Energy Optimal Dynamic Thermal Management for Green Computing
Existing thermal management systems for microprocessors assume that the thermal resistance of the heat-sink is constant and that the objective of the cooling system is simply to avoid thermal emergencies. But in fact the thermal resistance of the usual forcedconvection heat-sink is inversely proportional to the fan speed, and a more rational objective is to minimize the total power consumption of both processor and cooling system. Our new method of dynamic thermal management uses both the fan speed and the voltage/frequency of the microprocessor as control variables. Experiments show that tracking the energy-optimal steady-state temperature can saves up to 17.6% of the overall energy, when compared with a conventional approach that merely avoids overheating.
As the power density of microprocessors increases, more elaborate methods of thermal management are required. These cause from the air conditioners necessary in a large data center, to the cooling fans in a sub-notebook computer. The power consumption of active cooling systems is signiﬁcant, and can usually be adjusted dynamically. For instance, the thermal resistance of a forcedconvection heat-sink is determined by the rotational speed of the fan. A typical cooling fan is driven by a brushless DC motor with a feedback speed controller, so that the fan speed can be controlled by software. A higher speed produces a lower thermal resistance, but uses more power. Modern forced-convection heat-sinks for desktop computers have a thermal resistance from 0.2 to 0.6°C/W and can consume several watts. The number of fans required by multiple-node servers must be more than proportional to the number of nodes, in order to compensate for higher power densities