Low-power Task Scheduling for GPU Energy Reduction

Li Tang, Yiji Zhang
Department of Computer Science and Engineering, University of Notre Dame
University of Notre Dame, 2011


   title={Low-power Task Scheduling for GPU Energy Reduction},

   author={Tang, L. and Zhang, Y.},



Download Download (PDF)   View View   Source Source   



Graphics processing units (GPU) have been intensively used by high-performance computing applications. However, GPU’s large power consumption is a big issue coexisting with the high parallelism. Although Dynamic Voltage and Frequency Scaling (DVFS) [1] has been heavily studied and successfully applied to real products for saving CPU power consumption, DVFS is still relatively new for GPU energy studies. The lack of DVFS and other power management schemes on the GPU also makes its large power consumption significant in recent computer systems. In this Operating System course project, we propose a low-power scheme for GPU energy reduction. This project can be decomposed to GPU DVFS implementation and GPU linear regression power model. For enabling the DVFS on GPU, the settings of voltage and frequency in open-source Nouveau GPU driver [7] has been studied. For supporting the DVFS choosing voltage and frequency levels, a GPU linear regression power model has been built and evaluated.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2021 hgpu.org

All rights belong to the respective authors

Contact us: