CN217426042U

CN217426042U - Server structure supporting 8GPU

Info

Publication number: CN217426042U
Application number: CN202221041561.2U
Authority: CN
Inventors: 王聪
Original assignee: Suzhou Inspur Intelligent Technology Co Ltd
Current assignee: Suzhou Inspur Intelligent Technology Co Ltd
Priority date: 2022-04-30
Filing date: 2022-04-30
Publication date: 2022-09-13
Anticipated expiration: 2032-04-30

Abstract

The utility model discloses a server structure supporting 8GPU, which comprises a case, a power module, a storage module, an interaction module, a heat dissipation module, a GPU module and an expansion module; the upper part in the case is provided with a power supply module, two sides of the power supply module are respectively provided with an expansion module left support and an expansion module right support, the lower side of the power supply module is sequentially provided with a storage module and an interaction module, the interaction module is connected with an external computing module, the lower part in the case is provided with a GPU module, the wind scooper is arranged on the upper side of the GPU module, and the rear side of the GPU module is provided with a heat dissipation module; the tail part of the case is provided with a placing area. The utility model discloses compact structure adopts the modularized design, including power module, storage module, interactive module, GPU module, heat dissipation module and expansion module, support 19 inches and 21 inches rack deployment simultaneously. And the GPU module and the external computing module are interconnected through the interaction module.

Description

Server structure supporting 8GPU

Technical Field

The utility model relates to a server technical field specifically is a support 8 GPU's server structure.

Background

With the development of artificial intelligence, the demand of the AI server in the market is increasing. The currently deployed cabinets of a data center room are mainly 19-inch standard cabinets and 21-inch ODCC cabinets. The AI server in the current market integrates a computing module and a GPU module together, so that the cost is high, the AI server cannot be adapted to cabinets with various sizes, and the application scene is limited.

SUMMERY OF THE UTILITY MODEL

The utility model discloses be exactly to the above-mentioned not enough that prior art exists, provide a server structure that supports 8GPU, can select different usage pattern, it is nimble convenient to use.

In order to realize the purpose, the utility model provides the following technical scheme:

a server structure supporting 8 GPUs comprises a case, a power supply module, a storage module, an interaction module, a heat dissipation module, a GPU module and an expansion module; the upper part in the case is provided with a power supply module, two sides of the power supply module are respectively provided with an expansion module left support and an expansion module right support, the lower side of the power supply module is sequentially provided with a storage module and an interaction module, the interaction module is connected with an external computing module, the lower part in the case is provided with a GPU module, the upper side of the GPU module is provided with an air guide cover, and the rear side of the GPU module is provided with a heat dissipation module; the tail part of the case is provided with a placing area.

Preferably, the chassis comprises an upper cover and a chassis base, wherein the upper cover comprises an upper cover main body, an upper cover lock and I-shaped nails.

Preferably, the power supply module comprises a power supply tray, a power baffle, a CRPS power supply and a power supply board.

Preferably, the interaction module includes an interaction tray, a switch board, and a PCIE card.

Preferably, the storage module comprises a storage tray, an IO board, an NVMESSD hard disk and a hard disk backplane.

Preferably, the number of the placing areas is 10, and the two placing areas positioned on one side of the tail of the case are respectively provided with a crown clamp support/IO plate support and a false fan support.

Preferably, the crown clip bracket comprises a crown clip bracket and a power adapter plate.

Compared with the prior art, the beneficial effects of utility model are that:

1. the utility model discloses compact structure adopts the modularized design, including power module, storage module, mutual module, GPU module, radiating module and expansion module, supports 19 inches and 21 inches rack deployment simultaneously.

The GPU module and the external computing module are interconnected through the interaction module, and an existing server of a client site can be used as the computing module. Meanwhile, the width of the case is adjusted through the extension module, the requirements of different cabinet sizes are met, the application range of the server is enlarged, the development cost is reduced, and the market competitiveness is improved.

2. The utility model discloses according to not unidimensional rack deployment demand, whether adopt the electricity mode of getting that extends module influence power module (adopt CRPS power module or get the electricity from the rack copper bar), to the deployment demand of rack promptly, the adaptability change is made to other parts of quick-witted incasement (storage module, place region etc.), and the flexible operation ensures to use reliably.

Drawings

Fig. 1 is an exploded schematic view of the present invention;

FIG. 2 is a schematic view of the upper cover;

FIG. 3 is a schematic diagram of a power supply module;

FIG. 4 is a schematic diagram of a memory module;

FIG. 5 is a schematic diagram of an interaction module;

FIG. 6 is a schematic diagram of the rear of the chassis;

FIG. 7 is a schematic view of a crown clip holder;

fig. 8 is a schematic view of an IO board bracket.

