GPU direct connection substrate

Abstract

The utility model provides a GPU direct connection substrate. The surface of the direct connection substrate is provided with a high-precision slot and a high-speed precision direct connection port. Wherein the high-precision slot is used for directly inserting a GPU on the direct-connection substrate, the high-precision slot is provided with 12 PCIe 5.0 slots, the bandwidth of 10 PCIe slots is x16, the bandwidth of 2 PCIe slots is x8, the slot bandwidth x16 is suitable for being inserted into a single-width board card with the bandwidth of x8 or a single-width board card with the bandwidth of x16 and a double-width board card with the bandwidth of x16, and the slot bandwidth x8 is suitable for being inserted into the single-width board card with the bandwidth of x8; and the high-speed precise direct connection port is used for tightly connecting the direct connection substrate with a main board. Through the innovative design, the problems of bandwidth, delay, power consumption, heat dissipation and the like of a traditional GPU connection mode are effectively solved, and powerful technical support is provided for development of the high-performance computing field.

Description

TECHNICAL FIELD

[0001] The utility model relates to computer and server hardware technical field especially to integrated circuit board design technical field, concretely relates to a GPU direct connection substrate, aims at improving the data transmission efficiency and stability between GPU and mainboard. BACKGROUND

[0002] GPU (graphics processing unit) as the core component responsible for graphics rendering and parallel computing in computer system, its performance requirement is increasing. GPU direct connection substrate usually refers to the GPU board card through some kind of efficient, direct connection mode, such as high-speed bus, special interface and the like and the connecting method of the device connected with the substrate.

[0003] Traditional GPU is connected with mainboard through PCIe and the like, although it can satisfy general application demand, but in high load scene, interface bandwidth limitation and signal attenuation problem become the key factor restricting the overall performance of system.

[0004] Therefore, developing a kind of technology architecture that can directly, efficiently connect GPU and mainboard has important significance for improving system performance, expansibility, flexible application, reducing delay. SUMMARY

[0005] In view of the deficiencies of prior art, the utility model provides a kind of GPU direct connection substrate, through the innovative design, effectively solve the problem of traditional GPU connection mode in bandwidth, delay, power consumption and heat dissipation etc., provide strong technical support for the development of high-performance computing field.The utility model has significant technical progress and wide application prospect.

[0006] The technical scheme of the utility model is as follows:

[0007] A kind of GPU direct connection substrate, the direct connection substrate surface is provided with high-precision slot and high-speed precision direct connection port;Wherein, the high-precision slot is used for GPU directly inserted on the direct connection substrate;The high-precision slot designs 12 PCIe 5.0 slots, wherein 10 PCIe slots bandwidth is x16, 2 PCIe slots bandwidth is x8, slot bandwidth x16 is suitable for inserting bandwidth x8 single-width board card, or bandwidth x16 single-width board card and x16 double-width board card, slot bandwidth x8 is suitable for inserting bandwidth x8 single-width board card;The high-speed precision direct connection port is used for the direct connection substrate and mainboard close connection;And the direct connection substrate uses complex programmable logic device CPLD control management PCIe in-position state signal, reset signal and IIC data interaction.

[0008] In some realizable embodiments, the direct connection substrate adopts multilayer high-density wiring, and built-in high-speed signal transmission line, ensures the high-speed, stable transmission of data between GPU and mainboard.

[0009] In some implementable embodiments, the high-speed precision direct connection interface is based on the latest PCIe 5.0 communication protocol standard, uses differential signal transmission internally, has higher bandwidth and lower delay characteristics, effectively suppresses electromagnetic interference, and improves signal integrity.

[0010] In some implementable embodiments, the direct connection board is fixed by screws of the fixed board card and combined with the sheet metal part in the case when in use.

[0011] In some implementable embodiments, the 12 PCIe 5.0 slots on the direct connection board are arranged as follows: SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, SLT#4 PCIe x16, SLT#8 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, and SLT#12 PCIe x16 are double-wide size design, and SLT#5 PCIe x8, SLT#6 PCIe x8, and SLT#7 PCIe x16 are single-wide size design.

[0012] In some implementable embodiments, SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, and SLT#4 PCIe x16 are arranged on one side of the direct connection board, SLT#8 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, and SLT#12 PCIe x16 are arranged on the opposite side of the direct connection board, and SLT#5 PCIe x8, SLT#6 PCIe x8, and SLT#7 PCIe x16 are arranged in the middle of the direct connection board.

