A GPU server

Abstract

The embodiment of the application discloses a GPU server, which comprises a server case and a built-in GPU module; the server case is provided with a CPU module at the middle position in the depth direction of the server case, and a radiator is arranged above the CPU module; the built-in GPU module is arranged above the radiator and extends along the width direction of the server case; according to the standard width size of the server case and the size of a GPU card, four first GPU card modules can be arranged side by side along the width direction of the server case, and the first GPU card of each first GPU card module is a half-height single-width GPU card; by arranging the built-in GPU module capable of mounting four half-height single-width GPU cards side by side above the radiator, the GPU server of the embodiment increases the number of GPU cards and improves the computing performance, and solves the technical problem that the number of GPU cards in the existing GPU server is small and cannot meet the computing requirement.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of servers, and in particular to a GPU server. BACKGROUND

[0002] With the rapid development of artificial intelligence and big data, people have stronger demand for supercomputing capabilities, and for GPU servers, placing more GPU cards in limited space is the hardware basis for improving the computing capabilities of GPU servers.

[0003] The server chassis of a GPU server is usually of a standard size, such as a 2U chassis.

[0004] In related solutions, multiple GPU cards are usually stacked and installed at the rear end in the depth direction of the server chassis. Due to the size limitations of the server chassis and the GPU cards, currently, the server chassis can at most accommodate eight full-height or half-height single-width GPU cards, which cannot meet the computing requirements of high performance. SUMMARY

[0005] To at least one aspect of the above technical problems, the embodiments of the present application provide a GPU server, which sets a built-in GPU module above a heat sink for cooling a CPU module in a server chassis. The built-in GPU module can install four half-height single-width GPU cards side by side along the width direction of the server chassis, thereby increasing the number of GPU cards in the server chassis and improving the computing capability of the GPU server.

[0006] The embodiments of the present application provide a GPU server, which comprises a server chassis and a built-in GPU module.

[0007] The server chassis is provided with a CPU module at the middle position in the depth direction, and a heat sink for cooling the CPU module is installed above the CPU module.

[0008] In the height direction of the server chassis, the built-in GPU module is arranged above the heat sink and extends along the width direction of the server chassis.

[0009] The built-in GPU module comprises four first GPU card modules, which are installed side by side along the width direction of the server chassis, and each first GPU card module comprises one first GPU card, which is a half-height single-width GPU card.

[0010] In an embodiment, preferably, the built-in GPU module comprises a mounting bracket for the side-by-side installation of the four first GPU card modules.

[0011] The two ends of the mounting bracket are respectively provided with bracket side plates in the width direction of the server case, and the bracket side plates are arranged in parallel with the case side plates of the server case.

[0012] The bracket side plates are provided with hooks on the side facing the case side plates, and the case side plates are provided with clamping grooves on the side facing the bracket side plates, which are used for clamping the hooks.

[0013] In an embodiment, preferably, the GPU server comprises:

[0014] A heat dissipation module is arranged beside the built-in GPU module and on the side of the built-in GPU module facing the front end of the server case, so that the heat dissipation module dissipates heat from the heat sink and the built-in GPU module at the same time.

[0015] The heat dissipation module comprises four heat dissipation fans arranged side by side in the width direction of the server case.

[0016] The end of the mounting bracket facing the heat dissipation fans is provided with an air inlet, and the heat dissipation fans at the two ends in the width direction of the server case are provided with backflow prevention foam between the heat dissipation fans and the case side plates.

[0017] In an embodiment, preferably, the GPU server comprises:

[0018] A power supply module is fixedly installed at the rear end of the server case, and the power supply module is located at one end in the width direction of the server case.

[0019] A pair of second GPU card modules are arranged above the power supply module in the height direction of the server case and arranged side by side in the width direction of the server case.

[0020] A pair of third GPU card modules are located at the other end in the width direction of the server case, and the pair of third GPU card modules are arranged side by side in the width direction of the server case.

[0021] The second GPU card module comprises two second GPU cards stacked up and down, and the second GPU card is a half-height single-width GPU card.

[0022] The third GPU card module comprises three third GPU cards stacked up and down, and the third GPU card is a full-height single-width GPU card or a half-height single-width GPU card.

[0023] In an embodiment, preferably, the power supply module comprises a power supply housing and a power supply installed in the power supply housing; wherein the power supply housing is arranged at the end of the power supply facing the outside of the server case;

[0024] The second GPU card module comprises a second adapter support for fixing and installing the second GPU card; wherein the outer end of the second GPU card is arranged in a recessed manner compared with the outer end of the second adapter support along the depth direction of the server case, so that the second GPU card avoids the power supply housing.

[0025] In an embodiment, preferably, the second GPU card comprises a second baffle;

[0026] Wherein the second baffle is arranged in a bent manner towards the outside of the server case along one end in the width direction of the server case, and the bent part of the second baffle close to the power supply module is arranged in a protruding manner towards the power supply module relative to the second adapter support along the height direction of the server case;

[0027] And the power supply housing is provided with a plug-in notch at the position corresponding to the bent part, and the plug-in notch is used for plug-in installation of the bent part arranged in a protruding manner relative to the second adapter support.

[0028] In an embodiment, preferably, the end of the rear end of the server case is provided with a guide sliding groove extending along the height direction of the server case;

[0029] Wherein the two sides of the outer end of the second GPU card module and the third GPU card module are plugged into the guide sliding groove, and the inner end of the second GPU card module and the third GPU card module is fastened to the server case.

