此页面上的内容需要较新版本的 Adobe Flash Player。

获取 Adobe Flash Player

Power, Performance and Energy Efficiency Evaluation of GPU Overclocking When Executing CUDA applications

 

Jun-hee LEE, Joon-ho KONG, Sung-woo CHUNG

 

(Division of Computer and Communication Engineering, Korea University, Seoul 136-701, Korea)

 

Abstract-Processor performance has continuously advanced since Moore's law. However, power constraints have become one of the largest concerns for processor architects to further improve processor performance. With the recent introduction of GPGPU (general purpose graphic processing unit) technology, which uses GPU for general computing, techniques to maximize GPU processor parallelism and to minimize power consumption have become popular. However, end users requiring better performance often use overclocking to improve CPU and GPU performance. This paper aims to evaluate the effect of GPU and CPU overclocking when executing CUDA applications from three perspectives: performance, power consumption, and energy efficiency. The experiment shows that when executing matrix multiplication program with CUDA, while execution time of GPU (GPU core: 625 MHz, GPU memory: 450 MHz) overclocking was similar to that of CPU overclocking, its energy consumption was 12.8% less. In terms of energy efficiency, higher GPU overclocking leads to better efficiency. We verify the positive results of GPU overclocking in execution time and energy efficiency when executing CUDA applications and the potential of GPU overclocking.

 

Key words-overclocking; power efficiency; GPGPU computing

 

Manuscript Number: 1674-8042(2011)supp1.-0046-03

 

doi: 10.3969/j.issn.1674-8042.2011.supp1.010

 

References

 

[1]Borkar S. Low power design challenges for the decade. Proceedings of the 2001 Conference on Asia South Pacific Design Automation, IEEE Press, Januaury 2001.
[2]NVIDIA Corporation. Cuda project http://www. nvidia.com/cuda.
[3]Advanced Micro Devices Inc. ATI stream technology. http://www.amd.com/stream.
[4]Lee J  H, Kong J,  Suh T, et al. An effective CPU overclocking scheme considering energy efficiency. Journal of KSCI, 2009, 14(12): 17-24.
 

[full text view]