In the figure: 1-upper cover, 1-1 upper cover main body, 1-2 upper cover lock, 1-3I-shaped nail, 2-power module, 2-1 power tray, 2-2 power baffle, 2-3CRPS power, 2-4 power panel, 3-interactive module, 3-1 interactive tray, 3-2 exchange panel, 3-3PCIE card, 4-expansion module left support, 5-storage module, 5-1 storage tray, 5-2IO panel, 5-3NVMESSD hard disk, 5-4 hard disk backboard, 6-chassis base, 7-expansion module right support, 8-wind scooper, 9-GPU module, 10-heat dissipation module, 11-crown clamp support, 11-1 crown clamp bracket, 11-2 power adapter panel, 12-IO panel support, 12-1IO board bracket, 13-false fan bracket.

Detailed Description

The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments in the present invention, all other embodiments obtained by a person skilled in the art without creative work belong to the protection scope of the present invention.

Examples

A server structure supporting 8 GPUs has a whole height of 4U, and comprises a case, a power module 2, a storage module 5, an interaction module 3, a heat dissipation module 10 and a GPU module 9, as shown in figure 1. The case comprises an upper cover 1 and a case base 6, wherein the upper cover 1 comprises an upper cover main body 1-1, an upper cover lock 1-2 and I-shaped nails 1-3.

As shown in figure 2, the upper cover 1 is vertically installed downwards, after the upper cover lock 1-2 is matched with a nail on the power module 2 in place, a handle of the upper cover lock 1-2 is pressed flat, and the I-shaped nail 1-3 is clamped into an opening on the side wall of the case, so that the upper cover 1 is ensured to be installed firmly.

As shown in fig. 3, a power module 2 is disposed at an upper portion of the chassis, and the power module 2 is located at an uppermost layer 1U of the server. The power supply module 2 comprises a power supply tray 2-1, a power baffle 2-2, a CRPS power supply 2-3 and a power supply board 2-4. Two sides of the power supply tray 2-1 are fixed with the box body through screws, a power supply board 2-4 is arranged on the power supply tray and is responsible for supplying power to other modules of the server, and the power supply board 2-4 is fixed on studs and I-shaped nails at the bottom of the power supply tray 2-1 through two M3 screws.

When the server is deployed in a 21-inch ODCC cabinet, the server adopts the tail crown clamp bracket 11 to take electricity from the copper bar of the cabinet, and the power module 2 does not need to be inserted into the CRPS power supply 2-3. The front window of the power module 2 is blocked by a false screen plate, and the power baffle 2-2 is fixed on the front window of the power module 2 by using two M3 countersunk screws.

When the server is mounted in a 19 inch standard cabinet, the power module 2 can plug into a maximum of 4 CRPS power supplies 2-3. The whole power module 2 is fixed on the side wall of the case by using 6M 3 countersunk head screws.

As shown in fig. 1, the expansion module further comprises an expansion module left support 4 and an expansion module right support 7. And an expansion module left support 4 and an expansion module right support 7 are respectively arranged on two sides of the power module 2.

When the server is installed on a 21-inch cabinet, the expansion module left support 4 and the expansion module right support 7 are positioned through the I-shaped nails and the gourd holes in the side wall of the cabinet, and the expansion module left support 4 and the expansion module right support 7 are fixed through countersunk screws respectively. When the server is mounted in a 19 inch cabinet, no expansion module is required.

As shown in fig. 4, a storage module 5 and an interaction module 3 are sequentially arranged on the lower side of the power module 2, and the storage module 5 includes a storage tray 5-1, an IO board 5-2, an NVMESSD hard disk 5-3, and a hard disk backplane 5-4.

Two sides of the storage tray 5-1 are fixed with the box body through the pull nails and are used for installing an IO board 5-2 and 8 NVMESSD hard disks 5-3. The hard disk back plate 5-4 is fixed with the storage tray 5-1 by two hand screws, and 8 NVMESSD hard disks 5-3 are inserted into the storage tray 5-1 from the front window of the storage module 5.

When the server is installed in a 21-inch cabinet, the storage module 5 is in the front window position, and two M3 screws are required to fix the IO board 5-2. When the server is installed in a 19-inch cabinet, the storage module 5 is in the back window position, and the IO board 5-2 is not required to be installed. Two sides of the storage module 5 are fixed with the side wall of the case by means of the blind rivet.

