiency in supplying power to the various devices in the open acceleration module, such as the processor. Moreover, the cost of the open-loop power supply is also relatively low.

[0020] In some possible implementations, the open-loop power supply and the first connector of the open acceleration module are located on different sides of the processor. This minimizes the mutual interference between the placement of the open-loop power supply and the placement of the first connector. Furthermore, the routing within the first circuit board is easier when both the open-loop power supply and the first connector are electrically connected to the processor via traces within the first circuit board.

[0021] A second aspect of this application provides a server that includes the general-purpose substrate described in any of the above embodiments. Attached Figure Description

[0022] Figure 1 A schematic diagram of a server provided for an embodiment of this application;

[0023] Figure 2 A schematic diagram of the layout of a general-purpose substrate provided in an embodiment of this application;

[0024] Figure 3 A schematic diagram of a general-purpose substrate provided in an embodiment of this application;

[0025] Figure 4 This application provides a schematic diagram of the layout of an open acceleration module.

[0026] Figure 5 This is a schematic diagram of another server provided in an embodiment of this application.

[0027] Explanation of reference numerals in the attached figures:

[0028] 10. Computing node; 11. General-purpose baseboard; 20. Switching node; 21. Switching chip; 22. Third connector; 23. Communication interface; 24. Third circuit board; 25. Second cable; 26. Third cable; 27. Second external component; 30. Chassis;

[0029] 100. First load-bearing component; 110. First part; 120. Second part;

[0030] 200, Open acceleration module; 210, First circuit board; 220, Processor; 230, First connector; 240, Second connector; 250, Open-loop power supply; 251, Sub-power supply;

[0031] 300, First external component; 310, Second carrier; 320, High-speed connector; 330, Power connector;

[0032] 400, First Cable;

[0033] 500. Conductive components;

[0034] 600. Second circuit board. Detailed Implementation

[0035] The terminology used in the implementation section of this application is only for explaining specific embodiments of this application and is not intended to limit this application. The implementation of the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0036] This application provides a server, which may include, but is not limited to, a rack server, a high-density server, a tower server, a blade server, a full rack server, etc.

[0037] Figure 1 This is a schematic diagram of a server provided in an embodiment of this application. Figure 1 The server shown is a rack-mount server.

[0038] like Figure 1 As shown, the server includes compute nodes 10 and a chassis 30. The compute nodes 10 are located inside the chassis 30, which serves to house and protect the compute nodes 10. The chassis 30 may contain one or more compute nodes 10. In some examples, the chassis 30 may also be called a server rack (e.g., when the server is a rack-mount server).

[0039] Figure 2 This is a schematic diagram of the layout of a general-purpose substrate provided in an embodiment of this application.

[0040] like Figure 2 As shown, to meet the needs of high-performance computing scenarios such as artificial intelligence, machine learning, and data analysis, computing node 10 includes a general-purpose substrate 11. The general-purpose substrate 11 is a standardized substrate that meets the Open Compute Project specifications. The Open Compute Project specifications standardize the physical dimensions, power supply, and high-speed signals of the substrate.

[0041] For example, the server could be an artificial intelligence server.

[0042] like Figure 2 As shown, the general-purpose substrate 11 includes an open acceleration module 200 and a first external component 300. The open acceleration module 200 is an acceleration module based on the Open Compute Project standard. The open acceleration module 200 is electrically connected to the first external component 300, which is used to electrically connect to external devices independent of the computing node 10, so that external devices independent of the computing node 10 can supply power to the open acceleration module 200 and interact with the open acceleration module 200.

[0043] like Figure 2As shown, the first external component 300 includes a power connector 330, which is electrically connected to the open acceleration module 200, allowing external devices independent of the compute node 10 to supply power to the open acceleration module 200 via the power connector 330. For example, external devices independent of the compute node 10 can supply 48V or 54V current to the open acceleration module 200 via the power connector 330.

[0044] like Figure 2 As shown, the first external component 300 includes a high-speed connector 320, which is electrically connected to the open acceleration module 200, enabling the open acceleration module 200 to interact with external devices independent of the computing node 10 via the high-speed connector 320.

[0045] like Figure 2 As shown, the general-purpose substrate 11 includes multiple open acceleration modules 200. For example, the general-purpose substrate 11 may include eight open acceleration modules 200 arranged in an array. The first external component 300 includes multiple high-speed connectors 320 corresponding to the multiple open acceleration modules 200 respectively. The high-speed connectors 320 are electrically connected to the corresponding open acceleration modules 200, and the open acceleration modules 200 can interact with external devices independent of the computing node 10 through the corresponding high-speed connectors 320.

[0046] For example, each open acceleration module 200 may correspond to one or more high-speed connectors 320, and each open acceleration module 200 may be electrically connected to the corresponding one or more high-speed connectors 320.

[0047] like Figure 2 As shown, the general-purpose substrate 11 also includes a first carrier 100, on which the open acceleration module 200 is disposed. The open acceleration module 200 is fixed and supported by the first carrier 100, and is fixed to the chassis 30 via the first carrier 100. Multiple open acceleration modules 200 of the general-purpose substrate 11 can be arranged in an array on the first carrier 100. For example, when the general-purpose substrate 11 includes eight open acceleration modules 200, the eight open acceleration modules 200 can be arranged in a matrix of 4 rows and 2 columns on the first carrier 100.

[0048] like Figure 2As shown, the first external component 300 also includes a second carrier 310, which is an independent component from the first carrier 100. A high-speed connector 320 is disposed on the second carrier 310, and is fixed and supported by the second carrier 310. The high-speed connector 320 is fixed to the chassis 30 via the second carrier 310. By placing the open acceleration module 200 on the first carrier 100 and the high-speed connector 320 on the second carrier 310 (which is independent of the first carrier 100), the relative positions of the open acceleration module 200 and the high-speed connector 320 can be more flexible, and their arrangement is easier.

[0049] For example, multiple high-speed connectors 320 on the general-purpose substrate 11 can be arranged in a row.

[0050] like Figure 2 As shown, the general-purpose substrate 11 also includes a first cable 400, which is located outside the first carrier 100 and the second carrier 310. The high-speed connector 320 is electrically connected to the corresponding open acceleration module 200 through the first cable 400, and high-speed signals (e.g., SERDES signals) can be transmitted between the high-speed connector 320 and the corresponding open acceleration module 200 through the first cable 400. In this case, the high-speed connector 320 and the corresponding open acceleration module 200 do not need to transmit high-speed signals through traces within the motherboard of the general-purpose substrate 11. That is, traces for high-speed signal transmission between the high-speed connector 320 and the corresponding open acceleration module 200 do not need to be arranged within the motherboard of the general-purpose substrate 11.

[0051] Compared to the scheme where high-speed signals are transmitted through traces within the motherboard of the general-purpose substrate 11, the attenuation of high-speed signals transmitted via the first cable 400 is smaller. This results in less attenuation of the high-speed signals interacting between the open acceleration module 200 and the high-speed connector 320, thereby reducing the number of retimers required between the open acceleration module 200 and the high-speed connector 320, or even eliminating the need for retimers at all. This simplifies the transmission link of high-speed signals between the open acceleration module 200 and the high-speed connector 320, reduces the latency of high-speed signals between them, and thus facilitates an increase in the interaction rate of high-speed signals between the open acceleration module 200 and the high-speed connector 320. Reducing the number of retimers also helps to lower the cost of the general-purpose substrate 11 and reduce the power consumption of the server.

[0052] Since the high-speed connector 320 and the corresponding open acceleration module 200 are electrically connected via the first cable 400, the motherboard of the general-purpose substrate 11 does not need to have traces for high-speed signal transmission between the high-speed connector 320 and the corresponding open acceleration module 200. This reduces the wiring difficulty and complexity of the motherboard of the general-purpose substrate 11, thereby reducing the number of layers and size of the motherboard, and ultimately lowering the cost of the general-purpose substrate 11.

