Bridge, GPU interconnection structure and server device

Abstract

The utility model provides a bridge, a GPU interconnection structure and a server device. According to one embodiment of the invention, the bridge comprises two circuit boards which are arranged at an interval along a first direction; two ends of the connecting plate are respectively connected with the two circuit boards; wherein the connecting plate is made of a flexible material, a gap is formed in the connecting plate, the gap extends in the first direction and penetrates through the connecting plate in the thickness direction of the connecting plate, the minimum distance between the two circuit boards in the first direction is a first size, and the minimum distance between the two circuit boards in the second direction is a second size. The length of the connecting plate between the two circuit boards is a second size, and the first size is smaller than the second size. According to the scheme, the bridge and the GPU board card can be easily assembled.

Description

Technical Field

[0001] This application relates to the field of computer hardware technology, and more specifically, to a bridge, a GPU interconnect structure, and a server device. Background Technology

[0002] With the continuous development of computer technology, the computing demands of artificial intelligence and high-performance computing are constantly growing. Therefore, there is an increasing need for multiprocessor systems that can support seamless interconnection between processors so that they can work together as a giant accelerator.

[0003] A multiprocessor system integrates multiple GPUs (graphics processing units). These GPUs can share computing and / or storage resources through interconnection. Computational tasks can be distributed to each GPU in parallel and performed simultaneously, thereby improving computing efficiency.

[0004] Bridges connect multiple GPUs, enabling them to work together and improve graphics performance. However, current bridge interfaces are typically in fixed positions, each GPU's connector has design tolerances, and assembly tolerances may also exist between GPUs. This can lead to situations where the bridge interfaces and connectors cannot align precisely when connecting multiple GPUs, making assembly between the bridge and the GPUs difficult. Furthermore, even if the bridge is forcibly connected to the GPUs, the connection reliability is low due to these assembly tolerances. Summary of the Invention

[0005] This application provides a bridge, a GPU interconnect structure, and a server device, which can reduce the assembly difficulty between the bridge and the GPU board and improve the assembly reliability.

[0006] In a first aspect, this application provides a bridge, comprising:

[0007] Two circuit boards spaced apart along a first direction;

[0008] A connecting plate is provided, with two circuit boards connected to each end; wherein the connecting plate is made of a flexible material, and a gap is provided on the connecting plate, the gap extending along the first direction and penetrating the connecting plate along the thickness direction.

[0009] Optionally, the minimum distance between the two circuit boards along the first direction is the first dimension;

[0010] The length of the portion of the connecting plate between the two circuit boards is the second dimension;

[0011] The first dimension is smaller than the second dimension, and the difference between the second dimension and the first dimension is greater than or equal to 5 mm and less than or equal to 20 mm.

[0012] Optionally, the connecting plate includes:

[0013] A connecting portion extending along a first direction, wherein there are two connecting portions, and the two connecting portions are spaced apart along a second direction to form the gap; the second direction is perpendicular to the first direction and the thickness direction; each connecting portion is connected to two circuit boards spaced apart along the first direction at both ends;

[0014] The reinforcing part is fixed to the circuit board, and the two ends of the reinforcing part are respectively connected to the two connecting parts.

[0015] Optionally, the number of reinforcing parts is two, and the two reinforcing parts are respectively fixed to the two circuit boards.

[0016] Optionally, the dimension of the reinforcing part in the first direction is greater than or equal to 5 mm and less than or equal to 10 mm.

[0017] Optionally, the bridge may further include a lifting portion, which is fixed to the circuit board.

[0018] Optionally, the circuit board is provided with a fastening hole extending through its thickness direction, and the lifting part is provided with a through hole extending through its thickness direction;

[0019] The bridge has an assembled state and an unassembled state. In the assembled state, the fastening holes and the through holes are used to pass fasteners to fix the circuit board to the hardware to be interconnected, and the lifting part is attached to the circuit board. In the unassembled state, the lifting part and the circuit board have a gap along the thickness direction.

[0020] Secondly, this application provides a GPU interconnect structure, including a cross-region bridging structure and multiple GPU boards;

[0021] The cross-region bridging structure includes fasteners and any of the bridges described above; the bridge circuit board is provided with fastening holes for fixing, and the fasteners are used to cooperate with the fastening holes to fix the bridge to the GPU board.

[0022] At least two of the multiple GPU boards are spaced apart along the first direction, one of the two circuit boards is connected to one of the at least two GPU boards spaced apart along the first direction, and the other of the two circuit boards is connected to the other of the at least two GPU boards spaced apart along the first direction.