[0030] In an embodiment, preferably, the GPU server comprises:

[0031] A power supply module is fixedly installed at the rear end of the server case, and the power supply module is located at one end in the width direction of the server case;

[0032] Three fourth GPU card modules are arranged side by side along the width direction of the server case in the area above the power supply module;

[0033] A pair of fifth GPU card modules are arranged below the three fourth GPU card modules and beside the power supply module, and the pair of fifth GPU card modules are arranged side by side along the width direction of the server case;

[0034] The fourth GPU card module comprises a fourth GPU card, and the fourth GPU card is a full-height double-wide GPU card.

[0035] The fifth GPU card module comprises a fifth GPU card, and the fifth GPU card is a half-height single-wide GPU card or a full-height single-wide GPU card.

[0036] In an embodiment, preferably, the GPU server comprises:

[0037] Twelve hard disk modules are arranged at the front end of the server case.

[0038] Among them, along the height direction of the server case, three hard disk modules are arranged in a stacked manner, and along the width direction of the server case, four hard disk modules are arranged side by side.

[0039] And the hard disk module comprises a hard disk support and a hard disk module mounted on the hard disk support, the hard disk module is a 3.5-inch hard disk or a GPU computing power disk, and the GPU computing power disk comprises a GPU chip.

[0040] In an embodiment, preferably, the hard disk support comprises a support protection plate extending along the width direction of the server case, and both ends of the support protection plate extend in the same direction to form a pair of support fixing arms, which are used to be inserted into the server case along the depth direction of the server case after the hard disk module is mounted.

[0041] Among them, the support protection plate is provided with a honeycomb-shaped air inlet area.

[0042] In an embodiment, preferably, along the height direction of the server case, at least one end of the support fixing arm is provided with a damping protruding bridge; and / or,

[0043] The support fixing arm is provided with a first elastic sheet extending along the depth direction of the server case, one end of the first elastic sheet is riveted to the side of the support fixing arm away from the hard disk module, and the other end of the first elastic sheet passes through the support fixing arm and is limited to abut against the limiting groove on the side of the support fixing arm facing the hard disk module; and / or,

[0044] The support fixing arm is provided with a second elastic sheet arranged on the side of the support fixing arm away from the hard disk module and protruding.

[0045] In an embodiment, preferably, the first GPU card module comprises a first adapter support and a first card pressing support.

[0046] One end of the first GPU card is fastened to the first adapter support through a first blocking sheet along the depth direction of the server case, and the other end of the first GPU card is fastened to the first adapter support through the first card pressing support.

[0047] The one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0048] The GPU server provided by the embodiments of the present application comprises a server case, and a built-in GPU module is installed in the inside of the server case.

[0049] That is, the server case inside the GPU server is optimized in the embodiments, the radiator for cooling the CPU module can be replaced by a 1U radiator with a lower height, so that the space above the radiator can be used for installing the built-in GPU module, and the built-in GPU module is installed above the radiator, and the radiator fan beside the radiator can simultaneously cool the radiator and the built-in GPU module, so that the cooling requirement is met. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0051] Figure 1 The layout structure of the inside of the server case in the embodiments of the present application is shown in the figure.

[0052] Figure 2 The layout structure of the rear end of the server case in the embodiments of the present application is shown in the figure.

[0053] Figure 3 Structure diagram of the built-in GPU module installed in the middle position of the server case in the embodiment of the present application.

[0054] Figure 4 Structure diagram of the second GPU card module and the third GPU card module installed in the rear end of the server case in the embodiment of the present application.

[0055] Figure 5 Structure diagram of the built-in GPU module in the embodiment of the present application.

[0056] Figure 6 Structure diagram of the server case in the embodiment of the present application, wherein the rear end of the server case is not installed with the second GPU card module and the third GPU card module.

[0057] Figure 7 Structure diagram of the server case in the embodiment of the present application, wherein the rear end of the server case is installed with a pair of third GPU card modules, and is not installed with the second GPU card module.

[0058] Figure 8 Structure diagram of the second GPU card module in the embodiment of the present application.

[0059] Figure 9 Structure diagram of the server case in the embodiment of the present application, wherein the rear end of the server case is installed with a pair of third GPU card modules and a second GPU card module. Figure 8 Structure diagram from another perspective.

[0060] Figure 10 Structure diagram of the server case in the embodiment of the present application, wherein the rear end of the server case is installed with a pair of third GPU card modules and a second GPU card module.

[0061] Figure 11 Structure diagram of the server case in the embodiment of the present application, wherein the rear end of the server case is installed with a pair of third GPU card modules and a second GPU card module. Figure 10 Enlarged view of the structure of the server case in the embodiment of the present application.

[0062] Figure 12 Structure diagram of the layout of the rear end of the server case in another embodiment of the present application, wherein the rear end of the server case is installed with three fourth GPU card modules and a pair of fifth GPU card modules.

[0063] Figure 13 Structure diagram of the cooling fan in the embodiment of the present application.

[0064] Figure 14 Structure diagram of the hard disk module in the embodiment of the present application, wherein the hard disk support can selectively install two kinds of hard disk modules.

[0065] Figure 15 Possible structure diagram of the card pressing support in the embodiment of the present application.

