An energy-efficient parallel-processing method based on master-hibernating DVFS

A parallel-processing method, named “critical data-aware operation” (CDO), for higher energy efficiency (i.e., lower energy consumption in computer systems) is proposed. The proposed method achieves energy-efficient computation by aggressive voltage scaling with critical data-protection techniques. As a case study to evaluate the method, K-means clustering simulation was performed, and the simulation results show that the method achieves 23% better energy efficiency (with negligible accuracy degradation) than the conventional method.