[0023] Optionally, there are multiple cross-region bridging structures, including a first cross-region bridging structure and a second cross-region bridging structure. The connecting plates of the bridges in the first cross-region bridging structure and the second cross-region bridging structure are located on two sides of the circuit board that are opposite to each other along a second direction, and the second direction is perpendicular to the first direction and the thickness direction.

[0024] In this configuration, one of the circuit boards of the second cross-zone bridging structure is located in the area between the two circuit boards of the first cross-zone bridging structure, and the other circuit board of the second cross-zone bridging structure is located away from the first cross-zone bridging structure.

[0025] Optionally, the circuit board in the second cross-zone bridging structure located between the two circuit boards of the first cross-zone bridging structure serves as the first circuit board;

[0026] The connecting plates of the first cross-region bridging structure and the second cross-region bridging structure are spaced apart along the second direction to form a gap; the fastening holes on the first circuit board are exposed to the gap along the thickness direction.

[0027] Optionally, the spacing in the second direction is greater than or equal to 10 mm and less than or equal to 20 mm.

[0028] Optionally, the number of the cross-regional bridging structures can be multiple;

[0029] The bridge's connecting plate includes a connecting portion extending along a first direction, and two circuit boards spaced apart along the first direction are respectively connected to the two ends of the connecting portion.

[0030] The width of the connecting portion of at least two cross-regional bridging structures with different gap widths is the same.

[0031] Optionally, the GPU board includes a fixed position for engaging with the fastener;

[0032] At least one of the gaps in the cross-regional bridging structure is located above the fixed position, which is exposed to the gap along the thickness direction.

[0033] Optionally, there are multiple cross-region bridging structures, with at least one cross-region bridging structure located above at least another cross-region bridging structure, and the gap in the upper cross-region bridging structure exposes the fastening holes in the lower cross-region bridging structure along the thickness direction; and / or,

[0034] The GPU interconnect structure also includes a neighboring cell bridging structure, which includes a neighboring cell bridge and fasteners. The circuit board of the neighboring cell bridge has fastening holes for fixing, and the fasteners are used to cooperate with the fastening holes to fix the neighboring cell bridge to the GPU board. The neighboring cell bridge does not include gaps.

[0035] The cross-zone bridging structure is located above the adjacent zone bridging structure, and the gap in the upper cross-zone bridging structure exposes the fastening holes in the lower adjacent zone bridging structure along the thickness direction.

[0036] Thirdly, this application also provides a server device including the GPU interconnect structure as described in any of the preceding claims.

[0037] The solution provided in this application includes at least the following beneficial effects:

[0038] The flexible connector plate can be bent and unfolded appropriately, thereby adjusting the spacing between the two circuit boards to overcome design or assembly tolerances between the two GPU board connectors connected to the bridge. This allows the circuit boards to connect smoothly to the GPU boards, reducing the assembly difficulty between the bridge and the GPU boards and improving the connection reliability. Furthermore, the connector plate has a gap extending along a first direction and penetrating the thickness of the connector plate. The gap enhances the flexibility of the connector plate, making it easier to bend or twist, thus better meeting the adjustment requirements of the spacing between the two circuit boards in different assembly scenarios and better adapting to various tolerance conditions between GPU board connectors. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the structure of a bridge as shown in one embodiment;

[0040] Figure 2 This is a schematic diagram of a GPU interconnect structure as shown in one embodiment;

[0041] Figure 3 This is a schematic diagram of the first cross-regional bridging structure shown in one embodiment;

[0042] Figure 4 This is a schematic diagram of the second cross-region bridging structure shown in one embodiment;

[0043] Figure 5This is a schematic diagram of the bridge structure shown in another embodiment.

[0044] Explanation of reference numerals in the attached figures:

[0045] 10. Circuit board; 11. Interface; 12. Fastening hole; 20. Connecting plate; 21. Connecting part; 22. Reinforcing part; 30. Gap; 40. Lifting part; 41. Through hole; 50. Fastener; 60. GPU board; 61. First GPU board; 62. Second GPU board; 63. Third GPU board; 64. Fourth GPU board; 101. First cross-region bridging structure; 102. Second cross-region bridging structure; 103. Third cross-region bridging structure; 104. Neighboring region bridging structure. Detailed Implementation

[0046] The technical solutions in the embodiments (or "implementations") of this application will be clearly and completely described herein with reference to the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements.

