GPU card module and GPU server

Abstract

The embodiment of the present application discloses a GPU card module and a GPU server. After the GPU card module is plugged and installed with an adapter circuit board to form a GPU module, when a pair of GPU modules are opposite in front and back along the height direction of the server chassis and a pair of adapter circuit boards are placed relative to each other, the fixing bracket of the GPU card module is configured so as not to cause obstacles when the pair of GPU modules are plugged in side by side and staggered using the installation space between the packaging shell and the adapter circuit board. Therefore, when the GPU circuit board in the GPU card module is a full-height double-width GPU card, four GPU modules can be installed side by side and staggered along the width direction of the server chassis, thereby improving the space utilization rate in the width direction of the server chassis.

Description

Technical Field

[0001] The present invention relates to the technical field of servers, and in particular to a GPU card module and a GPU server. Background Art

[0002] The GPU server is a fast, stable, and flexible computing server based on GPU that is applied to various scenarios such as video encoding and decoding, deep learning, and scientific computing.

[0003] In a GPU server, since the server chassis is installed in a general cabinet, its width dimension is usually fixed and strictly limited to 446 mm; and the width of the front baffle of a current full-height double-width GPU card module is 126.3 mm, so that at most three full-height double-width GPU card modules can be installed side by side in the width direction inside the server chassis, and after installing three full-height double-width GPU card modules side by side, there is still insufficient remaining space in the width direction of the server chassis to install one full-height double-width GPU card module.

[0004] As a result, the space utilization in the width direction of the server chassis is not high. Summary of the invention

[0005] In response to at least one aspect of the above technical problems, an embodiment of the present application provides a GPU card module and a GPU server. After the GPU card module is plugged and installed with an adapter circuit board to form a GPU module, when a pair of GPU modules are arranged with their front and back sides opposite to each other in the height direction of the server chassis and a pair of adapter circuit boards are placed relative to each other, the fixing bracket of the GPU card module is configured to not cause any obstruction to the pair of GPU modules when they are plugged in side by side and staggered using the installation space between the packaging shell and the adapter circuit board. Therefore, when the GPU circuit board in the GPU card module is a full-height double-width GPU card, four GPU modules can be installed side by side and staggered along the width direction of the server chassis, thereby improving the space utilization in the width direction of the server chassis.

[0006] In a first aspect, an embodiment of the present application provides a GPU card module, the GPU card module comprising:

[0007] A GPU circuit board encapsulated by an encapsulation shell, wherein a PCI-E interface of the GPU circuit board is exposed from the encapsulation shell, and the PCI-E interface is used to be plugged into an adapter circuit board, so that the GPU circuit board and the adapter circuit board are plugged and installed to form a GPU module;

[0008] A fixing bracket fixedly mounted on the GPU circuit board, the fixing bracket being used to fix the GPU circuit board with the adapter circuit board after the GPU circuit board and the adapter circuit board are plugged and installed, and to install the GPU module formed after the GPU circuit board and the adapter circuit board are plugged and installed on a server chassis;

[0009] Among them, when a pair of the GPU modules are arranged with their front and back sides opposite to each other in the height direction of the server chassis and a pair of the adapter circuit boards are placed relative to each other, the fixing bracket is configured to not cause any obstruction to the side-by-side and staggered plug-in of the pair of GPU modules using the installation space between the packaging shell and the adapter circuit board, so that when the GPU circuit board is a full-height double-width GPU card, the four GPU modules are installed side-by-side and staggered in the width direction of the server chassis.

[0010] In one embodiment, preferably, the fixing bracket includes a first bracket and a second bracket which are separately arranged;

[0011] Wherein, the first bracket is fixedly installed at the front end of the GPU circuit board, and the second bracket is fixedly installed at the rear end of the GPU circuit board;

[0012] Furthermore, after the adapter circuit board is plugged and installed with the GPU circuit board, the first bracket and the second bracket are respectively fixedly installed with two ends of the adapter circuit board, so that the fixing bracket is used to be fixedly installed with the adapter circuit board after the GPU circuit board and the adapter circuit board are plugged and installed;

[0013] The first bracket and the second bracket are also used to be installed on the server chassis respectively, so that the fixed bracket is used to install the GPU module formed by plugging and installing the GPU circuit board and the adapter circuit board on the server chassis.

[0014] In one embodiment, preferably, the first bracket includes a first baffle and a first mounting ear connected by vertical bending, wherein the first baffle is fixedly mounted on the front end of the GPU circuit board and abuts against the front end surface of the packaging shell, and the first mounting ear is arranged parallel to the adapter circuit board and is used for fixed installation with one end of the adapter circuit board;

[0015] The second bracket includes a second baffle and a second mounting ear connected by a vertical bend, wherein the second baffle is fixedly mounted on the rear end of the GPU circuit board and abuts against the rear end surface of the packaging shell, and the second mounting ear is arranged parallel to the adapter circuit board and is used for fixed installation with the other end of the adapter circuit board.

