CPU Power Management via Weighted Usage and Frequency Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for managing power consumption in multi-core processors, such as CPU hot swapping based on usage rate and real-time processes, lead to frequent CPU turn-ons and turn-offs, increasing power consumption rather than reducing it due to the 'burst' characteristic of real-time processes.
Innovation Solution
A method that involves acquiring usage rates and operating frequencies of CPUs, applying different weights to calculate CPU power consumption sensitive factors, and determining whether to turn off or on CPUs based on these factors, ensuring stable conditions to minimize power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If CPU hot swapping is performed based on usage rate and real-time processes, then power consumption is reduced, but CPU stability deteriorates due to frequent turn-ons and turn-offs
Solution Approach 1:
The patent applies preliminary action by performing N consecutive judgments before actually turning off the CPU. Instead of immediately switching off based on a single usage rate threshold, the system requires the usage rate to remain below the threshold for N continuous monitoring periods. This preliminary sustained verification prevents premature shutdown decisions caused by transient low-usage states, thereby reducing frequent CPU turn-ons and turn-offs while still achieving power savings during genuinely idle periods.
2Loss of energy
If the number of real-time processes is used to determine CPU activation, then power consumption is reduced, but responsiveness to burst processes deteriorates
Solution Approach 1:
The patent applies dynamics by making the CPU activation threshold adaptive rather than fixed. The system dynamically adjusts the usage rate threshold based on system state and process characteristics. During burst process scenarios, the threshold can be temporarily raised to allow faster CPU activation, while during normal operation, a lower threshold maintains power efficiency. This dynamic adjustment enables the system to respond appropriately to both sustained workloads and transient burst processes.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention discloses a method for controlling a central processing unit CPU, where the method includes: acquiring a usage rate and an operating frequency of one CPU of operating CPUs, where the number of the operating CPUs is M; if the operating frequency of the one CPU is smaller than a core off-line frequency, multiplying the usage rate of the one CPU by a first weight to obtain a CPU power consumption sensitive factor corresponding to the one CPU; if the operating frequency of the one CPU is greater than or equal to the core off-line frequency, multiplying the usage rate of the one CPU by a second weight to obtain a CPU power consumption sensitive factor corresponding to the one CPU; executing the foregoing steps N times for each of M operating CPUs, where N is a positive integer greater than or equal to one; acquiring MxN CPU power consumption sensitive factors after the foregoing steps are executed N times and a mean of the MxN CPU power consumption sensitive factors; and turning off at least one operating CPU if the mean is smaller than a CPU turn-off power consumption sensitive factor threshold.