[0013] In some implementable embodiments, the direct connection board is also provided with four fan connectors and three temperature sensors to detect the temperature of the board in real time and adjust the number of revolutions of the fan, so as to ensure that the GPU can still maintain a lower working temperature when running for a long time under high load.

[0014] In some implementable embodiments, the direct-attached board connects 8 GPU double-wide board cards, the GPU double-wide board cards are inserted into the corresponding slots of SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, SLT#4 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, SLT#12 PCIe x16, and other type single-wide board cards are inserted into SLT#5 PCIe x8, SLT#6 PCIe x8, SLT#7 PCIe x16, SLT#8 PCIe x16.

[0015] In some implementable embodiments, the direct-attached board connects 9 GPU double-wide board cards, the GPU double-wide board cards are inserted into the corresponding slots of SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, SLT#4 PCIe x16, SLT#8 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, SLT#12 PCIe x16, and other type single-wide board cards are inserted into SLT#5 PCIe x8, SLT#6 PCIe x8, SLT#7 PCIe x16.

[0016] In some implementable embodiments, the direct-attached board connects 10 GPU double-wide board cards, the GPU double-wide board cards are inserted into the corresponding slots of SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, SLT#4 PCIe x16, SLT#7 PCIe x16, SLT#8 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, SLT#12 PCIe x16, and other type single-wide board cards are inserted into SLT#5 PCIe x8.

[0017] The GPU direct connection substrate provided by the utility model has the advantages and effects that the GPU direct connection substrate can directly and efficiently connect the GPU and the mainboard, and has important significance for improving system performance, expansibility, flexible application and reducing delay. The GPU direct connection substrate provided by the utility model can ensure the accuracy and reliability of connection, effectively suppress electromagnetic interference and improve signal integrity. The GPU direct connection substrate provided by the utility model effectively solves the problems of the traditional GPU connection mode in bandwidth, delay, power consumption and heat dissipation and the like, provides strong hardware support for high-performance computing, professional graphic design, game entertainment and artificial intelligence and the like, and has remarkable technical progressiveness and wide application prospect. It should be understood that the implementation of any embodiment of the utility model does not mean that multiple or all of the above-mentioned advantages are simultaneously achieved or reached. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the utility model or the technical solutions in the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and those skilled in the art can also obtain other implementation drawings according to the provided drawings without creating creative labor.

[0019] The structures, proportions, sizes and the like shown in the specification are only used to cooperate with the content disclosed in the specification, to enable those skilled in the art to understand and read, and are not used to limit the implementation conditions of the utility model, so they do not have technical substantive significance. Any modification of structure, change of proportion relationship or adjustment of size, without affecting the effects and purposes that the utility model can produce, should still fall within the scope of the technical content disclosed by the utility model.

[0020] Figure 1 An example of the principle block diagram of the GPU direct connection substrate provided by the utility model embodiment is shown;

[0021] Figure 2 An example of the structural design diagram of the GPU direct connection substrate provided by the utility model embodiment is shown.

[0022] Markings in the figure:

[0023] 1, SLT#1 PCIe x16;

[0024] 2, SLT#2 PCIe x16;

[0025] 3, SLT#3 PCIe x16;

[0026] 4. SLT #4 PCIe x16;

[0027] 5. SLT #5 PCIe x8;

[0028] 6. SLT #6 PCIe x8;

[0029] 7. SLT #7 PCIe x16;

[0030] 8. SLT #8 PCIe x16;

[0031] 9. SLT #9 PCIe x16;

[0032] 10. SLT #10 PCIe x16;

[0033] 11. SLT #11 PCIe x16;

[0034] 12. SLT #12 PCIe x16;

[0035] 13. MCIO x8 x11;

[0036] 14. Fan connector.

[0037] The same or corresponding reference signs in the figures indicate the same or corresponding parts. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical scheme and advantages of the embodiments of the utility model clearer and more apparent, the embodiments of the utility model will be further described in detail below in combination with the drawings. Herein, the illustrative embodiments of the utility model and their descriptions are used to explain the utility model, but not as a limitation to the utility model.

[0039] In the utility model, unless explicitly defined and limited, the terms "mounting", "connection", "connecting", "fixing" and other terms should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integrated; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through intermediate medium, it can be the communication inside two elements or the interaction relationship between two elements. For ordinary skilled in the art, the specific meaning of the above terms in the utility model can be understood according to specific circumstances.