[0053] The open acceleration module 200 and the high-speed connector 320, which are electrically connected by the first cable 400, are respectively disposed on the first carrier 100 and the second carrier 310. At least one of the first carrier 100 and the second carrier 310 is a component independent of the motherboard of the general-purpose substrate 11. That is to say, at least one of the open acceleration module 200 and the high-speed connector 320 is not disposed on the motherboard of the general-purpose substrate 11. This helps to reduce the difficulty and complexity of the device and wiring layout of the motherboard of the general-purpose substrate 11, reduces the number of layers of the motherboard of the general-purpose substrate 11, reduces the size of the motherboard of the general-purpose substrate 11, and reduces the cost of the general-purpose substrate 11.

[0054] The open acceleration module 200 is electrically connected to the corresponding high-speed connector 320 via a first cable 400. This allows the first carrier 100 and the second carrier 310 to be exempt from circuit connection functions, enabling more flexible material selection for the first carrier 100 and the second carrier 310 and reducing the cost of the general-purpose substrate 11. Furthermore, the component supporting the high-speed connector 320 does not need to perform the function of realizing the circuit connection of the high-speed connector 320. In other words, the high-speed connector 320 does not need to achieve circuit connection through the component supporting the high-speed connector 320. This eliminates the need to use a bendable connector to electrically connect the high-speed connector 320 to the component supporting the high-speed connector 320, making the selection of the high-speed connector 320 more flexible.

[0055] For example, after the open acceleration module 200 and the high-speed connector 320 are electrically connected through the first cable 400, the open acceleration module 200 and the high-speed connector 320 can support a data transmission rate of 112Gbps PAM4 or higher, so as to meet the design requirements of servers that support data transmission rates of 112Gbps PAM4 or higher.

[0056] For example, the first cable 400 is a low-loss high-speed cable, and the first cable 400 has a loss of less than 10dB per meter when transmitting a 224Gbps PAM4 signal.

[0057] For example, the second support member 310 can be a strip-shaped plate-like structural member.

[0058] like Figure 2As shown, in some possible embodiments, the general-purpose substrate 11 further includes a second circuit board 600, which is the mainboard of the general-purpose substrate 11. The first carrier 100 includes a first portion 110 and a second portion 120. The open acceleration module 200 is disposed in the first portion 110, and the second circuit board 600 is disposed in the second portion 120. The second circuit board 600 is carried by the first carrier 100 and fixed to the chassis 30 through the first carrier 100. The first cable 400 is located outside the second circuit board 600. At this time, both the first carrier 100 and the second carrier 310 are components independent of the mainboard of the general-purpose substrate 11. That is, the open acceleration module 200 and the high-speed connector 320 are not disposed on the mainboard of the general-purpose substrate 11. In this way, the difficulty and complexity of the device and wiring layout of the mainboard of the general-purpose substrate 11 can be further reduced, which is conducive to further reducing the number of layers and the size of the mainboard of the general-purpose substrate 11, and further reducing the cost of the general-purpose substrate 11. In addition, the second circuit board 600 and the open acceleration module 200 are both located on the first carrier 100. The second circuit board 600, the open acceleration module 200 and the first carrier 100 can be pre-assembled as a whole before assembly, making the disassembly and assembly of the universal base plate 11 easier.

[0059] For example, the main board of the general-purpose substrate 11 may be provided with low-speed connectors, other logic chips and other devices. That is, when the general-purpose substrate 11 includes a second circuit board 600, the second circuit board 600 may be provided with low-speed connectors, other logic chips and other devices.

[0060] For example, the second circuit board 600 and the open module are located on the same side of the thickness direction of the first carrier 100.

[0061] For example, the first carrier 100 may be a pallet.

[0062] For example, the first support member 100 can be a rectangular structure, and the second support member 310 can be disposed on one side of the length direction of the first support member 100.

[0063] For example, the first part 110 and the second part 120 may be arranged in a row along the length of the first carrier 100.

