Exploring GPGPUs Workload Characteristics and Power Consumption
Department of Embedded Systems Architecture, Technical University of Berlin, Einsteinufer 17, D-10587 Berlin, Germany
Technical University of Berlin, 2013
@article{lal2013exploring,
title={Exploring GPGPUs Workload Characteristics and Power Consumption},
author={Lal, Sohan and Lucas, Jan and Alvarez-Mesa, Mauricio and Elhossini, Ahmed and Juurlink, Ben},
year={2013}
}
While general purpose computing on GPUs continues to enjoy higher computing performance with every new generation. The high power consumption of GPUs is an increasingly important concern. To create power-efficient GPUs, it is important to thoroughly study its power consumption. The power consumption of GPUs varies significantly with workloads. Therefore, in this work we study GPU power consumption at a detailed level and its correlation with well-known workload characteristics such as IPC. The low IPC kernels are further explored for the possible bottlenecks.
June 22, 2013 by hgpu