[0047] If the embodiments of this application contain terms relating to directional indications or positional relationships (such as up, down, left, right, front, back, inside, outside, top, bottom, center, vertical, horizontal, longitudinal, transverse, length, width, counterclockwise, clockwise, axial, radial, circumferential, etc.), such terms are only used to explain the relative positional relationships and movement of the components in a specific posture (as shown in the attached figures); if the specific posture changes, the directional indications or positional relationships will also change accordingly. Furthermore, the terms "first" and "second" used in the embodiments of this application are only for descriptive convenience and should not be construed as indicating or implying relative importance.

[0048] Please refer to Figures 1 to 5 This application provides a bridge, a GPU interconnect structure, and a server device that makes the assembly between the bridge and the GPU board easier.

[0049] Please refer to this first. Figure 1 In a first aspect, embodiments of this application provide a bridge that can be used for connections between multiple interconnected hardware, such as connections between storage hard drives, connections between multiple GPU boards, and connections between multiple PLCs (Programmable Logic Controllers). The bridge of this application will be described in detail below using a GPU board as an example.

[0050] The bridge includes a connection plate 20 and two circuit boards 10 spaced apart along a first direction.

[0051] The connecting plate 20 has two circuit boards 10 connected to its two ends respectively. Specifically, one end of the connecting plate 20 connects to one of the two circuit boards 10, and the other end connects to the other. The two circuit boards 10 are used to connect two different GPU cards respectively. The connecting plate 20 interconnects the two GPU cards connected to the two circuit boards 10. For example, each circuit board 10 has an interface 11, and the GPU card has a connector. The interface 11 is adapted to the connector, which can be inserted into the interface 11 of the circuit board 10 to connect the circuit board 10 to the GPU card. The connector can be a PCIe (Peripheral Component Interconnect Express) connector, an M.2 connector, an HDMI (High Definition Multimedia Interface) connector, etc., but is not limited to these.

[0052] The connecting plate 20 is made of a flexible material, which may include, but is not limited to, metal foil (copper foil, aluminum foil, etc.), conductive rubber, flexible printed circuit materials, etc. Furthermore, the minimum distance between the two circuit boards 10 along the first direction X is the first dimension (i.e., the spacing between the two circuit boards 10 in the first direction X), and the length of the connecting plate 20 between the two circuit boards 10 is the second dimension. The first dimension is smaller than the second dimension.

[0053] The flexible connecting plate 20 can be bent appropriately from a fairly flat state to make itself bulge or arch, or it can be unfolded from a fairly bent state to make itself relatively flat, thereby adjusting the spacing between the two circuit boards 10 to overcome the design tolerances or assembly tolerances between the two GPU board connectors connected to the bridge, so that the bridge can be smoothly connected to the GPU board. This reduces the assembly difficulty between the bridge and the GPU board and improves the connection reliability between the bridge and the GPU board.

[0054] Furthermore, the connecting plate 20 is provided with a slot 30, which extends along the first direction X and penetrates the connecting plate 20 along its thickness direction. The slot 30 enhances the flexibility of the connecting plate 20, making it easier to bend and unfold, thereby better meeting the adjustment requirements of the spacing between the two circuit boards 10 in different assembly scenarios, and thus better adapting to various tolerance conditions between GPU board connectors.

[0055] In one embodiment, the difference between the second dimension and the first dimension is greater than or equal to 5 mm and less than or equal to 20 mm.

[0056] With this configuration, the connecting plate 20 has sufficient margin to accommodate potential positional changes between the two circuit boards 10, allowing for flexible adjustment of the spacing between them to accommodate tolerances in the GPU board connector. Furthermore, by limiting this margin (i.e., the difference between the second and first dimensions) to within the range of 5mm to 20mm, the connecting plate 20 is neither too long and occupying excessive space, nor too short and limiting the adjustable range of the spacing between the two circuit boards 10. Therefore, this solution, through these limitations, ensures that the connecting plate 20 does not occupy excessive unnecessary space while allowing for flexible adjustment of the spacing between the two circuit boards 10.

[0057] For example, the difference between the second dimension and the first dimension can be 5mm, 6mm, 7mm, 8mm, 9mm, 10mm, 11mm, 12mm, 13mm, 14mm, 15mm, 16mm, 17mm, 18mm, 19mm, 20mm, etc., but is not limited to these.

[0058] In one embodiment, the connecting plate 20 includes a connecting portion 21 and a reinforcing portion 22.

[0059] The connecting portion 21 extends along the first direction X, and there are two connecting portions 21. The two connecting portions 21 are spaced apart along the second direction Y and form the aforementioned gap 30. The two ends of each connecting portion 21 are respectively connected to two circuit boards 10 spaced apart along the first direction to form a circuit. The second direction Y is perpendicular to the first direction X and the aforementioned thickness direction.