[0066] wherein the reference signs:

[0067] 100 - server case, 101 - case side plate, 102 - card joint groove, 103 - case air inlet, 104 - case air outlet,

[0068] 300 - radiator,

[0069] 400 - heat dissipation module, 401 - heat dissipation fan,

[0070] 500 - power supply module, 501 - power supply case, 502 - power supply, 503 - plug-in gap,

[0071] 700 - guide sliding groove,

[0072] 10 - first GPU card module, 11 - first GPU card,

[0073] 20 - second GPU card module, 21 - second GPU card, 22 - second adapter support, 23 - second baffle,

[0074] 231 - bending part,

[0075] 30 - third GPU card module, 31 - third GPU card, 32 - third adapter support,

[0076] 40 - fourth GPU card module,

[0077] 50 - fifth GPU card module,

[0078] 60 - built-in GPU module, 61 - mounting support, 62 - support side plate, 63 - hanging nail, 64 - air inlet,

[0079] 70 - hard disk module, 71 - hard disk support, 72 - hard disk module,

[0080] 711 - support protection plate, 712 - support fixing arm, 713 - air inlet area, 714 - damping convex bridge, 715 - first elastic sheet, 716 - second elastic sheet,

[0081] 80 - card pressing support,

[0082] X - width direction, Y - height direction, Z - depth direction. DETAILED DESCRIPTION

[0083] In order to better understand the above technical solutions, the example embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments of the present application. It should be understood that the present application is not limited by the example embodiments described herein.

[0084] Regarding the layout of the existing GPU server, the front end of the server case is provided with 12 hard disk modules, the rear end of the server case can be provided with a maximum of 8 GPU cards, i.e. 8 half-height single-width GPU cards or 8 full-height single-width GPU cards, then, at the middle position of the depth direction of the server case, from low to high along the height direction, it is CPU module, radiator and wind deflector in turn, and the CPU module and the hard disk module are heat dissipation fans.

[0085] That is, the layout of the existing GPU server is to set the GPU card only at the rear end of the server case, and the number of GPU cards can be a maximum of 8, which is less and cannot meet the computing demand.

[0086] In view of the above situation, the GPU server of the present application optimizes the internal layout of the server case:

[0087] First, the built-in GPU module is arranged above the radiator for cooling the CPU module in the server case, and the built-in GPU module can install four half-height single-width GPU cards side by side along the width direction of the server case.

[0088] Second, ten GPU cards can be arranged at the rear end of the server case by optimizing the mounting structure.

[0089] Third, twelve hard disk modules can be arranged at the front end of the server case, and each hard disk module can choose to install a 3.5-inch hard disk or a GPU computing disk.

[0090] In this way, the number of GPU cards in the server case is increased, and the computing power of the GPU server is improved, thereby solving the above problems.

[0091] First of all, it needs to be pointed out that the width direction X, the height direction Y and the depth direction Z mentioned in the present application are relative to the horizontal insertion of the case frame into the server case.

[0092] The GPU card module is arranged inside the server case, and the GPU card module usually includes a GPU card and an adapter bracket for fixing the GPU card, and the adapter bracket can fix one GPU card, stack two or three GPU cards, etc.; wherein, when the GPU card module is installed in the server case, the GPU card is horizontally placed, i.e. the width direction of the server case is the height of the GPU card, the depth direction of the server case is the length of the GPU card, and the height direction of the server case is the width of the GPU card.

[0093] The standard GPU card refers to a standard card conforming to the PCI-E specification. At present, there are various commonly used specifications of GPU cards. On the basis of the standard GPU card, the height is halved to obtain a half-height GPU card, the length is halved to obtain a half-length GPU card, and the width is doubled to obtain a double-width GPU card. That is, the card with a standard PCI-E socket is referred to as a full-height single-width full-length GPU card. On the basis of the full-height single-width full-length GPU card, the height is halved (the dimension along the width direction of the server case is halved) to obtain a half-height single-width GPU card, and the width is doubled (the dimension along the height direction of the server case is doubled) to obtain a full-height double-width GPU card.

[0094] In addition, the server case in the present application is a standard 2U case. In the server and equipment room, U is the abbreviation of unit, which represents the unit of rack size or external size of the server, and 1U is equal to 4.445 cm.

[0095] Please refer to Figures 1-4 The first embodiment of the present application discloses a GPU server, which comprises a server case 100 and a built-in GPU module 60. The CPU module is arranged at the middle position of the server case 100 in the depth direction Z. The heat sink 300 for dissipating heat of the CPU module is arranged above the CPU module. The built-in GPU module 60 is arranged above the heat sink 300 in the height direction Y of the server case 100 and extends in the width direction X of the server case 100. The built-in GPU module 60 comprises four first GPU card modules 10. The four first GPU card modules 10 are arranged side by side in the width direction X of the server case 100. The first GPU card module 10 comprises a first GPU card 11, and the first GPU card 11 is a half-height single-width GPU card.

[0096] That is, as described above, in the existing server case, the CPU module, the heat sink for dissipating heat of the CPU module and the air duct are arranged at the middle position in the depth direction. In view of this layout, the air duct is removed, and the heat sink is replaced by a 1U height heat sink (which can be a 1U height fin heat sink). In this way, the space above the heat sink at the middle position of the server case is left, and the space can be used to install the built-in GPU module.

[0097] The built-in GPU module extends in the width direction of the server case, and the built-in GPU module can be provided with four half-height single-width GPU cards side by side.

