I am working on a thermal control system which varies core clock speed of a GPU/CPU to maintain a specific temperature. I want to include load into the calculation but to do so I need to know the normal temperature response (partial derivative) to load when all other variables (core speed, fan speed, etc.) are held constant. I intend to factor load into equations which increase clock speed, but not into equations that reduce clock speed. The purpose of this is that if load is <100% I don't want the system to crank up the core speed to maximum only to burn up the chip if the load spikes to 100%. I have considered using a straight line function from idle temperature to target temperature but it may not be the ideal solution.
T_target_applied = Target temperature to plug into existing optimization algorithm
T_idle = Idle temp
G_load = GPU Load (0 to 1)
T_target = Actual target temperature
T_target_applied = T_idle + (G_load * (T_target - T_idle))
To find an ideal solution I need to see the normal temperature response of a CPU/GPU to load.
T_target_applied = Target temperature to plug into existing optimization algorithm
T_idle = Idle temp
G_load = GPU Load (0 to 1)
T_target = Actual target temperature
T_target_applied = T_idle + (G_load * (T_target - T_idle))
To find an ideal solution I need to see the normal temperature response of a CPU/GPU to load.