[0064] In some other possible implementations, the second carrier 310 can be a circuit board. In this case, the second carrier 310 can be the main board of the general substrate 11, that is, the open acceleration module 200 can be disposed on the main board of the general substrate 11.

[0065] like Figure 2As shown, in some examples where the general-purpose substrate 11 includes a second circuit board 600, the open acceleration module 200 is disposed between the second circuit board 600 and the high-speed connector 320. In this case, the smaller spacing between the open acceleration module 200 and the high-speed connector 320 facilitates shortening the length of the first cable 400. Furthermore, the second circuit board 600 and the devices disposed thereon are less likely to affect the routing of the first cable 400, facilitating the arrangement of the first cable 400 and reducing its length. The shortened length of the first cable 400 results in a shorter transmission path for high-speed signals between the open acceleration module 200 and the high-speed connector 320, leading to less attenuation and higher speed in the high-speed signal interaction between them.

[0066] For example, the second circuit board 600 and the high-speed connector 320 are respectively disposed on both sides of the open acceleration module 200 in the length direction of the first carrier 100.

[0067] For example, the second carrier 310 is located on the side of the open acceleration module 200 away from the second circuit board 600.

[0068] Figure 3 This is a schematic diagram of a general-purpose substrate provided in an embodiment of this application.

[0069] like Figure 2 , Figure 3 As shown, in some possible embodiments, the power connector 330 is disposed on the second carrier 310. The universal substrate 11 also includes a conductive element 500, which is located outside the first carrier 100 and the second carrier 310. In an example where the universal substrate 11 includes a second circuit board 600, the conductive element 500 is located outside the second circuit board 600. The power connector 330 is electrically connected to the open acceleration module 200 via the conductive element 500, and the power connector 330 supplies power to the open acceleration module via the conductive element 500.

[0070] Compared to the power connector 330 which supplies power to the open acceleration module 200 via traces within the motherboard of the universal substrate 11, the conductive element 500 located outside the first carrier 100 and the second carrier 310 can have a larger current-carrying area, facilitating the flow of larger currents and enabling power supply to the higher-power open acceleration module 200. Furthermore, compared to the traces within the motherboard of the universal substrate 11, replacing the conductive element 500 located outside the first carrier 100 and the second carrier 310 is easier, allowing for flexible adjustment of the conductive element 500 according to the power of the open acceleration module 200 to adapt to different open acceleration modules 200. Additionally, compared to the traces within the motherboard of the universal substrate 11, the conductive element 500 located outside the first carrier 100 and the second carrier 310 can directly exchange heat with the external environment, resulting in higher heat dissipation efficiency and improving the stability and reliability of the universal substrate 11.

[0071] Since the power connector 330 supplies power to the open acceleration module 200 via the conductive element 500, there is no need to arrange traces for supplying power to the open acceleration module 200 within the motherboard of the universal substrate 11. This reduces the wiring difficulty and complexity of the motherboard of the universal substrate 11, thereby reducing the number of layers and the size of the motherboard, and ultimately lowering the cost of the universal substrate 11.

[0072] The open acceleration module 200 and the power connector 330 are respectively disposed on the independent first carrier 100 and the second carrier 310. At this time, at least one of the open acceleration module 200 and the power connector 330 is not disposed on the motherboard of the general substrate 11. This helps to reduce the difficulty and complexity of the device and wiring layout of the motherboard of the general substrate 11, reduces the number of layers of the motherboard of the general substrate 11, reduces the size of the motherboard of the general substrate 11, and reduces the cost of the general substrate 11.

[0073] The open acceleration module 200 is electrically connected to the power connector 330 via a conductive component 500. This allows the first carrier 100 and the second carrier 310 to be exempt from circuit connection functions, enabling greater flexibility in material selection for the first carrier 100 and the second carrier 310 and reducing the cost of the general-purpose substrate 11. Furthermore, the component carrying the power connector 330 does not need to perform the function of realizing the circuit connection of the power connector 330. In other words, the power connector 330 does not need to achieve circuit connection through the component carrying the power connector 330. This eliminates the need to use a bendable connector for the power connector 330 to electrically connect to the component carrying the power connector 330, making the selection of the power connector 330 more flexible.