[0098] Specifically, the built-in GPU module comprises four first GPU card modules arranged side by side. Each first GPU card module comprises a first GPU card, and the first GPU card is a half-height single-width GPU card.

[0099] The embodiment of the application provides a GPU server, which comprises a server case, and a built-in GPU module is arranged in the server case; wherein, a CPU module is arranged at the middle position of the server case in the depth direction, and a radiator is arranged above the CPU module for heat dissipation; then, the built-in GPU module of the embodiment is arranged above the radiator and extends along the width direction of the server case; according to the standard width size of the server case and the size of the GPU card, four first GPU card modules can be arranged side by side along the width direction of the server case, each first GPU card module comprises a first GPU card, and the first GPU card is a half-height single-width GPU card.

[0100] That is, the embodiment optimizes the inside of the server case of the GPU server, can replace the radiator for heat dissipation of the CPU module with a 1U radiator with lower height, so that the space above the radiator can be used for mounting the built-in GPU module; and the built-in GPU module is mounted above the radiator, and the radiator fan arranged beside the radiator can also dissipate heat for the radiator and the built-in GPU module, which meets the heat dissipation requirement, so that by arranging the built-in GPU module with four half-height single-width GPU cards arranged side by side above the radiator, the GPU server of the embodiment increases the number of GPU cards and improves the computing performance, and solves the technical problem that the number of GPU cards in the existing GPU server is small and cannot meet the computing requirement.

[0101] Regarding the installation of the built-in GPU module 60, in a specific embodiment, the built-in GPU module 60 comprises a mounting bracket 61 for mounting the four first GPU card modules 10 side by side; wherein, along the width direction of the server case 100, both ends of the mounting bracket 61 are respectively provided with bracket side plates 62, which are arranged in parallel with the case side plates 101 of the server case 100; and one side of the bracket side plate 62 facing the case side plate 101 is provided with a hanging nail 63, and one side of the case side plate 101 facing the bracket side plate 62 is provided with a clamping groove 102 for clamping and mounting the hanging nail 63.

[0102] In summary, the four first GPU card modules of the built-in GPU module can be modularly mounted through the mounting bracket.

[0103] Specifically, referring to Figure 3 and Figure 5, the mounting bracket can be, for example, in the form of a groove, the inside of the groove is used to place four first GPU card modules side by side, the mounting bracket extends along the width direction of the server case and is lapped on the case side plates on both sides of the server case; wherein, along the width direction of the server case, both ends of the mounting bracket are provided with bracket side plates, the bracket side plates are arranged in parallel with the case side plates, that is, arranged in the vertical direction, then, one side of the bracket side plate facing the case side plate is provided with a hanging nail, correspondingly, one side of the case side plate facing the bracket side plate is provided with a clamping groove; the clamping groove extends downward along the height direction of the server case from the end of the case side plate, and then extends horizontally towards, for example, the front end of the server case, so that when the mounting bracket is installed, it can be first inserted vertically into the clamping groove, and then pushed horizontally, that is, the clamping installation of the mounting bracket is realized, or in other words, pre-installation, and then the mounting bracket can be fastened to the server case by fastening screws and the like.

[0104] It can be seen that the built-in GPU module can realize the modular installation of four first GPU card modules by setting the mounting bracket, and simplify the installation and disassembly process.

[0105] Regarding the heat dissipation of the built-in GPU module 60, in a specific embodiment, the GPU server includes a heat dissipation module 400, which is arranged beside the built-in GPU module 60 and on the side of the built-in GPU module 60 facing the front end of the server case 100, so that the heat dissipation module 400 simultaneously dissipates heat for the heat sink 300 and the built-in GPU module 60; wherein, the heat dissipation module 400 includes four heat dissipation fans 401, which are installed side by side along the width direction of the server case 100; and one end of the mounting bracket 61 facing the heat dissipation fans 401 is provided with an air inlet 64, and the heat dissipation fans 401 at both ends along the width direction X of the server case 100 are provided with anti-backflow foam between the heat dissipation fans 401 and the case side plates 101 (not shown in the figure).

[0106] That is, since the built-in GPU module is arranged above the CPU module, the heat dissipation problem of the built-in GPU module can be solved by using the heat dissipation module for the CPU module; wherein, the heat dissipation module is arranged beside the built-in GPU module and on the side of the built-in GPU module facing the front end of the server case.

[0107] Specifically, the heat dissipation module includes four heat dissipation fans installed side by side, and one end of the above-mentioned mounting bracket facing the heat dissipation fans is provided with four air inlets, the four air inlets correspond one-to-one to the four first GPU card modules, and the other end of the mounting bracket away from the heat dissipation fans can be provided in a hollow manner, which is beneficial to the flow of cooling air.

[0108] Among them, considering the need to simultaneously dissipate heat for the built-in GPU module and the CPU module, refer to Figure 13The heat dissipation fan 401 can be specifically an 8056 fan formed by two 8028 fans connected in series, i.e., a single-rotor fan connected in series to form a double-rotor fan.

[0109] In addition, the server case can be provided with a case air inlet 103 corresponding to the position of the heat dissipation module, and a case air outlet 104 at the rear end of the server case. For example, the case air inlet 103 and the case air outlet 104 are specifically provided on the side plate of the server case.

[0110] Please refer to Figure 4 and Figures 6-11 The second embodiment of the present application discloses a GPU server, which comprises a server case 100 and a power supply module 500. The power supply module 500 is fixedly installed at the rear end of the server case 100, and the power supply module 500 is located at one end in the width direction of the server case 100. Along the height direction X of the server case 100, a pair of second GPU card modules 20 is arranged above the power supply module 500 and arranged side by side in the width direction X of the server case 100. At the other end in the width direction X of the server case 100, a pair of third GPU card modules 30 is arranged side by side in the width direction X of the server case 100. The second GPU card module 20 comprises two second GPU cards 21 stacked up and down, and the second GPU card 21 is a half-height single-width GPU card. The third GPU card module 30 comprises three third GPU cards 31 stacked up and down, and the third GPU card 31 is a full-height single-width GPU card or a half-height single-width GPU card.

[0111] That is, in summary, at the rear end of the server case, the present embodiment can arrange four half-height single-width GPU cards in the space above the power supply module, and at the same time, the present embodiment can arrange six half-height single-width or full-height single-width GPU cards in the space on one side of the power supply module, so that the number of GPU cards arranged at the rear end of the server case can reach ten.

[0112] Specifically, at the rear end of the server case, the power supply module is located at one end in the width direction of the server case, and then a pair of second GPU card modules can be arranged side by side in the space above the power supply module, wherein each second GPU card module comprises two second GPU cards stacked up and down, and the second GPU card should be a half-height single-width GPU card.

[0113] Then, a pair of third GPU card modules can be arranged side by side in the space on one side of the power supply module, wherein each third GPU card module comprises three third GPU cards stacked up and down, and the third GPU card can be a half-height single-width GPU card or a full-height single-width GPU card.

[0114] That is, considering the size of the GPU card (specifically the half-height and full-height size of the GPU card) and the width size of the server case (the width size of the server case is fixed at 446.6 mm), along the width direction of the server case, the width size of the server case is not enough to accommodate four full-height GPU cards side by side, because the height size of one full-height GPU card is 126.3 mm, and when four full-height GPU cards are placed side by side, the overall size will exceed the width size of the server case; therefore, the embodiment arranges a pair of second GPU card modules side by side in the space above the power module, wherein the second GPU cards in the second GPU card modules are half-height GPU cards, and the embodiment arranges a pair of third GPU card modules side by side in the space on one side of the power module, wherein the third GPU cards in the third GPU card modules can be full-height GPU cards.

[0115] That is, referring to Figure 3 , Figure 4 and Figure 10 , along the width direction of the server case, the embodiment arranges two full-height GPU cards and two half-height GPU cards side by side, and the overall size of the two full-height GPU cards and the two half-height GPU cards when placed side by side is actually 413 mm, which is less than the width size of the server case of 446.6 mm.

[0116] In a specific embodiment, the power module 500 includes a power case 501 and a power supply 502 mounted on the power case 501; wherein the power case 501 is arranged at the end of the power supply 502 facing the outside of the server case 100, and the second GPU card module 20 includes a second adapter bracket 22 for fixedly mounting the second GPU card 21; wherein along the depth direction Z of the server case 100, the outer end of the second GPU card 21 is arranged inwardly relative to the outer end of the second adapter bracket 22 (see Figure 8 and Figure 11 for details), so that the second GPU card 21 avoids the power case 501.

[0117] That is, above the power module, in order to successfully realize the installation of a pair of second GPU card modules along the height direction of the server case, the embodiment can specifically arrange the second GPU card in the second GPU card module inwardly relative to the power module, so that the outer end of the second GPU card avoids the power case, reducing the installation height, thereby realizing the installation of a pair of second GPU card modules.

[0118] Specifically, referring to Figure 11, the power supply shell of the power module is located at the outer end of the power supply, and then, in the installation, the outer end of the second adapter support of the second GPU card module is substantially flush with the power supply shell, and in the embodiment, the outer end of the second GPU card is arranged in a recessed manner compared with the outer end of the second adapter support. It can be understood that the recessed amplitude should be completely offset in the height direction of the server case relative to the outer end of the second GPU card and the power supply shell, so that the installation height of the pair of second GPU card modules can be reduced to meet the size requirements of the server case.

[0119] In a specific embodiment, the second GPU card 21 comprises a second baffle 23; wherein the second baffle 23 is arranged in a bending manner towards the outside of the server case 100 along one end in the width direction X of the server case 100, and the bending part 231 of the second baffle 23 close to the power module 500 is arranged in a protruding manner towards the power module 500 along the height direction Y of the server case 100 relative to the second adapter support 22; and the power supply shell 501 is provided with a plug-in notch 503 at a position corresponding to the bending part 231, and the plug-in notch 503 is used for plug-in installation of the bending part 231 arranged in a protruding manner relative to the second adapter support 22.

[0120] In actual installation, referring to Figures 8-11 , the bending part of the second baffle of the second GPU card close to the power module is protruded relative to the second adapter support, and then, the plug-in notch is provided on the power supply shell for the plug-in of the bending part, so that in the installation, the bending part protruded relative to the second adapter support can be plugged into the plug-in notch, so as not to increase the installation height due to the protruding bending part and cause installation failure.

[0121] In a specific embodiment, the end of the rear end of the server case 100 is provided with a guide sliding groove 700 extending along the height direction Y of the server case 100; wherein the two sides of the outer end of the second GPU card module 20 and the third GPU card module 30 are plugged into the guide sliding groove 700, and the inner end of the second GPU card module 20 and the third GPU card module 30 is fastened to the server case 100.

[0122] In the embodiment, referring back to Figure 6 , Figure 7 , for the fixed installation of the second GPU card module and the third GPU card module, a plurality of guide sliding grooves can be first arranged at the end of the rear end of the server case, then the two sides of the outer end of the second GPU card module and the third GPU card module can be plugged along the guide sliding groove, and then the inner end of the second GPU card module and the third GPU card module is fastened to the server case, so as to complete the fixed installation of the second GPU card module and the third GPU card module.

[0123] Please combineFigure 12 The third embodiment of the present application discloses a GPU server, which comprises a server case 100 and a power module 500. The power module 500 is fixedly installed at the rear end of the server case 100, and is located at one end of the server case 100 in the width direction X. Three fourth GPU card modules 40 are arranged side by side along the width direction of the server case 100 in the area above the power module 500. A pair of fifth GPU card modules 50 is arranged below the three fourth GPU card modules 40 and beside the power module 500, and the pair of fifth GPU card modules 50 is arranged side by side along the width direction X of the server case 100. The fourth GPU card module 40 comprises one fourth GPU card, and the fourth GPU card is a full-height double-wide GPU card. The fifth GPU card module 50 comprises one fifth GPU card, and the fifth GPU card is a half-height single-wide GPU card or a full-height single-wide GPU card.

[0124] That is, according to actual needs, the embodiment can also be provided with 3 full-height double-wide GPU cards and 2 half-height single-wide or full-height single-wide GPU cards at the rear end of the server case.

[0125] Specifically, please refer to Figure 12 On one side of the power module, two fifth GPU card modules can be installed side by side, wherein the fifth GPU card module comprises one fifth GPU card, which can be a half-height single-wide or full-height single-wide GPU card.

[0126] Then, in the space above the power module and the fifth GPU card module, three fourth GPU card modules can be installed side by side, wherein the fourth GPU card module comprises one fourth GPU card, which is a full-height double-wide GPU card.

[0127] Please refer to Figure 10 and Figure 14 The fourth embodiment of the present application discloses a GPU server, which comprises a server case 100. The front end of the server case 100 is provided with twelve hard disk modules 70. Along the height direction Y of the server case 100, three hard disk modules 70 are arranged in a stacked manner from top to bottom, and along the width direction X of the server case 100, four hard disk modules 70 are arranged side by side. The hard disk module 70 comprises a hard disk support 71 and a hard disk module 72 installed on the hard disk support 71. The hard disk module 72 is a 3.5-inch hard disk ( Figure 14 in the left side of the hard disk module 72) or a GPU computing power disk ( Figure 14 in the right side of the hard disk module 72), and the GPU computing power disk comprises a GPU chip.

[0128] In this embodiment, 12 hard disk modules are arranged side by side and stacked at the front end of the server case, wherein the 12 hard disk modules are arranged as follows:

[0129] Along the width direction of the server case, the four hard disk modules are installed side by side, and along the height direction of the server case, the three hard disk modules are installed stacked one above another, so as to form the 12 hard disk modules.

[0130] For each hard disk module, the hard disk module includes a hard disk support and a hard disk module installed on the hard disk support, and the hard disk module can be conveniently installed on the server case through the hard disk support.

[0131] Among them, the 12 hard disk modules can all be 3.5-inch hard disks, at this time, the front end of the server case is simultaneously installed with 12 3.5-inch hard disks; or the 12 hard disk modules can all be GPU computing disks, at this time, the front end of the server case is simultaneously installed with 12 GPU computing disks, the GPU computing disk includes a GPU chip, which can greatly improve the computing performance of the server case; or the 12 hard disk modules can be partially 3.5-inch hard disks and partially GPU computing disks.

[0132] Regarding the hard disk support 71, in a specific embodiment, the hard disk support 71 includes a support protection plate 711 extending along the width direction X of the server case 100, and a pair of support fixing arms 712 extending from both ends of the support protection plate 711 in the same direction, the support fixing arms 712 are used to be inserted into the server case 100 along the depth direction Z of the server case 100 after the hard disk module 72 is installed; wherein the support protection plate 711 is provided with a honeycomb-shaped air inlet area 713.

[0133] That is, referring to Figure 14 the hard disk support can specifically include a middle support protection plate, and then a support fixing arm is provided at both ends of the support protection plate in the same direction, wherein the hard disk module can be specifically fastened to the support fixing arm when the hard disk module is installed with the hard disk support; and then the hard disk support is inserted along the depth direction of the server case.

[0134] In a specific embodiment, at least one end of the support fixing arm 712 is provided with a damping protruding bridge 714 along the height direction Y of the server case 100; and / or the support fixing arm 712 is provided with a first elastic sheet 715 extending along the depth direction Z of the server case 100, one end of the first elastic sheet 715 is riveted to the side of the support fixing arm 712 away from the hard disk module 72, and the other end of the first elastic sheet 715 passes through the support fixing arm 712 and is limited to abut against the limiting groove on the side of the support fixing arm 712 facing the hard disk module 72; and / or the support fixing arm 712 is provided with a second elastic sheet 716, the second elastic sheet 716 is provided on the side of the support fixing arm 712 away from the hard disk module 72 and is protrudingly arranged.