[0060] The reinforcing part 22 is fixed to the circuit board 10, and two connecting parts 21 are respectively connected to both ends of the reinforcing part 22. The reinforcing part 22 can provide additional support to the two connecting parts 21, enhancing the ability of the connecting parts 21 to resist external forces.

[0061] Furthermore, there can be two reinforcing parts 22, which are fixed to two circuit boards 10 respectively. In other words, one of the reinforcing parts 22 is fixed to one of the two circuit boards 10, and the other reinforcing part 22 is fixed to the other of the two circuit boards 10. Each reinforcing part 22 is connected to two connecting parts 21 at both ends.

[0062] In one embodiment, the dimension of the reinforcing portion 22 in the first direction X is greater than or equal to 5 mm and less than or equal to 10 mm. A dimension of 5 mm or greater in the first direction X ensures that the reinforcing portion 22 provides sufficient support area and structural strength. A dimension of 10 mm or less in the first direction X avoids material waste and prevents the reinforcing portion 22 from excessively affecting the dimension of the gap 30 between the two connecting portions 21 in the first direction X, thereby avoiding excessive impact on the flexibility of the connecting plate 20.

[0063] In one embodiment, the bridge further includes a lifting portion 40, which is fixed to the circuit board 10. When installing or removing the bridge from the GPU board, the lifting portion 40 provides a convenient point of leverage. Installers can grasp the lifting portion 40 to lift the circuit board 40, or use a robotic arm or similar device to grab the lifting portion 40, thereby making it easier to place the bridge in the intended position and ensure accurate alignment between the two circuit boards 10 and the corresponding GPU board. It also makes it easier to remove the bridge from the GPU board.

[0064] Furthermore, the lifting portion 40 is fixed to both ends of the circuit board 10. Specifically, the circuit board 10 has two lifting portion mounting holes extending through its thickness direction, located at both ends of the circuit board 10 along the second direction Y. The lifting portion 40 includes two lifting portion ends and a lifting portion body. The lifting portion body is located on the front side of the circuit board 10. One of the two lifting portion ends passes through one of the two lifting portion mounting holes from the back side of the circuit board 10 and is fixed to the lifting portion body. The other lifting portion end passes through the other of the two lifting portion mounting holes from the back side of the circuit board 10 and is fixed to the lifting portion body. The lifting portion ends and the lifting portion body can be fixed by adhesive bonding, but are not limited to this method.

[0065] Please refer to Figure 1 and combined Figure 4 In one embodiment, the circuit board 10 is provided with a fastening hole 12 extending through its thickness direction, and the lifting part 40 is provided with a through hole 41 extending through its thickness direction.

[0066] The bridge has an assembled state and an unassembled state. In the assembled state, the fastening hole 12 and the through hole 41 are used to cooperate with the fastener 50, which can fix the circuit board 10 to the GPU board and other interconnect hardware, and press the lifting part 40 onto the circuit board 10. In the unassembled state, the lifting part 40 and the circuit board 10 have a gap formed in the aforementioned thickness direction.

[0067] With this configuration, in the assembled state, the lifting part 40 can be placed on the circuit board 10, reducing the space occupied and improving the aesthetics. In the unassembled state, there is a gap between the lifting part 40 and the circuit board 10, which allows installers to hold the lifting part 40 through the gap, making it easier to assemble or disassemble the circuit board 10 and the GPU card.

[0068] For example, when the circuit board 10 needs to be removed from the GPU board, after the installer removes the fastener 50, the fastener 50 has no tightening effect on the lifting part 40. Under the action of the restoring force, the lifting part 40 can automatically move away from the circuit board 10, forming the gap, so that the installer can remove the circuit board 10 from the GPU board through the lifting part 40.

[0069] In some embodiments, the number of lifting parts 40 can be two, with each circuit board 10 having one lifting part 40 fixed. The fixing method can be referred to the above embodiments, and will not be repeated in this embodiment.

[0070] Secondly, please refer to Figure 2 and combined Figure 1 This application also provides a GPU interconnect structure, including a cross-region bridging structure and multiple GPU boards 60.

[0071] The cross-region bridging structure includes a fastener 50 and a bridge as described in any of the above embodiments or implementations. The bridge's circuit board 10 has fastening holes 12 for fixing, and the fastener 50 is used to mate with the fastening holes 12 to fix the bridge to the GPU board 60.

