CN221008225U

CN221008225U - GPU heat abstractor and general server based on general server

Info

Publication number: CN221008225U
Application number: CN202323109056.4U
Authority: CN
Inventors: 王珉
Original assignee: Inspur Computer Technology Co Ltd
Current assignee: Inspur Computer Technology Co Ltd
Priority date: 2023-11-17
Filing date: 2023-11-17
Publication date: 2024-05-24
Anticipated expiration: 2033-11-17

Abstract

The utility model provides a general server-based GPU heat dissipation device and a general server, wherein the heat dissipation device comprises: the system comprises a main board, a GPU, an expansion module and a heat dissipation module; the CPU on the main board is in communication connection with the GPU through the expansion module; the baseboard management controller on the main board is connected with the input end of the expansion module; the first path of the output end of the expansion module is connected to the GPU, the second path of the output end of the expansion module is connected to the heat dissipation module, and the third path of the output end of the expansion module is connected to the temperature acquisition module. The heat dissipation module is arranged in the middle of the chassis between the CPU and the GPU; the expansion module adopts an I2CSWITCH chip; the CPU is in communication connection with the GPU through PCIE signals. The utility model further provides a general server, which comprises the GPU heat dissipation device based on the general server. The universal server main board uses PCIe expansion card small-size fans capable of supporting fan regulation and control, and achieves heat dissipation of the high-power consumption GPU of the universal server.

Description

GPU heat abstractor and general server based on general server

Technical Field

The utility model belongs to the technical field of server heat dissipation, and particularly relates to a general server-based GPU heat dissipation device and a general server.

Background

In current server systems, multiple high-performance GPUs are typically used to meet customer requirements in the image video computing and deep learning fields. For the server system, the high-performance GPU has high power consumption, and most of the GPUs are passively radiating, so that the whole machine needs to meet the radiating requirement of the GPU during design. GPU: (Graphics Processing Unit) a graphics processor.

In order to realize that the server system supports a plurality of high-performance GPUs, a common practice is to adopt a 4U-height chassis, and the heat dissipation of the GPUs is met by optimizing an air duct, improving an air guide cover, improving the fan configuration of the system and the like. However, in the design of the general server system, the 2U-height chassis is adopted in the industry, and because of the limited space and compact component layout, the heat dissipation of the high-power consumption GPU is met by optimizing the air duct, improving the air guide cover, improving the system fan configuration and the like, and the difficulty and the cost are high.

Disclosure of Invention

In order to solve the technical problems, the utility model provides a general-purpose server-based GPU heat dissipation device and a general-purpose server, which are used for realizing heat dissipation of a general-purpose server high-power consumption GPU.

In order to achieve the above purpose, the present utility model adopts the following technical scheme:

the GPU heat dissipation device based on the universal server comprises a main board, a GPU, an expansion module and a heat dissipation module;

The CPU on the main board is in communication connection with the GPU through the expansion module; the baseboard management controller on the main board is connected with the input end of the expansion module; the first path of the output end of the expansion module is connected to the GPU, the second path of the output end of the expansion module is connected to the heat dissipation module, and the third path of the output end of the expansion module is connected to the temperature acquisition module.

Further, the second path of the output end of the expansion module is connected to the heat dissipation module, which specifically includes: the second path of the output end of the expansion module is connected to the signal input end of the programmable logic chip, and the output end of the programmable logic chip is connected to the heat dissipation module through the fan connector.

Further, the heat dissipation module is installed in the middle of the chassis between the CPU and the GPU.

Further, the heat dissipation module adopts a fan; the model of the fan is 4028.

Further, the expansion module adopts an I2C SWITCH chip.

Further, the CPU on the motherboard is connected to the GPU through PCIE signals in a communication manner.

Further, the baseboard management controller is in communication connection with the I2C SWITCH chip through an I2C signal.

Further, the first path of the output end of the expansion module is connected to the GPU, which specifically includes: the first path of the output end of the I2CSWITCH chip is connected with the GPU through I2C signals in a communication mode.

Further, the output end of the programmable logic chip is connected to the heat dissipation module through a fan connector, which specifically includes: the programmable logic chip is connected with the fan in-place signal, the fan pulse signal and the fan rotating speed signal in a communication manner.

The utility model also provides a general server, which comprises the GPU heat dissipation device based on the general server.

