AI server heat dissipation regulation and control method and system, terminal and storage medium

Abstract

The invention provides an AI server heat dissipation regulation and control method and system, a terminal and a storage medium. The method comprises steps that a first temperature of a GPU and GPU related hardware is monitored through a first BMC disposed on a switch board, and a first target rotating speed is generated according to the first temperature; the first BMC receives a second target rotating speed sent by a second BMC through the I2C bus, wherein the second target rotating speed is generated by the second BMC according to a monitored second temperature of mainboard associated hardware; and the first BMC generates a fan regulation and control signal according to the first target rotating speed and the second target rotating speed, and sends the fan regulation and control signal to a fan controller. According to the method, the AI server with various hardware can be effectively cooled, and the whole machine cooling risk caused by hardware equipment increase is solved in a mode of not needing customer perception.

Description

technical field [0001] The present invention relates to the technical field of servers, in particular to a heat dissipation control method, system, terminal and storage medium for an AI server. Background technique [0002] Deep Learning, Machine Learning and High Performance Computing (HPC) are changing our world. From self-driving cars, optimizing retail logistics to global climate simulations, new challenges are emerging, the solutions of which require enormous computing resources. NVIDIA launched the HGX-2Next product in 2020, capable of handling the most computationally intensive workloads and implementing the "world's largest GPU." it consists of 16 A100GPU and NVIDIA NV Switch TM Acceleration, with unprecedented computing power, bandwidth and memory topology, can train massive models, analyze data sets and solve simulation problems faster and more efficiently. [0003] Relying on the HGX2-Next hardware, server manufacturers can develop AI servers with more powerf...

AI Technical Summary

Problems solved by technology

However, for AI servers with more complex hardware and functions, it seems to be stretched.

The more parts there are, the slower the speed of temperature acquisition will be. For complex devices (such as Raid controllers), more program time will be consumed. If the control is not timely, it may cause the risk of heat dissipation of the whole machine.

Images