[0016] In one embodiment, preferably, in an area corresponding to the installation space between the packaging shell and the adapter circuit board, the first baffle and the second baffle are respectively provided with a plugging notch;

[0017] Wherein, relative to the height direction of the server chassis, the plug-in gap is located at one end of the first baffle and the second baffle away from the GPU circuit board, so that the plug-in gap is used for a pair of GPU modules to be plugged in side by side and staggered when the front and back sides are opposite to each other along the height direction of the server chassis and a pair of adapter circuit boards are placed opposite to each other.

[0018] In one embodiment, preferably, along the height direction of the server chassis, the height of the insertion notch is not less than half of the height of the GPU module.

[0019] In one embodiment, preferably, along the width direction of the server chassis, a pair of first positioning columns are provided on both sides of the outer surface of the first baffle, and the first positioning columns are used to be inserted into first waist holes provided along the height direction of the server chassis;

[0020] In addition, the second baffle is provided with a mounting ear, and the mounting ear is used to be fastened to the server chassis after the first positioning column is inserted into the first waist hole.

[0021] In one embodiment, preferably, in the area between the pair of the first positioning pillars, the outer surface of the first baffle is further provided with a second positioning pillar and a third positioning pillar;

[0022] Among them, the second positioning column and the third positioning column are arranged up and down along the height direction of the server chassis, and the second positioning column and the third positioning column are staggered relative to the height direction of the server chassis, so that when the GPU card module is installed on the server chassis with the front and back sides opposite to each other along the height direction of the server chassis, the second positioning column and the third positioning column are respectively inserted into the second waist hole of the server chassis.

[0023] In one embodiment, preferably, the mounting ear is vertically bent relative to the second baffle, and the mounting ear is used to be fastened to a horizontally arranged mounting beam of the server chassis.

[0024] In a second aspect, an embodiment of the present application provides a GPU server, the GPU server comprising a server chassis and a first GPU card module installed inside the server chassis;

[0025] Among them, along the width direction of the server chassis, four first GPU card modules are installed side by side inside the server chassis, and the first GPU card module is the above-mentioned GPU card module.

[0026] In one embodiment, preferably, two second GPU card modules are also installed inside the server chassis;

[0027] Among them, the GPU circuit board in the second GPU card module is a half-height single-width GPU card.

[0028] In one embodiment, preferably, the server chassis includes a front bracket and a mounting beam for mounting the first GPU card module;

[0029] The front-end bracket is provided with a pair of first waist holes at positions corresponding to both sides of each of the first GPU card modules, the first waist holes are arranged along the height direction of the server chassis, and the first waist holes are used for inserting and installing a pair of first positioning columns in the first GPU card module; the front-end bracket is provided with a second waist hole in the area between the pair of first waist holes, and the second waist hole is used for inserting and installing a second positioning column or a third positioning column in the first GPU card module;

[0030] The mounting crossbeam is used for fastening and mounting the mounting ears in the first GPU card module.

[0031] In one embodiment, preferably, the GPU server includes a server mainboard and a transfer circuit board for plugging and installing with the first GPU card module;

[0032] Wherein, a transfer cable is welded on the transfer circuit board, and the transfer cable is electrically connected to the server mainboard through a transfer socket.

[0033] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0034] An embodiment of the present application provides a GPU card module and a GPU server. The GPU card module includes a GPU circuit board encapsulated by an encapsulation shell and a fixing bracket fixedly installed on the GPU circuit board. The GPU circuit board is exposed from the PCI-E interface of the encapsulation shell and is plugged and installed with an adapter circuit board to form a GPU module. The fixing bracket is used to fix the GPU circuit board with the adapter circuit board after the GPU circuit board and the adapter circuit board are plugged and installed, and the formed GPU module is installed in the server chassis.

[0035] It can be understood that after the GPU circuit board and the adapter circuit board are plugged and installed, an installation space is formed between the adapter circuit board and the packaging shell that encapsulates the GPU circuit board. At this time, when a pair of GPU modules are opposite in front and back along the height direction of the server chassis and a pair of adapter circuit boards are placed relatively, the pair of GPU modules can be plugged in side by side and staggered by virtue of the overlap of an adjacent pair of installation spaces. At this time, the fixed bracket in the GPU card module of this embodiment is configured not to hinder the side-by-side and staggered plugging of the pair of GPU modules. Therefore, when the GPU circuit board is a full-height double-width GPU card, four GPU modules formed by the GPU card module of this embodiment can be installed side by side and staggered along the width direction of the server chassis. Compared with the method of using existing GPU card modules in which only three GPU modules can be installed side by side in the width direction of the server chassis, the space utilization of the server chassis is improved, especially the space utilization in the width direction of the server chassis. In addition, due to the increase in the number of GPU modules, the performance and computing power of the GPU server are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0037] Figure 1 The schematic diagram of the structure of a GPU module is formed by installing an existing full-height double-width GPU card module and an adapter circuit board.

