Thermal modelling of modern processors using FEM and compact model

A variety of thermal models has been proposed to predict the temperatures inside modern processors. In this paper, we describe and compare two such approaches, a detailed FEMbased simulation and a simpler architectural compact model. It is shown that both models provide comparable results when it comes to predicting the maximal temperature, however there are also non-negligible differences when estimating thermal gradients within a chip. Furthermore, transient simulation results show some differences in temperature profile during processor heating