As shown in fig. 5, the interaction module 3 is located at the lower side of the storage module 5, the interaction module 3 includes an interaction tray 3-1, a switch board 3-2 and a PCIE card 3-3, and the interaction module 3 is connected to an external computing module. And power-assisted handles are designed on two sides of the interactive tray 3-1, so that the switch board card and the GPU board card are interconnected. The panel of the interaction module 3 supports a 16-port Mini SAS interface and is interconnected with the external computing module.

Two M3 screws are used to fix the exchange board 3-2 on the exchange tray 3-1, and 8 half-high and half-long PCIE cards 3-3 are inserted into the flat slot on the exchange board 3-2. After the interactive module 3 enters from the front window of the case, the high-density ports on the exchange board 3-2 are in good contact with the high-density ports on the GPU board by utilizing the power-assisted handles on the two sides.

As shown in fig. 1, a GPU module 9 is disposed at the lower portion of the chassis, and the GPU module 9 is an NVA100 GPU board and directly fixed to a stud on the chassis base 6. After the GPU module 9 is vertically placed in the chassis, its screw is fixed to the stud on the chassis base 6.

As shown in fig. 1, an air guiding cover 8 is disposed on the upper side of the GPU module 9, and the air guiding cover 8 is used for plugging a gap between the GPU board and the upper cover 1, so that air flow is mainly concentrated in a GPU board radiator area, and heat dissipation efficiency is improved. The GPU module 9 is provided with a heat dissipation module 10 at the rear side thereof, and the heat dissipation module 10 is 9 8056 dual-rotor fans.

As shown in fig. 6, the rear portion of the chassis is provided with 10 placement regions, two rows of the placement regions are provided, and the two placement regions located on one side of the rear portion of the chassis are respectively provided with a crown clamp bracket 11 (or an IO board bracket 12) and a false fan bracket 13. The placement area is normally available for installing fans, and when no fan is installed, a support called a false fan support 13 is placed to shield the wind.

The tail part of the case is divided into 10 placing areas through sheet metal parts, wherein the two placing areas 6-1 and 6-2 on the right side are in modular design.

When the chassis is installed on a 21-inch cabinet, the crown clamp bracket 11 is installed in the placement area 6-1, as shown in fig. 7, the crown clamp bracket 11 includes a crown clamp bracket 11-1 and a power adapter plate 11-2, and the power adapter plate 11-2 is fixed on 4 studs at the bottom of the crown clamp bracket 11-1 by using 4M 3 screws. A false fan bracket 13 is arranged in the placing area 6-2 for blocking an air duct to prevent backflow; the remaining 8 zones were each equipped with 8086 fans.

When the chassis is installed in a 19-inch cabinet, an IO board support 12 is installed in the placement area 6-1, as shown in fig. 8, the IO board support 12 includes an IO board bracket 12-1 and an IO board 5-2, and the IO board 5-2 is fixed to a stud and an i-shaped nail at the bottom of the IO board bracket 12-1 by using two M3 screws; the remaining 9 zones were each equipped with an 8086 fan module.

It will be apparent to those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims and their equivalents, the present invention is also intended to include such modifications and variations.

Claims

1. A server structure supporting 8 GPUs comprises a case, a power supply module, a storage module, an interaction module, a heat dissipation module and a GPU module, and is characterized by further comprising an expansion module; the upper part in the case is provided with a power supply module, two sides of the power supply module are respectively provided with an expansion module left support and an expansion module right support, the lower side of the power supply module is sequentially provided with a storage module and an interaction module, the interaction module is connected with an external computing module, the lower part in the case is provided with a GPU module, the upper side of the GPU module is provided with an air guide cover, and the rear side of the GPU module is provided with a heat dissipation module; the tail part of the case is provided with a placing area.

2. The server architecture that supports an 8GPU as recited in claim 1, wherein the chassis comprises a top cover and a chassis base, the top cover comprising a top cover body, a top cover lock, and a tee nail.

3. The architecture of claim 2 in which the power modules comprise power trays, power shelves, CRPS power supplies, and power boards.

4. The server architecture that supports an 8GPU as recited in claim 3, wherein the interaction module comprises an interaction tray, a switch board and a PCIE card.

5. The server structure supporting 8 GPUs according to claim 4, wherein the storage module comprises a storage tray, an IO board, an NVMESSD hard disk and a hard disk backplane.

6. The server structure supporting 8 GPUs according to claim 5, wherein the number of the placement areas is 10, and the two placement areas located at one side of the tail of the chassis are respectively provided with a crown clamp bracket/IO board bracket and a false fan bracket.

7. The 8 GPU-supported server architecture of claim 6, wherein the crown clip holder comprises a crown clip bracket and a power adapter plate.