Neural-Network Thermal Control for Multi-Parameter CPU Cooling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional thermal control methods in computing systems lack real-time accuracy and efficiency in managing temperature-related parameters, failing to consider multiple correlated factors such as power consumption and noise, leading to suboptimal performance and instability when CPUs operate at high clock rates.
Innovation Solution
A thermal control system utilizing a neural network algorithm framework, specifically a cerebellar model articulation controller (CMAC), to compute and adjust multiple correlated input parameters, including thermal, power consumption, and noise, to achieve precise control of target parameters like fan speed, LED color, and CPU/GPU underclocking, ensuring convergence to predetermined values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional table lookup method is used to control fan speed based on temperature, then the control system is simple to implement, but the control precision is insufficient and cannot achieve fine control for each temperature
Solution Approach 1:
The patent transitions from discrete temperature ranges in table lookup to continuous temperature parameter control through PID algorithm, enabling fine-grained adjustment of fan speed based on precise temperature measurements and dynamic error correction
Solution Approach 2:
The patent replaces the simple table lookup mechanical control method with an intelligent PID control algorithm that dynamically calculates optimal fan speed based on real-time temperature feedback, achieving higher precision without proportionally increasing hardware complexity
2Speed
If fixed temperature ranges are used to set fan speed, then the control logic is simple, but the system cannot respond in real-time and has switching losses
Solution Approach 1:
The patent implements continuous temperature monitoring and continuous fan speed adjustment through PID control, eliminating the discontinuous switching between fixed temperature ranges and maintaining optimal cooling performance without energy-wasting transitions
Solution Approach 2:
The patent employs feedback control where the actual temperature is continuously measured and compared with the target temperature, and the PID algorithm dynamically adjusts fan speed based on the error signal, enabling real-time response and eliminating switching losses
3Power
If air cooling with room temperature air is used, then the system is simple to implement, but the heat dissipation efficiency is limited when CPU operates at high clock rate
Solution Approach 1:
The system uses ambient air as the cooling medium, allowing the environment to provide cooling service free of charge, while the PID-controlled fan optimizes the utilization of this free cooling resource to achieve high heat dissipation efficiency without additional hardware complexity
Data Source
AI summary
In a thermal control system and a thermal control method thereof, the thermal control system includes a processing circuit, a control circuit, and a detection circuit. The processing circuit receives input parameters related to temperature and outputs a first signal. The input parameters are correlated with each other. The control circuit couples the processing circuit and receives the first signal. The control circuit performs a computation of a neural network algorithm based on the first signal and outputs a control signal to simultaneously adjust a plurality of target parameters based on the computation result of the neural network algorithm. The detection circuit couples the control circuit and the processing circuit, receives the updated target parameters, determines whether the updated target parameters converge to the corresponding pre-determined values, and outputs a feedback signal to the processing circuit.