[0074] For example, multiple open acceleration modules 200 can be connected to the same power connector 330 via conductive elements 500.

[0075] For example, the same conductive element 500 can be connected to multiple open acceleration modules 200.

[0076] In some examples, the conductive element 500 is a wire.

[0077] like Figure 3 As shown, in some other examples, the conductive element 500 is a conductive frame, a rigid structural component. The conductive frame is fixed and electrically connected to both the power connector 330 and the open acceleration module 200. This makes it convenient to connect multiple open acceleration modules 200 to the power connector 330 after the relative positions of the open acceleration module 200 and the power connector 330 are fixed. Furthermore, the rigid structure of the conductive frame prevents accidental contact with other conductive structures due to erratic movement. Additionally, because the conductive frame is less prone to accidental contact with other conductive structures, an insulating layer is not required on the outside of the conductive frame, allowing for higher heat dissipation performance. Moreover, since an insulating layer is not needed on the outside of the conductive frame, the conductive element 500 can have a larger current-carrying area, which is beneficial for powering higher-power open acceleration modules 200.

[0078] For example, the conductive component 500 can be made of metal materials such as copper or aluminum. For instance, when the conductive component 500 is a conductive frame, the conductive frame can be a copper frame.

[0079] like Figure 3 As shown, the open acceleration module 200 includes a first circuit board 210 and a processor 220. The first circuit board 210 is disposed on the first carrier 100, and the processor 220 is disposed on the first circuit board 210. The processor 220 can be used for computation. For example, the processor 220 can be a graphics processing unit (GPU).

[0080] In some possible implementations, the open acceleration module 200 further includes a second connector 240. The second connector 240 is located on the first circuit board 210 and is electrically connected to the processor 220. A conductive element 500 is located outside the first circuit board 210 and the processor 220, and is connected to the second connector 240. A power connector 330 is electrically connected to the second connector 240 via the conductive element 500. This facilitates the installation and removal of the conductive element 500 from the open acceleration module 200.

[0081] For example, a portion of the conductive element 500 may be inserted into the second connector 240, and the conductive element 500 and the second connector 240 are pluggable.

[0082] For example, the second connector 240 may be a duckbill connector. Specifically, the second connector 240 includes a first clamping portion and a second clamping portion, a clamping space is formed between the first clamping portion and the second clamping portion, a portion of the conductive element 500 is inserted into the clamping space and clamped by the first clamping portion and the second clamping portion, and the conductive element 500 is electrically connected to at least one of the first clamping portion and the second clamping portion.

[0083] In some examples, the second connector 240 can be electrically connected to the processor 220 via a cable outside the first circuit board 210.

[0084] In other examples, the second connector 240 and the processor 220 can be electrically connected through the first circuit board 210, in which case both the second connector 240 and the processor 220 are electrically connected to the first circuit board 210.

[0085] Figure 4 This is a schematic diagram of the layout of an open acceleration module provided in an embodiment of this application.

[0086] like Figure 4 As shown, in some possible implementations, the open acceleration module 200 also includes a power supply, which is a device for current conversion. In some examples, the power supply can be an open-loop power supply 250, which is a current conversion device with a fixed conversion ratio. The open-loop power supply 250 is located on the first circuit board 210 and is electrically connected to the second connector 240. The processor 220 is also electrically connected to the open-loop power supply 250, and the processor 220 is electrically connected to the second connector 240 through the open-loop power supply 250. In this way, through power conversion, the current from the power connector 330 is easily converted into the current required for the operation of the various devices of the open acceleration module 200, such as the processor 220, thereby facilitating the direct supply of power from the power connector 330 to the various devices of the open acceleration module 200, such as the processor 220, through the conductive element 500. When the power supply is the open-loop power supply 250, the structure of the open-loop power supply 250 is relatively simple, occupies less space, and is easy to arrange on the relatively small first circuit board 210. Furthermore, because the open-loop power supply 250 has a fixed conversion ratio, its operating efficiency is high, resulting in efficient power supply to the various devices in the open acceleration module 200, such as the processor 220. Moreover, the cost of the open-loop power supply 250 is also relatively low.