[0038] Figure 2 for Figure 1 Schematic diagram of the structure of the front baffle of the GPU card module.

[0039] Figure 3 This is a structural diagram of an existing GPU module installed in a server chassis.

[0040] Figure 4 This is a schematic diagram of the structure of a GPU module formed after the GPU card module and the adapter circuit board described in the embodiment of the present application are installed.

[0041] Figure 5 for Figure 4 Schematic diagram of the structure from another perspective.

[0042] Figure 6 This is a structural schematic diagram of the GPU module described in an embodiment of the present application installed in a server chassis.

[0043] Figure 7This is a comparison diagram of the width of the front baffle of the existing GPU card module and the first bracket in the embodiment of the present application.

[0044] Figure 8 This is a comparison diagram of the overall width of the four front end baffles of the existing GPU card module and the four first brackets in the embodiment of the present application.

[0045] Fig. 9 This is a schematic diagram of the structure of the staggered insertion of the fixed brackets described in the embodiment of the present application, wherein the dotted arrow in the figure indicates the insertion direction.

[0046] Fig.10 for Fig. 9 Schematic diagram of the structure after the fixing bracket is inserted.

[0047] Fig.11 This is a structural schematic diagram of the front end of the GPU module described in the embodiment of the present application being installed in a server chassis.

[0048] Fig.12 for Fig.11 A partial enlarged view of .

[0049] Fig.13 This is a structural schematic diagram of the rear end of the GPU module described in the embodiment of the present application being installed in a server chassis.

[0050] Fig.14 for Fig.13 A partial enlarged view of .

[0051] Fig.15 This is a schematic diagram of the structure of the half-height single-width GPU card module described in the embodiment of the present application.

[0052] Fig.16 for Figure 6 Schematic diagram of the side structure.

[0053] Wherein, the reference numerals are:

[0054] 10-Packaging shell,

[0055] 21-PCI-E interface,

[0056] 30-fixed bracket, 31-first bracket, 32-second bracket,

[0057] 311-first baffle, 312-first mounting ear,

[0058] 3111-first positioning column, 3112-second positioning column, 3113-third positioning column,

[0059] 321- second baffle, 322- second mounting ear,

[0060] 3211-Mounting ears,

[0061] 40-adapter circuit board,

[0062] 50-installation space,

[0063] 60-plug notch,

[0064] 70-fixing screw,

[0065] 100-server chassis, 101-first waist hole, 102-second waist hole, 103-front bracket, 104-mounting beam,

[0066] 201-first GPU card module, 202-second GPU card module,

[0067] 300-GPU module, 301-front baffle, 302-fixed sheet metal,

[0068] X-width direction, Y-height direction, Z-depth direction. DETAILED DESCRIPTION

[0069] In order to better understand the above technical solution, the example embodiments of the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited to the exemplary embodiments described here.

[0070] First of all, it should be noted that the width direction X, height direction Y and depth direction Z mentioned in this application are relative to the horizontal insertion of the server chassis into the chassis rack.

[0071] The GPU card module is arranged inside the server chassis. The GPU card or GPU circuit board in the GPU card module is placed horizontally, that is, the width direction of the server chassis is the height of the GPU card, the depth direction of the server chassis is the length of the GPU card, and the height direction of the server chassis is the width of the GPU card.

[0072] A standard GPU card refers to a standard card that complies with the PCI-E specification. Currently, there are many common specifications for GPU cards. Based on the standard GPU card, the height is halved to become a half-height GPU card, the length is halved to become a half-length GPU card, and the width is doubled to become a double-width GPU card; that is to say, a card with a standard PCI-E socket is called a full-height, single-width, full-length GPU card. On this basis, a half-height, single-width GPU card is halved in height (the size along the width of the server chassis is halved), and a full-height, double-width GPU card is doubled in width (the size along the height of the server chassis is doubled).

[0073] In addition, the server chassis mentioned in this application is a standard 2U chassis. In the server and equipment room, U is the abbreviation of unit, which represents the unit of rack size or server external size, and 1U is equal to 4.445 cm.

