Physical-aware predictive dynamic thermal management of multi-core processors

Abstract

The advances in silicon process technology have made it possible to have processors with larger number of cores. The increment of cores number has been hindered by increasing power consumption and heat dissipation due to high power expenditure in a small area die size. The high temperature can cause degradation in performance, reliability, transistor ageing, transition speed and increase in leakage current. In this paper, we present a method which considers different thermal behavior of cores and uses both physical sensors and performance counters simultaneously to improve thermal management of both SMT multi-core processors with a physical sensor per core and Non-SMT multi-core processors with only one physical sensor for the processor. The experimental results indicate that our technique can significantly decrease the average and peak temperature in most cases compared to Linux standard scheduler, and two well-known thermal management techniques: PDTM, and TAS.