[0135] That is, in order to enhance the shock absorption performance of the hard disk module, on the one hand, the embodiment can be provided with a shock absorption protruding bridge at one end or both ends of the support fixing arm along the height direction of the server; on the other hand, the embodiment can be provided with a first elastic sheet on the support fixing arm, wherein the first elastic sheet extends along the depth direction of the server case, one end of the first elastic sheet is riveted to the side of the support fixing arm away from the hard disk module, and then the other end of the first elastic sheet is limited to abut against the limiting groove on the side of the support fixing arm facing the hard disk module after passing through the support fixing arm, so that the first elastic sheet can at least absorb the shock between adjacent hard disk modules; on the other hand, the embodiment can be provided with a second elastic sheet on the support fixing arm, and the second elastic sheet is installed on the side of the support fixing arm away from the hard disk module and protrudes, so that the second elastic sheet can continue to absorb the shock between adjacent hard disk modules.

[0136] Based on the above embodiments, in one possible implementation, the GPU card module can also protect the GPU card fixed thereon by the card pressing support 80.

[0137] Generally, the GPU card is fixed to the adapter support by the baffle at one end, and the other end of the GPU card has no fixing structure. At this time, when the adapter support deforms during transportation, the fixed GPU card will also be raised. Therefore, the embodiment fastens the end of the GPU card away from the baffle to the adapter support by the card pressing support, so as to solve the problem of the tail end of the GPU card being raised due to the deformation of the adapter support during transportation.

[0138] In the GPU card module of the above embodiments, the card pressing support can be provided on the adapter support of the GPU card module.

[0139] For example, for the first GPU card module, the first GPU card module includes a first adapter support and a card pressing support; wherein along the depth direction of the server case, one end of the first GPU card is fastened to the first adapter support by the first baffle, and the other end of the first GPU card is fastened to the first adapter support by the card pressing support.

[0140] For example, for the second GPU card module, the second GPU card module includes a second adapter support and a card pressing support; wherein along the depth direction of the server case, one end of the second GPU card is fastened to the second adapter support by the second baffle, and the other end of the second GPU card is fastened to the second adapter support by the card pressing support.

[0141] Referring to Figure 15For example, for the third GPU card module 30, the third GPU card module 30 comprises a third adapter bracket 32 and a press-in bracket 80; wherein, along the depth direction of the server case, one end of the third GPU card 31 is fastened to the third adapter bracket 32 through a third blocking sheet, and the other end of the third GPU card 31 is fastened to the third adapter bracket 32 through the press-in bracket 80.

[0142] For example, for the fourth GPU card module, the fourth GPU card module comprises a fourth adapter bracket and a press-in bracket; wherein, along the depth direction of the server case, one end of the fourth GPU card is fastened to the fourth adapter bracket through a fourth blocking sheet, and the other end of the fourth GPU card is fastened to the fourth adapter bracket through the press-in bracket.

[0143] For example, for the fifth GPU card module, the fifth GPU card module comprises a fifth adapter bracket and a press-in bracket; wherein, along the depth direction of the server case, one end of the fifth GPU card is fastened to the fifth adapter bracket through a fifth blocking sheet, and the other end of the fifth GPU card is fastened to the fifth adapter bracket through the press-in bracket.

[0144] The above describes the basic principles of the present application in combination with specific embodiments, but it should be noted that the advantages, advantages, effects and the like mentioned in the present application are only examples and are not limited, and these advantages, advantages, effects and the like cannot be considered as the must-have of each embodiment of the present application. In addition, the above specific details are only for the purpose of example and for the purpose of understanding, and are not limited to the above specific details, and the above specific details do not limit the present application to be implemented with the above specific details.

[0145] The block diagrams of the devices, apparatuses, equipment, systems involved in the present application are only illustrative examples and are not intended to require or imply the connection, arrangement, configuration shown in the block diagram. As those skilled in the art will recognize, these devices, apparatuses, equipment, systems can be connected, arranged, configured in any way. Words such as "include", "contain", "have" and the like are open-ended words, which mean "include but not limited to", and can be used interchangeably. The words "or" and "and" used herein mean the word "and / or", and can be used interchangeably unless the context clearly indicates otherwise. The word "such as" used herein means the phrase "such as but not limited to", and can be used interchangeably.

[0146] It should also be noted that in the devices, equipment and methods of the present application, each component or each step can be decomposed and / or recombined. These decompositions and / or recombination should be considered as equivalent solutions of the present application.

[0147] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0148] The above description has been presented to enable any person skilled in the art to make or use the application. Numerous modifications and alterations to this application will be readily apparent to those skilled in the art, and it is intended to include all such modifications and alterations insofar as they come within the scope of the general concepts defined herein.