[0074] Figure 1 It is a structural diagram of an existing full-height double-width GPU card module after installing an adapter circuit board 40 to form a GPU module 300. Figure 2 for Figure 1 A schematic diagram of the structure of the front baffle 301 of the GPU card module, Figure 3 for Figure 1 The schematic diagram of the structure of the GPU module installed in the server chassis; it can be seen that in the existing GPU card module, the GPU circuit board (not shown in the figure) is encapsulated with an encapsulation shell 10, and a front end baffle 301 is installed at the front end of the GPU circuit board. After the adapter circuit board 40 is plugged and installed with the GPU card module, the front end baffle 301 is not fixedly connected to the adapter circuit board 40, but is fixedly installed with the adapter circuit board 40 through an additional fixed sheet metal 302, and the adapter circuit board 40 is directly plugged into the server motherboard; wherein, Figure 2 is a schematic diagram of the structure of the front end baffle 301, combined with Figure 1 and Figure 2 It can be seen that the front baffle is relatively long and usually exceeds the adapter circuit board along the width direction. Combined with the feature that the adapter circuit board is directly plugged into the server motherboard, Figure 3 As shown, when the existing GPU card module is used, only three GPU modules can be installed in the server chassis along the width direction, and after the three GPU modules are installed, a certain amount of free space will be left in the width direction of the server chassis, resulting in low space utilization of the server chassis.

[0075] In view of the above situation, the present application discloses a GPU card module and a GPU server having the GPU card module. After the GPU card module of the present application is plugged and installed with an adapter circuit board to form a GPU module, a pair of GPU modules can be plugged and installed side by side in an staggered manner. Specifically, the present application places a pair of GPU modules with their front and back sides opposite to each other in the height direction of a server chassis, and a pair of adapter circuit boards are opposite to each other. By virtue of the overlap of an adjacent pair of installation spaces (the space between the packaging shell and the adapter circuit board), the pair of GPU modules can be plugged and installed side by side in an staggered manner. In this way, the overall width of four GPU modules when installed side by side can be reduced, so that they can be installed in the server chassis.

[0076] Figure 4 , Figure 5 This is a schematic diagram of the structure of the GPU module formed after the GPU card module and the adapter circuit board are installed. Figure 4 , Figure 5 A GPU card module, the GPU card module includes a GPU circuit board encapsulated by an encapsulation shell 10 and a fixing bracket 30 fixedly installed on the GPU circuit board; wherein a PCI-E interface 21 of the GPU circuit board is exposed from the encapsulation shell 10, and the PCI-E interface 21 is used to be plugged into an adapter circuit board 40, so that the GPU circuit board and the adapter circuit board 40 are plugged and installed to form a GPU module 300; the fixing bracket 30 is used to be fixedly installed with the adapter circuit board 40 after the GPU circuit board and the adapter circuit board 40 are plugged and installed, and the GPU circuit board and the adapter circuit board 40 are plugged and installed. The GPU module 300 formed after the installation is installed in the server chassis 100; wherein, when a pair of GPU modules 300 are opposite to each other in the height direction of the server chassis, and a pair of adapter circuit boards 40 are placed opposite to each other, the fixing bracket 30 is configured to not cause any obstruction to the pair of GPU modules 300 when they are plugged in side by side and staggered using the installation space 50 between the packaging shell 10 and the adapter circuit board 40, so that when the GPU circuit board is a full-height double-width GPU card, the four GPU modules 300 are installed side by side and staggered along the width direction of the server chassis 100.

[0077] It should be noted that the GPU card module mentioned in each embodiment of the present application is a GPU circuit board encapsulated by a packaging shell, and the GPU circuit board is installed with a fixed bracket; wherein, the GPU card module forms a GPU module after the adapter circuit board is plugged in and installed, and then the GPU module is installed in the server chassis.

[0078] Among them, the server chassis mentioned in this application is, for example, a 2U server chassis, and the GPU module is horizontally installed in the server chassis; thus, the GPU circuit board of the GPU card module is placed horizontally in the server chassis, and the adapter circuit board is placed vertically.

[0079] Therefore, relative to the GPU card module, the direction perpendicular to the GPU circuit board is the height direction Y of the server chassis, the direction of one side of the PCI-E interface of the GPU circuit board is the depth direction Z of the server chassis, and the direction of the other side of the GPU circuit board is the width direction X of the server chassis.

[0080] The GPU card module includes a GPU circuit board, a packaging shell is provided on the GPU circuit board to package the components on the GPU circuit board, and a fixing bracket is also installed on the GPU circuit board, and the fixing bracket is located outside the packaging shell.

[0081] It can be understood that the PCI-E interface of the GPU circuit board is not encapsulated by the packaging shell, but is exposed from the packaging shell. The PCI-E interface is used to be plugged and installed with the adapter circuit board to form a GPU module. Then, the adapter circuit board is electrically connected to the server motherboard to complete the electrical connection between the GPU circuit board and the server motherboard.

[0082] See also Figure 4 and Figure 5 After being plugged and installed with the adapter circuit board, an installation space is formed between the packaging shell and the adapter circuit board. That is, the installation space mentioned in this embodiment can be understood as the space between the adapter circuit board on one side of the packaging shell and the side wall of the packaging shell.