[0040] It should be understood that the terms "include", "comprise", "consist of" or any other variant thereof are intended to cover a non-exclusive inclusion, such that a product, device, process or method that includes a list of elements is not necessarily limited to those elements, but can include other elements not expressly listed or inherent to such product, device, process or method. Without more limitations, the elements defined by the statement "include", "comprise", "consist of" do not exclude the presence of additional identical elements in the product, device, process or method that includes the said elements.

[0041] It should also be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device, component or structure referred to must have a particular orientation, be constructed or operated in a particular orientation, and cannot be understood as a limitation on the present application.

[0042] In addition, the terms "first" and "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0043] Noun explanation:

[0044] GPU, Graphics Processing Unit (GPU);

[0045] PCIe, peripheral component interconnect express (PCIe) high-speed serial computer expansion bus standard.

[0046] GPU (Graphics Processing Unit) direct connection board generally refers to the technology that GPU board card is connected with the board through some efficient and direct connection mode, such as high-speed bus, special interface, etc. The GPU board of the present application adopts PCIe (Peripheral Component Interconnect Express) high-speed bus mode. This connection mode aims to reduce the delay and loss of signal transmission, improve the efficiency and stability of data transmission, and further improve the performance and efficiency of GPU. Through innovative design, the present application realizes the direct, high-speed and low-delay connection between GPU and motherboard, providing strong hardware support for high-performance computing, professional graphic design, game entertainment and artificial intelligence and other fields.

[0047] The implementation of the utility model is described in detail in combination with the preferred embodiments.

[0048] As Figures 1-2 , Figure 1 A principle block diagram of a GPU direct connection substrate provided by an embodiment of the utility model is shown, Figure 2 A structural design diagram of a GPU direct connection substrate provided by an embodiment of the utility model is shown.

[0049] The GPU direct connection substrate provided by the utility model adopts multi-layer high-density wiring in the design of the substrate body, has built-in high-speed signal transmission lines, and ensures high-speed and stable data transmission between the GPU and the mainboard.

[0050] In order to realize direct, high-speed and low-delay connection between the GPU and the mainboard, reduce the delay and loss of signal transmission, and improve the efficiency and stability of data transmission, the utility model is provided with high-precision slots and high-speed precise direct connection interfaces on the surface of the direct connection substrate.

[0051] The high-precision slots are, for example Figure 1 , Figure 2 SLT#1 PCIe x16 1 and the like, which are used for directly plugging the GPU on the direct connection substrate, and ensure the accuracy and reliability of the connection; as shown in Figure 1 , the high-precision slots are designed with 12 PCIe 5.0 slots, of which 10 PCIe slots have a bandwidth of x16, and 2 PCIe slots have a bandwidth of x8; the slot bandwidth x16 is suitable for inserting a single-width board card with a bandwidth of x8, or a single-width board card with a bandwidth of x16 and a double-width board card with a bandwidth of x16; and the slot bandwidth x8 is suitable for inserting a single-width board card with a bandwidth of x8.

[0052] Specifically, the structural layout of the 12 PCIe 5.0 slots on the direct connection substrate is as follows: SLT#1 PCIe x16 1, SLT#2 PCIe x16 2, SLT#3 PCIe x16 3, SLT#4 PCIe x16 4, SLT#8 PCIe x16 8, SLT#9 PCIe x16 9, SLT#10 PCIe x16 10, SLT#11 PCIe x16 11, and SLT#12 PCIe x16 12 are designed with a double-width size, and SLT#5 PCIe x8 5, SLT#6 PCIe x8 6, and SLT#7 PCIe x16 7 are designed with a single-width size.

[0053] Further referring to Figure 2More specifically, SLT#1 PCIe x16 1, SLT#2 PCIe x16 2, SLT#3 PCIe x16 3, SLT#4 PCIe x16 4 are arranged on one side (left side in the figure) of the direct connection board, SLT#8 PCIe x16 8, SLT#9 PCIe x16 9, SLT#10 PCIe x16 10, SLT#11 PCIe x16 11, SLT#12 PCIe x16 12 are arranged on the other side (right side in the figure) of the direct connection board, and SLT#5 PCIe x8 5, SLT#6 PCIe x8 6, SLT#7 PCIe x8 7 are arranged in the middle of the direct connection board. According to the layout and wiring of the whole board, the x16 is distributed on both sides, which can balance the gravity and is more conducive to heat dissipation.

