Evaluating System Identification Methods for Predicting Thermal Dissipation of Heterogeneous SoCs

TL;DR

This paper compares three system identification methods for predicting the thermal behavior of heterogeneous SoC platforms, demonstrating that polynomial regressors outperform neural network models in accuracy.

Contribution

It evaluates and compares the effectiveness of linear and neural network-based system identification methods for thermal prediction of heterogeneous SoCs.

Findings

Polynomial regressor model outperforms neural network models in accuracy.

Shorter training data (1 hour) yields better predictions.

Model can predict temperature based on clock frequency and core utilization.

Abstract

In this paper we evaluate the use of system identification methods to build a thermal prediction model of heterogeneous SoC platforms that can be used to quickly predict the temperature of different configurations without the need of hardware. Specifically, we focus on modeling approaches that can predict the temperature based on the clock frequency and the utilization percentage of each core. We investigate three methods with respect to their prediction accuracy: a linear state-space identification approach using polynomial regressors, a NARX neural network approach and a recurrent neural network approach configured in an FIR model structure. We evaluate the methods on an Odroid-XU4 board featuring an Exynos 5422 SoC. The results show that the model based on polynomial regressors significantly outperformed the other two models when trained with 1 hour and 6 hours of data.