[0083] Then, it can be understood that when a pair of GPU modules are placed with their front and back sides opposite to each other along the height direction of the server chassis and a pair of adapter circuit boards are placed opposite to each other, the pair of GPU modules can overlap a pair of adjacent installation spaces to achieve a pair of GPU modules being staggered and plugged in side by side. Fig. 9 The fixing bracket of the present embodiment is arranged not to hinder the side-by-side staggered plug-in, so that, on the basis of reducing the width by the side-by-side staggered plug-in, when the GPU circuit board is a full-height double-width GPU card, four GPU modules can be installed side by side in the width direction of the server chassis using the GPU card module of the present embodiment.

[0084] It can be seen that compared with the existing GPU card modules, the GPU card module of this embodiment can add a full-height double-width GPU card in the server chassis, which maximizes the space utilization of the server chassis (especially in the width direction) as a whole, thereby improving the performance and computing power of the GPU server.

[0085] It is understandable that the present embodiment does not limit the specific structure and shape of the fixing bracket, as long as it does not cause any obstruction to the above-mentioned side-by-side staggered plug-in.

[0086] In actual use, please refer to Figure 7 and Figure 8 The width dimension of the front baffle in the existing full-height double-width GPU card module is actually 126.3 mm. When four front baffles are placed side by side, the overall width dimension is 505.2 mm, which is larger than the width of the server chassis (446 mm). However, the width dimension of the fixed bracket in this embodiment is actually 115 mm. When four fixed brackets are placed side by side, the overall width dimension is 430.8 mm, which is smaller than the width of the server chassis (446 mm).

[0087] In addition, it should be noted that, in the present embodiment, the function of the fixing bracket is, on the one hand, to fix the plugged-in adapter circuit board and the GPU card module together, and on the other hand, the fixing bracket is also used to install the GPU module formed after plugging in the server chassis without the need to use additional fixing sheet metal for fixing, thereby reducing the weight of the GPU module.

[0088] An embodiment of the present application provides a GPU card module and a GPU server. The GPU card module includes a GPU circuit board encapsulated by an encapsulation shell and a fixing bracket fixedly installed on the GPU circuit board. The GPU circuit board is exposed from the PCI-E interface of the encapsulation shell and is plugged and installed with an adapter circuit board to form a GPU module. The fixing bracket is used to fix the GPU circuit board with the adapter circuit board after the GPU circuit board and the adapter circuit board are plugged and installed, and the formed GPU module is installed in the server chassis.

[0089] Among them, it can be understood that after the GPU circuit board and the adapter circuit board are plugged and installed, an installation space is formed between the adapter circuit board and the packaging shell that encapsulates the GPU circuit board. At this time, when a pair of GPU modules are opposite in front and back along the height direction of the server chassis and a pair of adapter circuit boards are placed relatively, the pair of GPU modules can be plugged in side by side and staggered by virtue of the overlap of an adjacent pair of installation spaces. At this time, the fixed bracket in the GPU card module of this embodiment is configured not to hinder the side-by-side and staggered plugging of the pair of GPU modules. Therefore, when the GPU circuit board is a full-height double-width GPU card, four GPU modules formed by the GPU card module of this embodiment can be installed side by side and staggered along the width direction of the server chassis. Compared with the method of using existing GPU card modules in which only three GPU modules can be installed side by side in the width direction of the server chassis, the space utilization of the server chassis is improved, especially the space utilization in the width direction of the server chassis. In addition, due to the increase in the number of GPU modules, the performance and computing power of the GPU server are improved.

[0090] In one possible implementation, the fixed bracket 30 includes a first bracket 31 and a second bracket 32 ​​which are separately arranged; wherein the first bracket 31 is fixedly installed at the front end of the GPU circuit board, and the second bracket 32 ​​is fixedly installed at the rear end of the GPU circuit board; and after the adapter circuit board 40 is plugged and installed with the GPU circuit board, the first bracket 31 and the second bracket 32 ​​are used to be fixedly installed with both ends of the adapter circuit board 40 respectively, so that the fixed bracket 30 is used to be fixedly installed with the adapter circuit board 40 after the GPU circuit board and the adapter circuit board 40 are plugged and installed; the first bracket 31 and the second bracket 32 ​​are also used to be installed on the server chassis 100 respectively, so that the fixed bracket 30 is used to install the GPU module 300 formed after the GPU circuit board and the adapter circuit board 40 are plugged and installed on the server chassis 100.

[0091] That is, see Figure 5 The fixed bracket of this embodiment can be specifically divided into a first bracket and a second bracket, wherein the first bracket is fixed to the front end of the GPU circuit board, and the second bracket is fixed to the rear end of the GPU circuit board; then, after the adapter circuit board and the GPU circuit board are plugged and installed, the first bracket at the front end and the second bracket at the rear end can be fixed and installed at both ends of the adapter circuit board respectively; in addition, combined with Fig.11 , Fig.12 The first bracket and the second bracket are also used to be installed on the server chassis respectively.