[0087] In some examples, the second connector 240 can be electrically connected to the power supply via a cable outside the first circuit board 210.

[0088] In other examples, the second connector 240 and the power supply can be electrically connected through the first circuit board 210, in which case both the second connector 240 and the power supply are electrically connected to the first circuit board 210.

[0089] In some examples, the processor 220 and the power supply can be electrically connected via cables outside the first circuit board 210.

[0090] In other examples, the processor 220 and the power supply can be electrically connected through the first circuit board 210, in which case both the processor 220 and the power supply are electrically connected to the first circuit board 210.

[0091] For example, the open-loop power supply 250 may include one or more sub-power supplies 251, the second connector 240 is electrically connected to all sub-power supplies 251, and the processor 220 is electrically connected to at least one sub-power supply 251.

[0092] For example, sub-power supply 251 can be a multi-phase power supply.

[0093] For example, sub-power supply 251 can be an unregulated bus converter.

[0094] like Figure 4 As shown, in some possible embodiments, the open acceleration module 200 further includes a first connector 230, which is disposed on the first circuit board 210, and the processor 220 is electrically connected to the first connector 230. A first cable 400 is located outside the first circuit board 210 and the processor 220, and is connected to the first connector 230. A high-speed connector 320 is electrically connected to the first connector 230 via the first cable 400. This facilitates the connection and disconnection of the first cable 400 from the open acceleration module 200.

[0095] For example, the first connector 230 may be a board-to-board connector (BTB).

[0096] For example, the pitch between two adjacent contact points (e.g., pins or pads) of the first connector 230 can be greater than 0.6 mm and less than 1 mm, making the first connector 230 more compact and saving space. In addition, the first connector 230 also has better signal integrity and electrical performance.

[0097] For example, the center-to-center distance between two adjacent contact points of the first connector 230 is approximately 0.8 mm. For instance, the center-to-center distance (pitch) between two adjacent contact points of the first connector 230 can be 0.7 mm, 0.75 mm, 0.8 mm, 0.85 mm, 0.9 mm, etc. This allows the first connector 230 to have a small size while ensuring electrical performance and signal transmission integrity, facilitating its placement on the space-constrained first circuit board 210.

[0098] For example, a plurality of first connectors 230 may be provided on the first circuit board 210, and the processor 220 may be electrically connected to the plurality of first connectors 230.

[0099] For example, multiple first connectors 230 can be arranged in a row.

[0100] For example, an array of multiple first connectors 230 is arranged in a high-density orthogonal layout on a first circuit board 210.

[0101] For example, the first connector 230 is located between the processor 220 and the high-speed connector 320. In this case, the distance between the first connector 230 and the high-speed connector 320 is small, which helps to shorten the length of the first cable 400. Furthermore, the processor 220 is less likely to affect the routing of the first cable 400, facilitating the arrangement of the first cable 400 and reducing its length. The shortened length of the first cable 400 results in a shorter transmission path for high-speed signals between the first connector 230 and the high-speed connector 320, leading to less attenuation and a higher speed for high-speed signal interaction between them.

[0102] For example, the processor 220 and the high-speed connector 320 may be respectively located on both sides of the first connector 230 in the length direction of the first carrier 100.

[0103] For example, the second carrier 310 is located on the side of the first connector 230 away from the processor 220.

[0104] In some possible implementations, the open-loop power supply 250 and the first connector 230 are located on different sides of the processor 220.