The effects provided in the summary of the invention are merely effects of embodiments, not all effects of the invention, and one of the above technical solutions has the following advantages or beneficial effects:

The utility model provides a general server-based GPU heat dissipation device and a general server, wherein the heat dissipation device comprises: the system comprises a main board, a GPU, an expansion module and a heat dissipation module; the CPU on the main board is in communication connection with the GPU through the expansion module; the baseboard management controller on the main board is connected with the input end of the expansion module; the first path of the output end of the expansion module is connected to the GPU, the second path of the output end of the expansion module is connected to the heat dissipation module, and the third path of the output end of the expansion module is connected to the temperature acquisition module. The heat dissipation module is arranged in the middle of the chassis between the CPU and the GPU; the expansion module adopts an I2CSWITCH chip; the CPU on the main board is in communication connection with the GPU through PCIE signals; the baseboard management controller is in communication connection with the I2C SWITCH chip through an I2C signal; the first path of the output end of the I2C SWITCH chip is connected with the GPU through I2C signals in a communication mode. The utility model further provides a general server, which comprises the GPU heat dissipation device based on the general server. The universal server main board uses PCIe expansion card small-size fans capable of supporting fan regulation and control, and achieves heat dissipation of the high-power consumption GPU of the universal server.

In the utility model, the PCIe expansion card can be matched with other PCIe devices besides the GPU to be used, and the quick switching multiplexing of the general configuration and the GPU configuration can be realized.

Drawings

Fig. 1 is a schematic connection diagram of a GPU heat sink device based on a general server according to embodiment 1 of the present utility model;

Fig. 2 is a schematic diagram of a fan according to embodiment 1 of the present utility model.

Detailed Description

In order to clearly illustrate the technical features of the present solution, the present utility model will be described in detail below with reference to the following detailed description and the accompanying drawings. The following disclosure provides many different embodiments, or examples, for implementing different structures of the utility model. In order to simplify the present disclosure, components and arrangements of specific examples are described below. Furthermore, the present utility model may repeat reference numerals and/or letters in the various examples. This repetition is for the purpose of simplicity and clarity and does not in itself dictate a relationship between the various embodiments and/or configurations discussed. It should be noted that the components illustrated in the figures are not necessarily drawn to scale. Descriptions of well-known components and processing techniques and processes are omitted so as to not unnecessarily obscure the present utility model.

Example 1

The embodiment 1 of the utility model provides a general server-based GPU heat dissipation device, which is used for solving the problem of heat dissipation of a high-power consumption GPU in a general server system.

Fig. 1 is a schematic connection diagram of a GPU heat sink device based on a general server according to embodiment 1 of the present utility model; the device comprises: the system comprises a main board, a GPU, an expansion module and a heat dissipation module;

The expansion module adopts an I2C SWITCH chip.

And a CPU on the main board is in communication connection with the GPU through PCIE signals.

The baseboard management controller is in communication connection with the I2C SWITCH chip through I2C signals.

The first path of the output end of the expansion module is connected to the GPU, and specifically comprises: the first path of the output end of the I2C SWITCH chip is connected with the GPU through I2C signals in a communication mode.

In fig. 1, a PCIe expansion card is used on the motherboard to connect the fan and GPU. The PCIe expansion card acquires PCIe signals from the CPU of the main board and sends the PCIe signals to the GPU, so that high-speed communication between the CPU and the GPU is realized. The PCIe expansion card and the main board BMC are communicated through I2C, and the I2C_BMC on the expansion card is divided into three paths through SWITCH: i2c_0 is connected to GPU 0; i2c_1 is connected with the CPLD; i2c_3 is connected to the temperature sensor. The fan and the CPLD realize related signal interaction through a fan connector on the PCIe expansion card. Therefore, the function of cooling the GPU card by the fan is realized on the hardware design.

The heat dissipation module is arranged in the middle of the chassis between the CPU and the GPU. The heat dissipation module adopts a fan; the fan model number is 4028. Fig. 2 is a schematic diagram of a fan according to embodiment 1 of the present utility model, where a heat dissipation module is a fan; the model of the fan is 4028, and the small-size fan is arranged in the middle of the chassis between the CPU and the GPU.

The working process of the embodiment 1 of the utility model is as follows: the CPLD firstly judges whether the Fan is in place according to the state of the fan_present signal level, and writes the state value into the fan_present register, wherein a low level 0 represents that the Fan is in place, and a high level 1 represents that the Fan is out of place.

If high, the CPLD will first default to writing 100 into the internally set fan_duty register, thereby setting the fan_PWM signal to full rotation of the Fan. The CPLD can also know the current rotation speed value according to the fan_ Tach signal fed back by the Fan and write the value into the fan_speed register.