[0092] It can be seen that by adopting the first bracket and the second bracket which are separately arranged, the weight of the GPU module can be further reduced.

[0093] In a specific embodiment, the first bracket 31 includes a first baffle 311 and a first mounting ear 312 connected by a vertical bend, wherein the first baffle 311 is fixedly installed at the front end of the GPU circuit board and abuts the front end surface of the packaging shell 10, and the first mounting ear 312 is arranged parallel to the adapter circuit board 40 and is used to be fixedly installed with one end of the adapter circuit board 40; the second bracket 32 ​​includes a second baffle 321 and a second mounting ear 322 connected by a vertical bend, wherein the second baffle 321 is fixedly installed at the rear end of the GPU circuit board and abuts the rear end surface of the packaging shell 10, and the second mounting ear 322 is arranged parallel to the adapter circuit board 40 and is used to be fixedly installed with the other end of the adapter circuit board 40.

[0094] That is, combined Figure 5 , Fig.14 The first bracket and the second bracket can specifically use a baffle and a mounting ear connected by a vertical bend, wherein the baffle abuts the packaging shell and is fixedly installed with the GPU circuit board, and the mounting ear abuts the adapter circuit board and is fastened and installed. The structure is simple and easy to install and use.

[0095] In one possible implementation, in the area of ​​the installation space 50 between the corresponding packaging shell 10 and the adapter circuit board 40, the first baffle 311 and the second baffle 321 are respectively provided with a plug-in notch 60; wherein, relative to the height direction of the server chassis, the plug-in notch 60 is located at an end of the first baffle 311 and the second baffle 321 away from the GPU circuit board, so that the plug-in notch 60 is used for a pair of GPU modules 300 to be plugged in side by side and staggered when the front and back sides are opposite to each other along the height direction of the server chassis and a pair of adapter circuit boards 40 are placed relative to each other.

[0096] See also Figure 4 , Figure 5In order to ensure the smooth parallel and staggered plugging mentioned above, the first baffle and the second baffle should be provided with plugging notches in the area corresponding to the installation space, and the plugging notches should be arranged away from the GPU circuit board ( Figure 4 and Figure 5 The GPU circuit board in the package is located at one end of the package shell.

[0097] In a specific implementation, along the height direction of the server chassis, the height of the insertion notch 60 is not less than half of the height of the GPU module 300 .

[0098] It can be understood that a pair of GPU modules can be staggered and plugged in horizontally only when the height of the plug-in gap is not less than half the height of the GPU module.

[0099] In the specific implementation, see Figure 4 or Figure 5 Considering that the height of the first baffle is usually consistent with the height of the packaging shell, at this time, the insertion gap of the first baffle in the corresponding installation space area can be specifically realized in the form of a gap; see Fig. 9 and Fig.14 Since the second baffle is located inside the server chassis, its height is not required to be consistent with the height of the packaging shell. At this time, the second baffle can be formed as a whole or at least the second baffle is set in the area corresponding to the installation space in the form of a beam to form a plug-in gap.

[0100] In one possible implementation, along the width direction of the server chassis, a pair of first positioning columns 3111 are provided on both sides of the outer surface of the first baffle 311, and the first positioning columns 3111 are used to be inserted into the first waist hole 101 set along the height direction of the server chassis 100; and the second baffle 321 is provided with a mounting ear 3211, and the mounting ear 3211 is used to be fastened to the server chassis 100 after the first positioning column 3111 is inserted into the first waist hole 101.

[0101] That is, combined Fig.11 and Fig.12 In order to achieve the purpose of installing the GPU module on the server chassis through the fixing bracket, this embodiment specifically sets a pair of first positioning posts on both sides of the outer surface of the first baffle, and then, correspondingly, a first waist hole for the first positioning post to be inserted is provided on the server chassis; wherein, the first waist hole is arranged along the height direction of the server chassis; thus, it can be understood that when installing the GPU module on the server chassis, the front end of the GPU module can be firstly inserted into the pair of first waist holes through the pair of first positioning posts respectively, and then, combined with Fig.13 and Fig.14, and then slowly lower the rear end of the GPU module. After it is placed, the GPU module can be fastened to the server chassis through the mounting ears of the second baffle.

[0102] It can be seen that in this embodiment, the GPU module can be quickly installed on or removed from the server chassis by using the first baffle and the second baffle, thereby achieving simple and convenient rapid disassembly and assembly.

[0103] Specifically, the mounting ears 3211 of the second blocking piece 321 may be bent vertically relative to the second blocking piece 321 , and the mounting ears 3211 are used to be fastened to the horizontally arranged mounting beams 104 of the server chassis 100 .

[0104] In a specific embodiment, in the area between a pair of first positioning posts 3111, a second positioning post 3112 and a third positioning post 3113 are further provided on the outer surface of the first baffle 311; wherein the second positioning post 3112 and the third positioning post 3113 are arranged up and down along the height direction of the server chassis, and the second positioning post 3112 and the third positioning post 3113 are staggered relative to the height direction of the server chassis, so that when the GPU card module is installed on the server chassis 100 with the front and back sides opposite to each other along the height direction of the server chassis, the second positioning post 3112 and the third positioning post 3113 are respectively plugged into the second waist hole 102 of the server chassis 100.

[0105] That is, in order to ensure the stable installation of the front end of the GPU module, the present embodiment further provides a second positioning post and a third positioning post on the outer surface of the first baffle.

[0106] Among them, see Fig.12 The second positioning column and the third positioning column are located in the area between the pair of first positioning columns, for example, approximately in the middle of the pair of first positioning columns, and then the second positioning column and the third positioning column are arranged up and down along the height direction of the server chassis, and relative to the height direction of the server chassis, the second positioning column and the third positioning column are also arranged in a certain offset; thus, on the basis that the second waist hole is also provided at the installation position corresponding to each GPU module of the server chassis, the second positioning column and the third positioning column can be respectively plugged into the second waist hole of the server chassis when the GPU module is installed in the server chassis with the front and back sides opposite to each other along the height direction of the server chassis.

[0107] For example, combining Fig.12 , along the height direction of the server chassis, if the middle position of the first baffle is far from the GPU circuit board end is the second positioning column, and the one close to the GPU circuit board end is the third positioning column, then, Fig.12 The GPU module shown in the figure is specifically plugged into the second waist hole of the server chassis through the second positioning column. Fig.12Since the GPU module next to the GPU module shown in the figure needs to be placed inversely along the height direction of the server chassis, the GPU module next to it needs to be plugged into the second waist hole of the server chassis through the third positioning column.

[0108] Based on the above GPU card module, the present application also discloses a GPU server, see Fig.16 , the GPU server includes a server chassis 100 and a first GPU card module 201 installed inside the server chassis 100;

[0109] Among them, along the width direction of the server chassis, four first GPU card modules 201 are installed side by side inside the server chassis 100, and the first GPU card modules 201 are the above-mentioned GPU card modules.

[0110] In other implementations, two second GPU card modules 202 are further installed inside the server chassis 100; wherein the GPU circuit board in the second GPU card module 202 is a half-height single-width GPU card.

[0111] In one possible implementation, the server chassis 100 includes a front-end bracket 103 and a mounting beam 104 for mounting a first GPU card module 201; wherein the front-end bracket 103 is provided with a pair of first waist holes 101 at positions corresponding to both sides of each first GPU card module 201, the first waist holes 101 are arranged along the height direction of the server chassis 100, and the first waist holes 101 are used for plugging and installing a pair of first positioning columns 3111 in the first GPU card module 201; the front-end bracket 103 is provided with a second waist hole 102 in the area between the pair of first waist holes 101, and the second waist hole 102 is used for plugging and installing a second positioning column 3112 or a third positioning column 3113 in the first GPU card module 201; the mounting beam 104 is used for fastening and installing the mounting ears 3211 in the first GPU card module 201.

[0112] In one possible implementation, the GPU server includes a server mainboard and an adapter circuit board 40 for plugging and installing with a first GPU card module; wherein, unlike the existing solution in which the adapter circuit board is directly plugged into the server mainboard, in this embodiment, since adjacent GPU modules need to be placed in opposite directions in height, at this time, the adapter circuit board should be electrically connected to the server mainboard via an adapter cable. Specifically, the adapter circuit board can be welded with an adapter cable (not shown in the figure), and then the adapter cable is plugged into the server mainboard via an adapter socket, thereby being electrically connected to the server mainboard.

[0113] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0114] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0115] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0116] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0117] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize that some variations, modifications, changes, additions and sub-combinations thereof should be included within the scope of protection of the present invention.

