Performance Analysis of an Embedded Chipset on a Multi-screen based Automotive Applications Environment

Abstract

There is a growing need for a System-on-Chip (SoC) targeting in-vehicle infotainment (IVI) and advanced driver assistance system (ADAS) products. SoC is superior to multi-chip systems in terms of space efficiency and has the advantage of minimizing power consumption. However, since different kinds of operations are done on a single chip, individual devices may do not achieve expected performance. Therefore, the purpose of this study is to examine the performance issue of an embedded chipset in the multi-screen-based automotive scenario. We observe two performance anomalies of shared resources and quantitatively analyze them using synthetic workload. From experimental results, we derive that performance can be maximized by only co-balancing CPU and GPU workload of given task sets' resource demand.