[0105] In this way, the mutual influence between the arrangement of the open-loop power supply 250 and the arrangement of the first connector 230 is small. When both the open-loop power supply 250 and the first connector 230 are electrically connected to the processor 220 through traces in the first circuit board 210, the arrangement of traces in the first circuit board 210 is relatively easy.

[0106] For example, the processor 220 is disposed within the space formed by the open-loop power supply 250 and the first connector 230. For instance, a plurality of sub-power supplies 251 and a plurality of first connectors 230 are arranged circumferentially around the processor 220, and the processor 220 is disposed within the space formed by the plurality of sub-power supplies 251 and the plurality of first connectors 230.

[0107] For example, the first circuit board 210 can be a high-density interconnect printed circuit board (HDIPCB) to facilitate a high-density layout of the traces on the first circuit board 210, so as to arrange devices such as the first connector 230, the second connector 240, and the open-loop power supply 250 on the relatively small space of the first circuit board 210. For example, the first circuit board 210 can be a 4th-order high-density interconnect printed circuit board, a 5th-order high-density interconnect printed circuit board, or a 6th-order high-density interconnect printed circuit board.

[0108] In some examples, compute node 10 may include a housing, within which first carrier 100, second carrier 310, open acceleration module 200, and second circuit board 600 may be housed. A portion of first external component 300 may be housed within the housing for electrical connection to the open acceleration module 200 within the housing, while a portion of the first external component 300 may be located outside the housing for connection to external devices independent of compute node 10. The first carrier 100, second carrier 310, open acceleration module 200, second circuit board 600, and first external component 300 may be connected to chassis 30 via the housing.

[0109] In other examples, the compute node 10 may not include a housing, and the first carrier 100 and the second carrier 310 may be directly connected to the chassis 30.

[0110] In some examples, the second carrier 310 and the first carrier 100 can be fixedly connected by a connector. In examples where the computing node 10 includes a housing, the connector can be the housing itself. In examples where the computing node 10 does not include a housing, the connector can be a connecting frame or other structural components. This facilitates flexible insertion and removal of the entire universal substrate 11.

[0111] In other examples, the second carrier 310 and the first carrier 100 may also be fixedly connected to the chassis 30 independently.

[0112] Figure 5 This is a schematic diagram of another server provided in an embodiment of this application.

[0113] like Figure 5As shown, in some examples, the server also includes a switching node 20, which can be located inside the chassis 30. In this case, the server can be a rack-mount server, etc. A high-speed connector 320 can be electrically connected to the switching node 20, enabling the computing node 10 to interact with the switching node 20 via high-speed signals. Specifically, the switching node 20 includes a switching chip 21, a second external component 27, and a third circuit board 24. The switching chip 21 is located on the third circuit board 24 and is electrically connected to the second external component 27. The high-speed connector 320 is also connected to the second external component 27.

[0114] For example, the high-speed connector 320 can be directly mated with the second external component 27, or the high-speed connector 320 and the second external component 27 can be electrically connected via a backplane.

[0115] In some examples, the switching node 20 also includes a third connector 22 and a second cable 25. The third connector 22 is located on the third circuit board 24 and is electrically connected to the switching chip 21. The second cable 25 is located outside the third circuit board 24 and is electrically connected to the second external component 27 through the second cable 25. This reduces the attenuation of the high-speed signal between the switching chip 21 and the second external component 27, simplifies the transmission link of the high-speed signal between the switching chip 21 and the second external component 27, reduces the delay of the high-speed signal between the switching chip 21 and the second external component 27, and thus helps to improve the interaction rate of the high-speed signal between the switching chip 21 and the second external component 27.

[0116] For example, the third connector 22 can be a board-to-board connector.

[0117] In some examples, the third connector 22 and the switching chip 21 can be electrically connected through the third circuit board 24. In this case, both the third connector 22 and the switching chip 21 are electrically connected to the third circuit board 24.

[0118] In other examples, the third connector 22 and the switching chip 21 can be electrically connected via a cable located outside the third circuit board 24.

[0119] like Figure 5As shown, in some examples, the switching node 20 also includes a communication interface 23 and a third cable 26. The third cable 26 is located outside the third circuit board 24. The communication interface 23 is electrically connected to the third connector 22 through the third cable 26. This reduces the attenuation of the high-speed signal between the switching chip 21 and the communication interface 23, simplifies the transmission link of the high-speed signal between the switching chip 21 and the communication interface 23, reduces the delay of the high-speed signal between the switching chip 21 and the communication interface 23, and thus helps to improve the interaction rate of the high-speed signal between the switching chip 21 and the communication interface 23.

[0120] For example, communication interface 23 can be an optical module interface.

[0121] In other examples, the server may not include switch node 20; in this case, the server may be a rack server, etc.

[0122] In the description of the embodiments of this application, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, an indirect connection through an intermediate medium, or the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this application according to the specific circumstances.

[0123] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of this application, and are not intended to limit them. Although the embodiments of this application have been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A universal substrate (11), characterized in that, include: First load-bearing component (100); An open acceleration module (200) is provided on the first support member (100); Second bearing member (310); A high-speed connector (320) is disposed on the second carrier (310); A first cable (400) is located outside the first carrier (100) and the second carrier (310), and the high-speed connector (320) is electrically connected to the open acceleration module (200) via the first cable (400).

2. The universal substrate (11) according to claim 1, characterized in that, The open acceleration module (200) includes a first circuit board (210), a processor (220), and a first connector (230); The first circuit board (210) is disposed on the first carrier (100), the processor (220) and the first connector (230) are disposed on the first circuit board (210), and the processor (220) is electrically connected to the first connector (230); The first cable (400) is located outside the first circuit board (210) and the processor (220), and the first cable (400) is connected to the first connector (230). The high-speed connector (320) is electrically connected to the first connector (230) through the first cable (400).

3. The universal substrate (11) according to claim 1 or 2, characterized in that, It also includes a second circuit board (600); The first carrier (100) includes a first part (110) and a second part (120), the open acceleration module (200) is disposed in the first part (110), and the second circuit board (600) is disposed in the second part (120); The first cable (400) is located outside the second circuit board (600).

4. The universal substrate (11) according to claim 3, characterized in that, The open acceleration module (200) is located between the second circuit board (600) and the high-speed connector (320).

5. The universal substrate (11) according to any one of claims 1-4, characterized in that, It also includes a power connector (330) and conductive components (500); The power connector (330) is disposed on the second carrier (310); The conductive element (500) is located outside the first carrier (100) and the second carrier (310), and the power connector (330) is electrically connected to the open acceleration module (200) through the conductive element (500).

6. The universal substrate (11) according to claim 5, characterized in that, The conductive component (500) is a conductive frame, which is fixed and electrically connected to both the power connector (330) and the open acceleration module (200).

7. The universal substrate (11) according to claim 5 or 6, characterized in that, The open acceleration module (200) includes a second connector (240); The second connector (240) is disposed on the first circuit board (210) of the open acceleration module (200), and the second connector (240) is electrically connected to the processor (220) of the open acceleration module (200); The conductive element (500) is located outside the first circuit board (210) and the processor (220). The conductive element (500) is connected to the second connector (240). The power connector (330) is electrically connected to the second connector (240) through the conductive element (500).

8. The universal substrate (11) according to claim 7, characterized in that, The open acceleration module (200) also includes an open-loop power supply (250); The open-loop power supply (250) is located on the first circuit board (210). The open-loop power supply (250) is electrically connected to the second connector (240). The processor (220) is electrically connected to the open-loop power supply (250). The processor (220) is electrically connected to the second connector (240) through the open-loop power supply (250).

9. The universal substrate (11) according to claim 8, characterized in that, The first connector (230) of the open-loop power supply (250) and the open acceleration module (200) is located on different sides of the processor (220).

10. A server, characterized in that, Includes the general-purpose substrate (11) as described in any one of claims 1-9.

Citation Information

Cited By

  • An artificial intelligence server node and an artificial intelligence server

    CN122470556A