Claims

1. A GPU server, comprising: The GPU server comprises a server case and a built-in GPU module and a heat dissipation module, the server case is a 2U case; The CPU module is arranged at the middle position of the server case in the depth direction, a heat sink for dissipating heat of the CPU module is arranged above the CPU module, the height of the heat sink is 1U, and the heat sink is a fin heat sink; In the height direction of the server case, the built-in GPU module is arranged above the heat sink and extends in the width direction of the server case; The built-in GPU module comprises four first GPU card modules, the four first GPU card modules are arranged side by side in the width direction of the server case, the first GPU card module comprises one first GPU card, and the first GPU card is a half-height single-width GPU card; the first GPU card module comprises a first adapter support and a first card pressing support; in the depth direction of the server case, one end of the first GPU card is fastened to the first adapter support through a first baffle, and the other end of the first GPU card is fastened to the first adapter support through the first card pressing support; The server case is provided with a power module and a pair of second GPU card modules and a pair of third GPU card modules at the rear end position in the depth direction; or, The server case is provided with a power module and three fourth GPU card modules and a pair of fifth GPU card modules at the rear end position in the depth direction; The built-in GPU module comprises a mounting support, and the mounting support is used for side-by-side mounting of the four first GPU card modules; In the width direction of the server case, the two ends of the mounting support are respectively provided with support side plates, and the support side plates are arranged in parallel with the case side plates of the server case; One side of the support side plate facing the case side plate is provided with a hanging nail, and one side of the case side plate facing the support side plate is provided with a clamping groove for clamping the hanging nail; The heat dissipation module is arranged beside the built-in GPU module and on the side of the built-in GPU module facing the front end of the server case, so that the heat dissipation module dissipates heat for the heat sink and the built-in GPU module at the same time; and the heat dissipation module comprises four heat dissipation fans arranged side by side in the width direction of the server case; One end of the mounting support facing the heat dissipation fan is provided with an air inlet, and anti-backflow foam is arranged between the heat dissipation fan at the two ends of the server case in the width direction and the case side plate.

2. The GPU server of claim 1, wherein, When the server case is provided with a power module and a pair of second GPU card modules and a pair of third GPU card modules at the rear end position in the depth direction, the power module is located at one end of the server case in the width direction; In the height direction of the server case, the pair of second GPU card modules is arranged above the power module and arranged side by side in the width direction of the server case; The pair of third GPU card modules is located at the other end of the server case in the width direction, and the pair of third GPU card modules is arranged side by side in the width direction of the server case; The second GPU card module comprises two second GPU cards stacked in up and down directions, and the second GPU card is a half-height single-width GPU card. The third GPU card module comprises three third GPU cards stacked in up and down directions, and the third GPU card is a full-height single-width GPU card or a half-height single-width GPU card.

3. The GPU server of claim 2, wherein The power supply module comprises a power supply shell and a power supply installed on the power supply shell, and the power supply shell is arranged at an end of the power supply facing the outside of the server case. The second GPU card module comprises a second adapter support for fixedly installing the second GPU card, and an outer end of the second GPU card is arranged in a recessed manner relative to an outer end of the second adapter support in a depth direction of the server case, so that the second GPU card avoids the power supply shell.

4. The GPU server of claim 3, wherein, The second GPU card comprises a second baffle. The second baffle is arranged in a bent manner towards the outside of the server case at one end in a width direction of the server case, and a bent part of the second baffle close to the power supply module is arranged in a protruding manner towards the power supply module relative to the second adapter support in a height direction of the server case. The power supply shell is provided with a plug-in notch at a position corresponding to the bent part, and the plug-in notch is used for plug-in installation of the bent part arranged in a protruding manner relative to the second adapter support.

5. The GPU server of claim 2, wherein, An end of the rear end of the server case is provided with a guide sliding groove extending in the height direction of the server case. The outer ends of the second GPU card module and the third GPU card module are plugged into the guide sliding groove on both sides, and the inner ends of the second GPU card module and the third GPU card module are fastened to the server case.

6. The GPU server of claim 1, wherein, When the rear end of the server case in the depth direction is provided with a power supply module and three fourth GPU card modules and a pair of fifth GPU card modules, the power supply module is located at one end in the width direction of the server case. The three fourth GPU card modules are arranged side by side in the width direction of the server case in a region above the power supply module. The pair of fifth GPU card modules are arranged below the three fourth GPU card modules and beside the power supply module, and the pair of fifth GPU card modules are arranged side by side in the width direction of the server case. The fourth GPU card module comprises one fourth GPU card, and the fourth GPU card is a full-height double-width GPU card. The fifth GPU card module comprises one fifth GPU card, and the fifth GPU card is a half-height single-width GPU card or a full-height single-width GPU card.

7. The GPU server of claim 1, wherein, The GPU server comprises: Twelve hard disk modules arranged at the front end of the server case. In the height direction of the server case, three hard disk modules are arranged in a stacked manner in up and down directions, and in the width direction of the server case, four hard disk modules are arranged side by side. And the hard disk module includes a hard disk support and a hard disk module installed on the hard disk support, the hard disk module is a 3.5-inch hard disk or a GPU computing power disk, and the GPU computing power disk includes a GPU chip.

8. The GPU server of claim 7, wherein, The hard disk support includes a support protection plate extending along the width direction of the server cabinet, and a pair of support fixing arms extending from the two ends of the support protection plate in the same direction, and the support fixing arms are used to be inserted into the server cabinet along the depth direction of the server cabinet after the hard disk module is installed. The support protection plate is provided with a honeycomb-shaped air inlet area.

9. The GPU server of claim 8, wherein, At least one end of the support fixing arm is provided with a damping protruding bridge along the height direction of the server cabinet; and / or The support fixing arm is provided with a first elastic sheet extending along the depth direction of the server cabinet, one end of the first elastic sheet is riveted to the side of the support fixing arm away from the hard disk module, and the other end of the first elastic sheet passes through the support fixing arm and is limited to abut against a limiting groove on the side of the support fixing arm facing the hard disk module; and / or The support fixing arm is provided with a second elastic sheet, and the second elastic sheet is arranged on the side of the support fixing arm away from the hard disk module and protrudes.

Images