FPGA-Based AI Server Heat Dissipation Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional server control systems, relying on Baseboard Management Controllers (BMC), are unable to respond timely to the fast-changing power consumption of Graphics Processing Units (GPU) in AI servers, leading to potential mainboard damage due to inadequate heat dissipation.
Innovation Solution
A safety control method and system utilizing a Field Programmable Gate Array (FPGA) to independently sample and monitor the GPU's electrical current and power, taking control of the heat dissipation system when certain conditions are met, ensuring timely and effective cooling measures are taken.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the BMC is used to control heat dissipation by periodically scanning mainboard information, then the system structure is simple and easy to operate, but the response speed is slow due to sampling frequency limitations, causing delayed heat dissipation adjustment
Solution Approach 1:
The patent segments the heat dissipation control function from the BMC and implements it in the FPGA. The FPGA independently monitors GPU power consumption and controls the heat dissipation system, while the BMC retains its original management functions. This segmentation allows the heat dissipation control to operate at a higher sampling frequency without increasing overall system complexity.
Solution Approach 2:
The patent introduces an intermediary mechanism where the FPGA acts as a mediator between the GPU and the heat dissipation system. The FPGA receives GPU power consumption information, makes real-time control decisions, and adjusts the heat dissipation system accordingly, bridging the gap between the GPU's fast-changing power consumption and the heat dissipation system's response.
2Reliability
If the sampling frequency is increased to respond to fast-changing GPU power consumption, then the response timeliness is improved, but the system complexity and resource consumption of the BMC increase
Solution Approach 1:
The patent segments the high-frequency sampling function from the BMC and assigns it to the FPGA. This allows the system to achieve millisecond-level or even microsecond-level sampling frequency for GPU power consumption monitoring without increasing the BMC's workload or system complexity. The FPGA's segmentation of this function enables high-reliability heat dissipation control.
Solution Approach 2:
The FPGA performs self-service by independently monitoring GPU power consumption and controlling the heat dissipation system without requiring BMC intervention. This self-service mechanism ensures reliable heat dissipation control at high sampling frequencies while keeping the BMC's resource consumption low.
3Measurement precision
If the FPGA takes over heat dissipation control from the BMC, then the response speed and control precision are improved, but the control system becomes more complex
Solution Approach 1:
The patent segments the monitoring and control functions into distinct modules within the FPGA: a monitoring module that reads GPU power consumption data, a judgment module that compares actual values with preset thresholds, and a control module that adjusts the heat dissipation system. This segmentation achieves precise power consumption monitoring and control while maintaining clear system architecture.
Solution Approach 2:
The FPGA performs multiple functions: it monitors GPU power consumption, controls the heat dissipation system, and can communicate with the BMC when needed. This multi-functionality allows the FPGA to achieve precise control without requiring separate dedicated components for each function, thereby limiting the increase in system complexity.
Data Source
AI summary
Provided is a safety control method for an AI server, which is applied in an FPGA. The method includes: obtaining a current electrical current and a current power of a GPU in the AI server according to a preset frequency; determining whether the GPU satisfies a first control privilege transfer requirement; when the GPU satisfies the first control privilege transfer requirement, taking over control privilege of a heat dissipation system from a BMC; and controlling the heat dissipation system according to the current electrical current and the current power of the GPU; wherein the first control privilege transfer requirement includes: the current electrical current of the GPU exceeds a preset electrical current, or a rate of change of the current electrical current of the GPU exceeds a preset rate of change of electrical current, or the current power of the GPU exceeds a first preset power.