Claims

1. A GPU card module, characterized in that: The GPU card module includes: A GPU circuit board encapsulated by an encapsulation shell, wherein a PCI-E interface of the GPU circuit board is exposed from the encapsulation shell, and the PCI-E interface is used to be plugged into an adapter circuit board, so that the GPU circuit board and the adapter circuit board are plugged and installed to form a GPU module; A fixing bracket fixedly mounted on the GPU circuit board, the fixing bracket being used to fix the GPU circuit board with the adapter circuit board after the GPU circuit board and the adapter circuit board are plugged and installed, and to install the GPU module formed after the GPU circuit board and the adapter circuit board are plugged and installed on a server chassis; Wherein, when a pair of the GPU modules are arranged with their front and back sides opposite to each other along the height direction of the server chassis, and a pair of the adapter circuit boards are placed opposite to each other, the fixing bracket is configured so as not to hinder the pair of GPU modules from being inserted side by side and staggered using the installation space between the packaging shell and the adapter circuit board, so that when the GPU circuit board is a full-height double-width GPU card, four GPU modules are installed side by side and staggered along the width direction of the server chassis; Wherein, the fixed bracket includes a first bracket and a second bracket which are separately arranged; The first bracket is fixedly mounted on the front end of the GPU circuit board, and the first bracket includes a first baffle and a first mounting ear connected by vertical bending, the first baffle is fixedly mounted on the front end of the GPU circuit board and abuts against the front end surface of the packaging shell, and the first mounting ear is arranged parallel to the adapter circuit board and is used to be fixedly mounted on one end of the adapter circuit board; The second bracket is fixedly mounted on the rear end of the GPU circuit board, the second bracket comprises a second baffle and a second mounting ear connected by vertical bending, the second baffle is fixedly mounted on the rear end of the GPU circuit board and abuts against the rear end surface of the packaging shell, and the second mounting ear is arranged parallel to the adapter circuit board and is used to be fixedly mounted on the other end of the adapter circuit board; Furthermore, in an area corresponding to the installation space between the packaging shell and the adapter circuit board, the first baffle and the second baffle are respectively provided with plug-in notches; relative to the height direction of the server chassis, the plug-in notches are located at one end of the first baffle and the second baffle away from the GPU circuit board, so that the plug-in notches are used for a pair of GPU modules to be plugged in side by side and staggered when the front and back sides are opposite to each other along the height direction of the server chassis and a pair of adapter circuit boards are placed relative to each other.

2. The GPU card module according to claim 1, characterized in that: After the adapter circuit board is plugged and installed with the GPU circuit board, the first bracket and the second bracket are used to be fixedly installed with the two ends of the adapter circuit board respectively, so that the fixing bracket is used to be fixedly installed with the adapter circuit board after the GPU circuit board and the adapter circuit board are plugged and installed; The first bracket and the second bracket are also used to be installed on the server chassis respectively, so that the fixed bracket is used to install the GPU module formed by plugging and installing the GPU circuit board and the adapter circuit board on the server chassis.

3. The GPU card module according to claim 1, characterized in that: Along the height direction of the server chassis, the height of the insertion notch is not less than half of the height of the GPU module.

4. The GPU card module according to claim 1, characterized in that: Along the width direction of the server chassis, a pair of first positioning columns are provided on both sides of the outer surface of the first baffle, and the first positioning columns are used to be plugged into first waist holes provided along the height direction of the server chassis; In addition, the second baffle is provided with a mounting ear, and the mounting ear is used to be fastened to the server chassis after the first positioning column is inserted into the first waist hole.

5. The GPU card module according to claim 4, characterized in that: In the area between the pair of the first positioning posts, the outer surface of the first baffle is further provided with a second positioning post and a third positioning post; Among them, the second positioning column and the third positioning column are arranged up and down along the height direction of the server chassis, and the second positioning column and the third positioning column are staggered relative to the height direction of the server chassis, so that when the GPU card module is installed on the server chassis with the front and back sides opposite to each other along the height direction of the server chassis, the second positioning column and the third positioning column are respectively inserted into the second waist hole of the server chassis.

6. The GPU card module according to claim 4, characterized in that: The mounting ear is bent vertically relative to the second baffle, and the mounting ear is used to be fastened to a horizontally mounted crossbeam of the server chassis.

7. A GPU server, characterized in that: The GPU server includes a server chassis and a first GPU card module installed inside the server chassis; Among them, along the width direction of the server chassis, four first GPU card modules are installed side by side inside the server chassis, and the first GPU card module is the GPU card module according to any one of claims 1 to 6.

8. The GPU server according to claim 7, characterized in that: Two second GPU card modules are also installed inside the server chassis; Among them, the GPU circuit board in the second GPU card module is a half-height single-width GPU card.

9. The GPU server according to claim 7, characterized in that: The server chassis includes a front bracket and a mounting crossbeam for mounting the first GPU card module; The front-end bracket is provided with a pair of first waist holes at positions corresponding to both sides of each of the first GPU card modules, the first waist holes are arranged along the height direction of the server chassis, and the first waist holes are used for inserting and installing a pair of first positioning columns in the first GPU card module; the front-end bracket is provided with a second waist hole in the area between the pair of first waist holes, and the second waist hole is used for inserting and installing a second positioning column or a third positioning column in the first GPU card module; The mounting crossbeam is used for fastening and mounting the mounting ears in the first GPU card module.

10. The GPU server according to claim 7, characterized in that: The GPU server includes a server mainboard and a transfer circuit board for plugging and installing with the first GPU card module; Wherein, a transfer cable is welded on the transfer circuit board, and the transfer cable is electrically connected to the server mainboard through a transfer socket.

Images