Skip to main content
eScholarship
Open Access Publications from the University of California

System, Component and Subcomponent Power Estimation

  • Author(s): Hovhannisyan, Davit
  • Advisor(s): Kurdahi, Fadi
  • et al.
Abstract

This work focused on the power estimation of plug load devices, and in particular on Personal Computers. As a result, neural network classification estimated power with less than 5.4% errors. Study showed that internal performance counters properly described the overall system and the main component (CPU and GPU) power. Furthermore, neural networks model demonstrated higher precision on test data than the linear regression model.

Main Content
Current View