Heterogeneous computing exploits several disparate technologies within a single system. The different components of a heterogeneous system are often placed within separate temperature zones. Selecting an appropriate operating temperature strongly affects the dissipated power, cooling power (heat load), system performance, and ambient temperature. To this date, no multitemperature design methodology exists. To overcome this limitation, a framework for thermal optimization of heterogeneous computing systems is presented in this article. The effects of operating temperature on delay and power consumption are characterized based on a graph representation of the system. In addition, thermal interactions among the components within a system are considered to accurately evaluate the total power consumption and heat load. In a practical case study, the target temperature of each component within a quantum computing system is determined to minimize the total power under target performance constraints.
Read full abstract