The BMC reads the fan_present register in the CPLD through the I2C signal to determine whether the fan is in place. If the fan is in place, the temperature parameters of the current GPU and the Temp Sensor on the PCIe expansion card are obtained through the I2C, and the heat dissipation requirement is combined. Thereby setting the fan_PWM signal to the Fan to realize the regulation of the rotation speed of the Fan.

It should be noted that: the heat dissipation engineer may provide the fan_duty value at the corresponding temperature in advance to ensure that the GPU maintains within a stable operating temperature range, and output the corresponding fan_duty value, i.e. 0-100, to the CPLD through the I2C.

According to the GPU heat dissipation device based on the universal server, provided by the embodiment 1 of the utility model, the universal server main board is clamped with the small-size fan by using the PCIe expansion capable of supporting the regulation and control of the fan, so that the heat dissipation of the high-power consumption GPU of the universal server is realized.

According to the GPU heat dissipation device based on the general server, provided by the embodiment 1 of the utility model, the PCIe expansion card can be matched with other PCIe equipment for use besides the GPU, and the quick switching multiplexing of general configuration and GPU configuration can be realized.

Example 2

The embodiment 2 of the utility model also provides a general server, which comprises the GPU heat dissipating device based on the general server provided by the embodiment 1 of the utility model, and the device comprises: the system comprises a main board, a GPU, an expansion module and a heat dissipation module;

The expansion module adopts an I2C SWITCH chip.

The operation of the GPU heat sink based on the general server may refer to the operation in embodiment 1 of the present utility model.

According to the general server provided by the embodiment 2 of the utility model, the PCIe expansion card which can support the regulation and control of the fan is used for connecting the small-size fan by the general server main board, so that the heat dissipation of the high-power consumption GPU of the general server is realized.

According to the general server provided by the embodiment 2 of the utility model, the PCIe expansion card can be matched with other PCIe devices besides the GPU, and the quick switching multiplexing of general configuration and GPU configuration can be realized.

The description of the relevant parts in the general server provided in embodiment 2 of the present application can be referred to the detailed description of the corresponding parts in the GPU heat dissipating device based on the general server provided in embodiment 1 of the present application, and will not be repeated here.

It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements is inherent to. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. In addition, the parts of the above technical solutions provided in the embodiments of the present application, which are consistent with the implementation principles of the corresponding technical solutions in the prior art, are not described in detail, so that redundant descriptions are avoided.

While the specific embodiments of the present utility model have been described above with reference to the drawings, the scope of the present utility model is not limited thereto. Other modifications and variations to the present utility model will be apparent to those of skill in the art upon review of the foregoing description. It is not necessary here nor is it exhaustive of all embodiments. On the basis of the technical scheme of the utility model, various modifications or variations which can be made by the person skilled in the art without the need of creative efforts are still within the protection scope of the utility model.

Claims

1. The GPU heat dissipation device based on the universal server is characterized by comprising a main board, a GPU, an expansion module and a heat dissipation module;

2. The GPU heat sink of claim 1, wherein the second path of the output of the expansion module is connected to the heat sink module, specifically comprising: the second path of the output end of the expansion module is connected to the signal input end of the programmable logic chip, and the output end of the programmable logic chip is connected to the heat dissipation module through the fan connector.

3. The general-purpose server-based GPU heat sink of claim 1, wherein the heat sink module is mounted in a middle portion of the chassis between the CPU and the GPU.

4. The general-purpose server-based GPU heat sink of claim 1, wherein the heat dissipation module employs a fan; the model of the fan is 4028.

5. The GPU heat sink of claim 2, wherein the expansion module is an I2C SWITCH chip.

6. The general-purpose server-based GPU heat sink of claim 5, wherein the CPU on the motherboard is communicatively coupled to the GPU via PCIE signals.

7. The general purpose server-based GPU heat sink of claim 5, wherein the baseboard management controller is communicatively coupled to the I2C SWITCH chip via I2C signals.

8. The GPU heat sink of claim 5, wherein the first path of the output of the expansion module is connected to the GPU specifically comprises: and the first path of the output end of the I2C SWITCH chip is connected with the GPU through I2C signals in a communication way.

9. The GPU heat sink of claim 5, wherein the output of the programmable logic chip is connected to the heat sink module via a fan connector, specifically comprising: the programmable logic chip is connected with the fan in-place signal, the fan pulse signal and the fan rotating speed signal in a communication manner.

10. A general-purpose server comprising the general-purpose server-based GPU heat sink of any of claims 1 to 9.