Hiroshi Sasaki

Graphics processing units (GPUs) provide signifi-cant improvements in performance and performance-per-watt as compared to traditional multicore CPUs. This energy-efficiency of GPUs has facilitated the use of GPUs in many application domains. Albeit energy effi-cient, GPUs consume non-trivial power independently of CPUs. Therefore, we need to analyze the power and per-formance characteristic of GPUs and their causal relation with CPUs in order to reduce the total energy consump-tion of the system while sustaining high performance. In this paper, we provide a power and performance anal-ysis of GPU-accelerated systems for better understand-ings of these implications. Our analysis on a real system discloses that system energy can be reduced by 28 % re-taining a decrease in performance within 1 % by control-ling the voltage and frequency levels of GPUs. We show that energy savings can be achieved when GPU core and memory clock frequencies are appropriately scaled con-sidering the workload characteristics. Another interest-ing finding is that voltage and frequency scaling of CPUs is trivial for total system energy reduction, and even should not be applied in state-of-the-art GPU-accelerated systems. We believe that these findings are useful to de-velop dynamic voltage and frequency scaling (DVFS) al-gorithms for GPU-accelerated systems.