[0054] For high-speed precision direct connection interfaces such as MCIO x8 x11 13 in Figure 2 , which are used for the close connection of the direct connection board and the mainboard. The utility model designs a domestic high-speed precision direct connection interface, which is based on the latest PCIe 5.0 communication protocol standard and has higher bandwidth and lower delay characteristics. The interface supports the impedance of differential signals, and the differential signal transmission technology is used inside the interface to effectively suppress electromagnetic interference and improve signal integrity.

[0055] In the utility model, the direct connection board uses a complex programmable logic device CPLD to control and manage the in-place state signal, reset signal and IIC data interaction of PCIe. As shown in Figure 1 , the CPLD is used to control and manage the IIC signal, RESET and PRSNT signal of PCIe; the external connector BMC / IIC interacts with the CPLD data.

[0056] In some implementable embodiments, as shown in Figure 1 , the GPU direct connection board is designed with 4 fan connectors 14 and 3 temperature sensors, the 4 fan connectors 14 can detect the GPU board temperature in real time, and adjust the fan speed according to the temperature, so as to ensure that the GPU can still maintain a low working temperature when running for a long time under high load, thereby improving the performance and reducing the power consumption.

[0057] The GPU direct connection board of the utility model is used by inserting the GPU directly into the corresponding slot of the direct connection board, and fixed by a fastening mechanism (such as screws for fixing the board card and inner panel metal parts of the case), and then the direct connection board is closely connected with the mainboard through the direct connection interface. During the connection process, the system automatically identifies and configures the corresponding communication protocol to realize seamless connection.

[0058] The GPU direct connection board provided by the utility model mainly includes the following connection architecture in specific use scenarios:

[0059] (1) Using 8 GPU double-wide board cards, at this time, the GPU double-wide board cards are inserted into the corresponding slots of SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, SLT#4 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, and SLT#12 PCIe x16, and SLT#5 PCIe x8, SLT#6 PCIe x8, SLT#7 PCIe x16, and SLT#8 PCIe x16 can be inserted into other types of single-wide board cards.

[0060] Note:

[0061] SLT#5 PCIe x8 and SLT#6 PCIe x8 can only be inserted into x8 single-wide board cards.

[0062] SLT#7 PCIe x16 and SLT#8 PCIe x16 can be inserted into x8 single-wide board cards, or x16 single-wide board cards and x16 double-wide board cards; and in combination with Figure 1 It can be seen that:

[0063] When an x8 single-wide board card is inserted, the MCIO cable only needs to connect MCIO x8 SLT#7-0 and MCIO x8 SLT#8-0 to realize the current bandwidth usage.

[0064] When an x16 single-wide board card is inserted, the MCIO cable needs to connect MCIO x8 SLT#7-0 and MCIO x8 SLT#7-1, and MCIO x8 SLT#8-0 and MCIO x8 SLT#8-1 to realize the current bandwidth usage.

[0065] When an x16 double-wide board card is inserted, that is, in the following usage scenarios 2 and 3, it is applied when the GPU board card is expanded.

[0066] (2) Using 9 GPU double-wide board cards, at this time, the GPU double-wide board cards are inserted into the corresponding slots of SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, SLT#4 PCIe x16, SLT#8 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, and SLT#12 PCIe x16, and SLT#5 PCIe x8, SLT#6 PCIe x8, SLT#7 PCIe x16 can be inserted into other types of single-wide board cards.

[0067] (3) using 10 GPU double-wide board cards, at this time, the GPU double-wide board cards are inserted into the corresponding slots of SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, SLT#4 PCIe x16, SLT#7 PCIe x16, SLT#8 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, SLT#12 PCIe x16, and SLT#5 PCIe x8 can be inserted into other types of single-wide board cards.

[0068] The above has described various embodiments of the present application, the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes are obvious to those skilled in the art without departing from the scope and spirit of the described embodiments.

[0069] Although several implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present application. Certain features described in the context of separate embodiments can also be implemented in combination. Conversely, various features described in the context of a single embodiment can also be implemented separately or in any suitable subcombination.