[0072] At least two of the multiple GPU boards 60 are spaced apart along a first direction X. One of the two circuit boards 10 is connected to one of the at least two GPU boards 60 spaced apart along the first direction X, and the other of the two circuit boards 10 is connected to the other of the at least two GPU boards 60 spaced apart along the first direction X.

[0073] The cross-region bridging structure includes a connecting plate 20 made of flexible material. The connecting plate 20 can be bent or unfolded appropriately along the first direction X as needed to adjust the spacing between the two circuit boards 10 of the cross-region bridging structure, thereby overcoming the design tolerances or assembly tolerances between the connectors of the two GPU boards 60 connected to the two circuit boards 10. In this way, the circuit boards 10 can be smoothly connected to the GPU boards 60, so that the cross-region bridging structure can be smoothly connected to the corresponding two GPU boards 60, thereby interconnecting the two GPU boards.

[0074] Please refer to Figure 3 and Figure 4 and combined Figure 2In one embodiment, there are multiple cross-region bridging structures, including a first cross-region bridging structure 101 and a second cross-region bridging structure 102. The connecting plates 20 of the bridges in the first and second cross-region bridging structures 101 and 102 are located on opposite sides of the circuit board 10 along a second direction Y, where the second direction Y is perpendicular to the first direction X and the thickness direction. In other words, the connecting plate 20 of the bridge in the first cross-region bridging structure 101 is located on one side of the circuit board 10 along the second direction Y, and the connecting plate 20 of the bridge in the second cross-region bridging structure 102 is located on the other side of the circuit board 10 along the second direction Y. That is, the connecting plates 20 of the bridges in the second and first cross-region bridging structures 101 are positioned away from each other.

[0075] In this structure, one of the circuit boards 10 of the first cross-region bridging structure 101 is located in the area between the two circuit boards 10 of the first cross-region bridging structure 101, and the other circuit board 10 of the second cross-region bridging structure 102 is located away from the first cross-region bridging structure.

[0076] Thus, the connecting plates 20 of the bridge in the first cross-region bridging structure 101 and the second cross-region bridging structure 102 are positioned away from each other in the second direction Y, which frees up the space between the two circuit boards 10 of the first and second cross-region bridging structures 101 and 102. One of the circuit boards 10 of the second cross-region bridging structure 102 is located in the area between the two circuit boards 10 of the first cross-region bridging structure 101, utilizing the space freed up by the connecting plates 20 of the first and second bridge structures. Therefore, the solution provided in this embodiment can improve the space utilization of the entire GPU interconnect structure, helping to achieve miniaturization of the GPU interconnect structure or integrate more functional components in a limited space.

[0077] In one embodiment, the circuit board 10 in the second cross-region bridging structure 102 located in the region between the two circuit boards 10 of the first cross-region bridging structure 101 serves as the first circuit board 1021; the connecting plates 20 of the first cross-region bridging structure 101 and the second cross-region bridging structure 102 are spaced apart along the second direction Y to form a gap; the fastening holes 12 on the first circuit board 1021 are exposed to the gap along the thickness direction.

[0078] With this configuration, the first circuit board 1021 in the second cross-region bridging structure 102 utilizes the space between the two circuit boards 10 of the first cross-region bridging structure 101. The fastening holes 12 on the first circuit board 1021 are exposed through the aforementioned spacing, allowing operators to directly see and manipulate them. For example, fasteners 50 can be passed through the fastening holes 12 to fix the first circuit board 1021 to the corresponding GPU board 60, or the fasteners 50 can be removed from the fastening holes 12 to detach the first circuit board 1021 from the corresponding GPU board 60. Therefore, this embodiment, through the above design, reduces the difficulty and complexity of installing the first circuit board 1021 while utilizing the aforementioned space, thus improving the efficiency of installing and removing the first circuit board 1021.

[0079] In a further embodiment, the aforementioned spacing in the second direction Y is greater than or equal to 10 mm and less than or equal to 20 mm. This spacing of greater than or equal to 10 mm in the second direction Y ensures sufficient space for installation, maintenance, and disassembly operations. When a technician needs to operate through this spacing on the exposed fastening holes 12 on the first circuit board 1021, for example, by tightening or loosening fasteners 50 with tools, the minimum spacing of 10 mm ensures that tools can be easily inserted and operated, avoiding difficulties caused by insufficient space. The limitation of less than or equal to 20 mm in the second direction Y is to prevent wasted space.

[0080] For example, the spacing can be 10mm, 11mm, 12mm, 13mm, 14mm, 15mm, 16mm, 17mm, 18mm, 19mm, or 20mm.

