Dynamic Voltage Frequency Scaling User Driven Process Profiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Dynamic Voltage and Frequency Scaling (DVFS) techniques in high-performance processors are pessimistic, assuming worst-case manufacturing process variations and operating temperatures, and ignore user preferences, leading to inefficient power management and performance optimization.
Innovation Solution
The implementation of User-Driven Frequency Scaling (UDFS) and Process-Driven Voltage Scaling (PDVS) systems, which determine operating frequencies and voltages based on direct user input and individual processor profiles, respectively, to dynamically adapt power management according to user needs and processor-specific conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional DVFS techniques use worst-case assumptions for manufacturing process variation and temperature, then reliability is improved, but power consumption increases
Solution Approach 1:
The patent changes the parameters used for DVFS decision-making from static worst-case assumptions to dynamic measurements of actual process variation and temperature. By measuring the actual operating conditions and adjusting voltage/frequency settings accordingly, the system achieves both high reliability and reduced power consumption, eliminating the need for conservative over-provisioning
Solution Approach 2:
The patent implements self-service by having the processor automatically measure its own process variation and temperature characteristics, then use this information to optimize its own operating parameters. This eliminates the need for external calibration or conservative manufacturer-specified settings, allowing the processor to serve itself with optimal power-efficient settings while maintaining reliability
2Productivity
If traditional DVFS techniques base frequency selection on CPU utilization, then productivity is improved, but user satisfaction deteriorates
Solution Approach 1:
The patent introduces feedback by directly querying the user about their performance expectations and satisfaction levels. This feedback loop allows the system to adjust frequency settings based on actual user needs rather than relying solely on CPU utilization metrics, thereby maintaining productivity while improving user satisfaction through personalized performance optimization
3Reliability
If DVFS techniques use loose worst-case bounds for voltage setting, then reliability is improved, but power consumption increases
Solution Approach 1:
The patent changes the voltage setting approach from using fixed loose worst-case bounds to using dynamically measured process variation parameters. By determining the actual process variation of the specific processor instance and calculating the minimum required voltage accordingly, the system eliminates energy loss from excessive voltage headroom while ensuring reliable operation through measurement-based parameter optimization
Data Source
AI summary
Certain embodiments of the present invention provide a method for power management including determining at least one of an operating frequency and an operating voltage for a processor and configuring the processor based on the determined at least one of the operating frequency and the operating voltage. The operating frequency is determined based at least in part on direct user input. The operating voltage is determined based at least in part on an individual profile for processor.