The Power Efficiency of GPUs in Multi Nodes Environment with Molecular Dynamics

Author

Udagawa, Takuro ; Sekijima, Masakazu

Author_Institution

Dept. of Comput. Sci., Tokyo Inst. of Technol., Tokyo, Japan

fYear

2011

fDate

13-16 Sept. 2011

Firstpage

399

Lastpage

405

Abstract

Energy efficient systems are highly demanded as the power consumption in HPC region increase. The use of GPUs has attracted attention as a possible solution to these problems because of their parallel performance and power efficiency. However, it is uncertain how much improvement can be obtained by applying GPUs to applications. In this study, we developed a molecular dynamics simulation program for CPU parallel and GPU parallel, and executed it on our computer cluster. Then we compared the performance of CPUs and GPUs. We obtained the result that the GPUs were about 10 times faster and 5 times more power efficient than the CPUs.

Keywords

computer graphic equipment; coprocessors; energy conservation; molecular dynamics method; parallel processing; power aware computing; CPU parallel; GPU parallel; HPC region; computer cluster; energy efficient systems; molecular dynamics simulation program; multinodes environment; parallel performance; power consumption; power efficiency; Atomic clocks; Atomic measurements; Central Processing Unit; Graphics processing unit; Instruction sets; Message systems; Power demand; GPU; Molecular Dynamics;

fLanguage

English

Publisher

ieee

Conference_Titel

Parallel Processing Workshops (ICPPW), 2011 40th International Conference on

Conference_Location

Taipei City

ISSN

1530-2016

Print_ISBN

978-1-4577-1337-8

Electronic_ISBN

1530-2016

Type

conf

DOI

10.1109/ICPPW.2011.43

Filename

6047055