[0081] Please continue to refer to this. Figure 2 and combined Figure 1 and Figure 3 In one embodiment, there are multiple cross-region bridging structures, and the connecting portions 21 of at least two cross-region bridging structures with different gap 30 widths have the same width. In other words, the gap 30 widths of at least two cross-region bridging structures are different, but the connecting portions 21 of these at least two cross-region bridging structures have the same width. That is, the gap 30 of one of the at least two cross-region bridging structures has a different width than the gap 30 of the other, and the connecting portion 21 of one of the at least two cross-region bridging structures has the same width as the connecting portion 21 of the other. For example, as... Figure 2 As shown, and in combination Figure 1 and Figure 3At least two cross-region bridging structures include a first cross-region bridging structure 101 and a third cross-region bridging structure 103. The width of the gap 30 in the first cross-region bridging structure 101 is smaller than the width of the gap 30 in the third cross-region bridging structure 103, but the width of the connecting portion 21 of the first cross-region bridging structure 101 is the same as the width of the connecting portion 21 of the third cross-region bridging structure 103.

[0082] At least two cross-region bridging structures have different gap widths 30. The cross-region connection structure with a wider gap 30 has greater flexibility in its connecting portion 21, which can meet larger tolerance requirements. The cross-region connection structure with a narrower gap 30 is suitable for situations with smaller tolerances. At least two cross-region bridging structures have the same width of connecting portion 21. This ensures that the electrical characteristics (e.g., bandwidth, transmission rate) of the two cross-region bridging structures remain consistent during signal transmission, guaranteeing accurate and consistent data signal transmission between at least two GPU boards 60. Therefore, the solution provided in this embodiment allows at least two cross-region bridging structures to maintain signal stability and consistency even when the gaps 30 are different.

[0083] like Figure 2 In the illustrated embodiment, at least two cross-region bridging structures include a first cross-region bridging structure 101 and a third cross-region bridging structure 103. The first cross-region bridging structure 101 connects the first GPU board 61 and the third GPU board 63 of the plurality of GPU boards 60, and the third cross-region bridging structure 103 connects the first GPU board 61 and the fourth GPU board 64 of the plurality of GPU boards 60. Since the third cross-region bridging structure 103 sequentially crosses the second GPU board 62 and the third GPU board 63 along the first direction X, while the second cross-region bridging structure 102 only crosses the second GPU board 62 along the first direction X, the tolerance that the third cross-region bridging structure 103 needs to overcome is generally larger. Therefore, the gap width of the first cross-region bridging structure 101 is smaller than the gap width of the third cross-region bridging structure 103. Furthermore, the width of the connecting portion 21 of the first cross-zone bridging structure 101 is equal to the width of the connecting portion 21 of the third cross-zone bridging structure 103, so as to maintain the consistency of the transmission signals of the first cross-zone bridging structure 101 and the third cross-zone bridging structure 103.

[0084] In one embodiment, the GPU board 60 includes a mounting position for engaging with a fastener 50. At least one gap 30 in the cross-sectional bridging structure is located above the mounting position, with the mounting position exposed along its thickness direction through the gap 30. During installation, the operator can more clearly observe the mounting position through the relatively open space of the gap 30, facilitating accurate engagement of the fastener 50 with the mounting position and simplifying installation and removal.

[0085] It should be noted that the fixing position may include, but is not limited to, the fastening hole 12. For example, the fixing position may also include the space around the fastening hole 12 so that fastening tools such as screwdrivers and wrenches can also be exposed through the gap 30, thereby facilitating the operator to use the fastening tools to fix the fastener 50.

[0086] In one embodiment, there are multiple cross-zone bridging structures, with at least one cross-zone bridging structure located above at least one cross-zone bridging structure, and the gap 30 in the upper cross-zone bridging structure exposes the fastening hole 12 in the lower cross-zone bridging structure along the thickness direction.

[0087] In this embodiment, the multiple cross-region bridging structures are arranged with at least one cross-region bridging structure above another, forming a layered spatial structure. This layered layout makes full use of vertical space, allowing more cross-region bridging structures and connected GPU boards and other components to be accommodated within a limited volume. Furthermore, the gap 30 in the upper cross-region bridging structure exposes the fastening holes 12 in the lower cross-region bridging structure along its thickness, facilitating installation, maintenance, and troubleshooting. During installation, operators can clearly see the fastening holes 12 of the lower cross-region bridging structure through the gap 30 in the upper structure, enabling more accurate insertion of fasteners 50 (such as screws and bolts) through the holes 12, reducing installation difficulties caused by obstructed vision or limited operating space.

