Hotspot Avoidance Through Runtime Reconfiguration in Network-On-Chip Designs

As technology scales, thermal issues are becoming a significant factor in chip design. Ever increasing transistor densities and increasing leakage currents result in high power dissipation, and require increasingly large and expensive cooling systems to maintain chip operating temperatures within safe bounds. Many existing thermal management techniques focus on reducing the overall power consumption of the chip by throttling performance, eventually resulting in an overall reduction in chip temperature. These techniques, while effective, often do not address location-specific temperature problems referred to as hotspots. Recent research into hotspots has shown that different functional units in general purpose processors can have significantly different temperature profiles, and that moving workloads between units can reduce the creation of hotspots on the die. In this paper, we demonstrate that hotspots also exist in more uniform reconfigurable architectures such as Network on chip (NoC) Designs. We propose the use of dynamic runtime reconfiguration to shift the hotspot-inducing computation periodically and make the thermal profile more uniform. Different approaches to reconfiguration are proposed and evaluated for their effectiveness using a target NoC designed for wireless communication. Our analysis shows that dynamic reconfiguration is an effective technique in reducing hotspots for NoCs.