Linpack evaluation on a supercomputer with heterogeneous accelerators

Author

Endo, Toshio ; Nukada, Akira ; Matsuoka, Satoshi ; Maruyama, Naoya

Author_Institution

Grad. Sch. of Inf. Sci. & Eng., Tokyo Inst. of Technol., Tokyo, Japan

fYear

2010

fDate

19-23 April 2010

Firstpage

1

Lastpage

8

Abstract

We report Linpack benchmark results on the TSUBAME supercomputer, a large scale heterogeneous system equipped with NVIDIA Tesla GPUs and ClearSpeed SIMD accelerators. With all of 10,480 Opteron cores, 640 Xeon cores, 648 ClearSpeed accelerators and 624 NVIDIA Tesla GPUs, we have achieved 87.01TFlops, which is the third record as a heterogeneous system in the world. This paper describes careful tuning and load balancing method required to achieve this performance. On the other hand, since the peak speed is 163 TFlops, the efficiency is 53%, which is lower than other systems. This paper also analyses this gap from the aspect of system architecture.

Keywords

benchmark testing; coprocessors; parallel machines; performance evaluation; resource allocation; ClearSpeed SIMD accelerators; ClearSpeed accelerators; Linpack benchmark; NVIDIA Tesla GPU; Opteron cores; TSUBAME supercomputer; Xeon cores; heterogeneous accelerators; load balancing; supercomputer evaluation; system architecture; Acceleration; Computer architecture; Energy consumption; High performance computing; Informatics; Information science; Large-scale systems; Load management; Supercomputers; Switches;

fLanguage

English

Publisher

ieee

Conference_Titel

Parallel & Distributed Processing (IPDPS), 2010 IEEE International Symposium on

Conference_Location

Atlanta, GA

ISSN

1530-2075

Print_ISBN

978-1-4244-6442-5

Type

conf

DOI

10.1109/IPDPS.2010.5470353

Filename

5470353