Claims

1. A GPU direct-connect substrate, characterized in that, The surface of the direct-connect substrate is provided with high-precision slots and high-speed precision direct-connect interfaces. The high-precision slots are used for direct GPU insertion onto the substrate. These slots feature 12 PCIe 5.0 slots: 10 with a bandwidth of x16 and 2 with a bandwidth of x8. The x16 slots are suitable for inserting single-width x8 cards, or both x16 single-width and x16 double-width cards. The x8 slots are suitable for inserting single-width x8 cards. The high-speed precision direct-connect interfaces ensure a tight connection between the substrate and the motherboard. Furthermore, the substrate uses a complex programmable logic device (CPLD) to control and manage the PCIe presence status signals, reset signals, and IIC data interaction.

2. The GPU direct-connect substrate according to claim 1, characterized in that, The direct-connect substrate uses multi-layer high-density wiring and has built-in high-speed signal transmission lines to ensure high-speed and stable data transmission between the GPU and the motherboard.

3. The GPU direct-connect substrate according to claim 1, characterized in that, The high-speed precision direct connection interface is based on the latest PCIe 5.0 communication protocol standard and uses differential signal transmission internally to achieve higher bandwidth and lower latency, while effectively suppressing electromagnetic interference and improving signal integrity.

4. The GPU direct-connect substrate according to claim 1, characterized in that, The direct-connect base plate is fixed in use by screws on the fixing board and by the sheet metal parts inside the chassis.

5. The GPU direct-connect substrate according to claim 1, characterized in that, The 12 PCIe 5.0 slots on this direct-connect substrate are arranged as follows: SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, SLT#4 PCIe x16, SLT#8 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, and SLT#12 PCIe x16 are double-width designs, while SLT#5 PCIe x8, SLT#6 PCIe x8, and SLT#7 PCIe x16 are single-width designs.

6. The GPU direct-connect substrate according to claim 5, characterized in that, SLT#1 PCIe x16, SLT#2 PCIe x16, SLT#3 PCIe x16, and SLT#4 PCIe x16 are arranged on one side of the direct-connect substrate, SLT#8 PCIe x16, SLT#9 PCIe x16, SLT#10 PCIe x16, SLT#11 PCIe x16, and SLT#12 PCIe x16 are arranged on the opposite side of the direct-connect substrate, and SLT#5 PCIe x8, SLT#6 PCIe x8, and SLT#7 PCIe x16 are arranged in the middle of the direct-connect substrate.

7. The GPU direct-connect substrate according to claim 1, characterized in that, The direct-connect substrate is also equipped with four fan connectors and three temperature sensors to detect the substrate temperature in real time and adjust the fan speed to ensure that the GPU can maintain a low operating temperature even when running under high load for a long time.

8. The GPU direct-connect substrate according to any one of claims 1 to 7, characterized in that, The direct-connect substrate connects eight GPU dual-width cards. GPU dual-width cards are inserted into the corresponding slots of SLT#1PCIe x16, SLT#2PCIe x16, SLT#3PCIe x16, SLT#4PCIe x16, SLT#9PCIe x16, SLT#10PCIe x16, SLT#11PCIe x16, and SLT#12PCIe x16. Other types of single-width cards are inserted into SLT#5PCIe x8, SLT#6PCIe x8, SLT#7PCIe x16, and SLT#8PCIe x16.

9. The GPU direct-connect substrate according to any one of claims 1 to 7, characterized in that, The direct-connect substrate connects to nine GPU dual-width cards. GPU dual-width cards are inserted into SLT#1PCIe x16, SLT#2PCIe x16, SLT#3PCIe x16, SLT#4PCIe x16, SLT#8PCIe x16, SLT#9PCIe x16, SLT#10PCIe x16, SLT#11PCIe x16, and SLT#12PCIe x16, while other types of single-width cards are inserted into SLT#5PCIe x8, SLT#6PCIe x8, and SLT#7PCIe x16.

10. The GPU direct-connect substrate according to any one of claims 1 to 7, characterized in that, The direct-connect substrate connects 10 GPU dual-width cards. GPU dual-width cards are inserted into the corresponding slots of SLT#1PCIe x16, SLT#2PCIe x16, SLT#3PCIe x16, SLT#4PCIe x16, SLT#7PCIe x16, SLT#8PCIe x16, SLT#9PCIe x16, SLT#10PCIe x16, SLT#11PCIe x16, and SLT#12PCIe x16. Other types of single-width cards are inserted into SLT#5PCIe x8.

Images