[0088] In one embodiment, please refer to Figure 5 and combined Figure 1 and Figure 2 The GPU interconnect structure also includes a neighboring cell bridging structure 104, which includes fasteners and a neighboring cell bridge. The neighboring cell bridge does not include the aforementioned gap 30; otherwise, the neighboring cell bridge can have the same structure as the bridge in the above embodiment. The circuit board 10 of the neighboring cell bridge also has fastening holes 12 for fixing, and the fasteners are used to engage with the fastening holes 12 to fix the neighboring cell bridge to the GPU board 60. The cross-cell bridging structure is located above the neighboring cell bridging structure 104, and the gap 30 in the upper cross-cell bridging structure exposes the fastening holes 12 in the lower cross-cell bridging structure along the thickness direction.

[0089] In this embodiment, the cross-zone bridging structure located above the adjacent zone bridging structure 104 can form a layered spatial structure to fully utilize the vertical space. Furthermore, the gap 30 in the cross-zone bridging structure exposes the fixing holes of the lower adjacent zone bridging structure 104 along its thickness, facilitating installation, maintenance, and troubleshooting. During installation, operators can clearly see the fastening holes 12 of the lower adjacent zone bridging structure 104 through the gap 30 in the upper cross-zone bridging structure, allowing for more accurate fastening of the fasteners 50 through the fastening holes 12 and reducing installation difficulties caused by obstructed vision or limited operating space.

[0090] It should be noted that, in this embodiment, a cross-region bridging structure refers to a situation where, among multiple GPU boards 60, there is at least one other GPU board 60 between the two GPU boards 60 connected by the cross-region bridging structure. A neighboring region bridging structure 104 refers to a situation where, among multiple GPU boards 60, there are no other GPU boards 60 between the two GPU boards 60 connected by the neighboring region bridging structure 104; that is, the neighboring region bridging structure 104 connects two adjacent GPU boards 60.

[0091] like Figure 2 In the illustrated embodiment, combined with Figure 1 , Figures 3 to 5 The multiple GPU boards 60 include a first GPU board 61, a second GPU board 62, a third GPU board 63, and a fourth GPU board 64, which are arranged sequentially along a first direction; the number of cross-region bridging structures is 3, and the number of adjacent region bridging structures 104 is 3.

[0092] The three cross-regional bridging structures are the first cross-regional bridging structure 101, the second cross-regional bridging structure 102, and the third cross-regional bridging structure 103.

[0093] The two circuit boards 10 of the first neighboring cell bridging structure 104 are respectively connected to the first GPU board 61 and the second GPU board 62.

[0094] The two circuit boards 10 of the first cross-region bridging structure 101 are respectively connected to the first GPU board 61 and the third GPU board 63.

[0095] The two circuit boards 10 of the second cross-region bridging structure 102 are respectively connected to the second GPU board 62 and the fourth GPU board 64.

[0096] The two circuit boards 10 of the second neighboring cell bridging structure 104 are respectively connected to the second GPU board 62 and the third GPU board 63.

[0097] The two circuit boards 10 of the third cross-zone bridging structure 103 are respectively connected to the first GPU board 61 and the fourth GPU board 64.

[0098] The two circuit boards 10 of the third neighboring cell bridging structure 104 are respectively connected to the third GPU board 63 and the fourth GPU board 64.

[0099] The second neighboring zone bridging structure 104 is located below the second spanning zone bridging structure 102. The gap 30 of the second spanning zone bridging structure 102 exposes the fastening holes of the second neighboring zone bridging structure 104 along the thickness direction.

[0100] Of course, in some other embodiments, the number of cross-zone bridging structures is not limited to three, and the number of neighboring zone bridging structures 104 is not limited to three.

[0101] Thirdly, this application also provides a server device, which includes the GPU interconnect structure described in any of the above embodiments or implementations.

[0102] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A bridge, characterized in that, include: Two circuit boards spaced apart along a first direction; A connecting plate is provided, with two circuit boards connected to each end; wherein the connecting plate is made of a flexible material, and a gap is provided on the connecting plate, the gap extending along the first direction and penetrating the connecting plate along the thickness direction; The minimum distance between the two circuit boards along a first direction is a first dimension, and the length of the connecting plate between the two circuit boards is a second dimension, wherein the first dimension is smaller than the second dimension.

2. The bridge according to claim 1, characterized in that, The difference between the second dimension and the first dimension is greater than or equal to 5 mm and less than or equal to 20 mm.

3. The bridge according to claim 1, characterized in that, The connecting plate includes: A connecting portion extending along a first direction, wherein there are two connecting portions, and the two connecting portions are spaced apart along a second direction to form the gap; the second direction is perpendicular to the first direction and the thickness direction; each connecting portion is connected to two circuit boards spaced apart along the first direction at both ends; The reinforcing part is fixed to the circuit board, and the two ends of the reinforcing part are respectively connected to the two connecting parts.

4. The bridge according to claim 3, characterized in that, The number of the reinforcing parts is two, and the two reinforcing parts are respectively fixed to the two circuit boards.

5. The bridge according to claim 3, characterized in that, The dimension of the reinforcing part in the first direction is greater than or equal to 5 mm and less than or equal to 10 mm.

6. The bridge according to claim 1, characterized in that, The bridge also includes a lifting section, which is fixed to the circuit board.

7. The bridge according to claim 6, characterized in that, The circuit board is provided with a fastening hole that extends through its thickness direction, and the lifting part is provided with a through hole that extends through its thickness direction. The bridge has an assembled state and an unassembled state. In the assembled state, the fastening holes and the through holes are used to pass fasteners to fix the circuit board to the hardware to be interconnected, and the lifting part is attached to the circuit board. In the unassembled state, the lifting part and the circuit board have a gap along the thickness direction.

8. A GPU interconnect structure, characterized in that, This includes cross-region bridging structures and multiple GPU boards; The cross-region bridging structure includes fasteners and a bridge as described in any one of claims 1 to 6; the bridge's circuit board is provided with fastening holes for fixing, and the fasteners are used to cooperate with the fastening holes to fix the bridge to the GPU board. At least two of the multiple GPU boards are spaced apart along the first direction, one of the two circuit boards is connected to one of the at least two GPU boards spaced apart along the first direction, and the other of the two circuit boards is connected to the other of the at least two GPU boards spaced apart along the first direction.

9. The GPU interconnect structure according to claim 8, characterized in that, The number of cross-regional bridging structures is multiple, including a first cross-regional bridging structure and a second cross-regional bridging structure. The connecting plates of the bridges in the first cross-regional bridging structure and the second cross-regional bridging structure are located on two opposite sides of the circuit board along a second direction, which is perpendicular to the first direction and the thickness direction. In this configuration, one of the circuit boards of the second cross-zone bridging structure is located in the area between the two circuit boards of the first cross-zone bridging structure, and the other circuit board of the second cross-zone bridging structure is located away from the first cross-zone bridging structure.

10. The GPU interconnect structure as described in claim 9, characterized in that, The circuit board in the second cross-zone bridging structure located between the two circuit boards in the first cross-zone bridging structure serves as the first circuit board; The connecting plates of the first cross-region bridging structure and the second cross-region bridging structure are spaced apart along the second direction to form a gap; the fastening holes on the first circuit board are exposed to the gap along the thickness direction.

11. The GPU interconnect structure as described in claim 10, characterized in that, The spacing in the second direction is greater than or equal to 10 mm and less than or equal to 20 mm.

12. The GPU interconnect structure as described in claim 8, characterized in that, The number of cross-regional bridging structures is multiple; The bridge's connecting plate includes a connecting portion extending along a first direction, and two circuit boards spaced apart along the first direction are respectively connected to the two ends of the connecting portion. The width of the connecting portion of at least two cross-regional bridging structures with different gap widths is the same.

13. The GPU interconnect structure as described in claim 8, characterized in that, The GPU board includes a fixed position for engaging with the fastener; At least one of the gaps in the cross-regional bridging structure is located above the fixed position, which is exposed to the gap along the thickness direction.

14. The GPU interconnect structure as described in claim 13, characterized in that, The number of the cross-zone bridging structures is plurality of, at least one of the cross-zone bridging structures is located above at least another cross-zone bridging structure, and the gap in the upper cross-zone bridging structure exposes the fastening hole in the lower cross-zone bridging structure along the thickness direction; and / or, The GPU interconnect structure also includes a neighboring cell bridging structure, which includes a neighboring cell bridge and fasteners. The circuit board of the neighboring cell bridge has fastening holes for fixing, and the fasteners are used to cooperate with the fastening holes to fix the neighboring cell bridge to the GPU board. The neighboring cell bridge does not include gaps. The cross-zone bridging structure is located above the adjacent zone bridging structure, and the gap in the upper cross-zone bridging structure exposes the fastening holes in the lower adjacent zone bridging structure along the thickness direction.

15. A server device, characterized in that, Includes the GPU interconnect structure as described in any one of claims 8 to 13.

Images