Evaluation of clusters with AMD Barcelona and Shanghai CPUs and Infiniband DDR and QDR interconnects

Authors

  • M.V. Kudryavtsev
  • D.V. Moshkin
  • M.A. Polunin
  • L.K. Aisymont

Keywords:

clusters
superclusters
processors
performance
gigaflops
APEX-MAP
STREAM
INFINIBAND
BARCELONA
SHANGHAI

Abstract

A methodology for the performance evaluation of clusters with quadsocket and dualsocket Supermicro boards with AMD Barcelona (clock rate 2,0 GHz, peak performance 32 Gflops) and AMD Shanghai (clock rate 2,6 GHz, peak performance 41,6 Gflops) CPUs, Infiniband 4X DDR and 4X QDR interconnects is proposed. APEX-MAP, STREAM, EuroBen, NPB and other benchmarks are used. The results are compared with the clusters and nodes based on Intel Clovertown CPUs. In the nearest future, the methodology is planned to be applied to clusters based on Nehalem CPUs. The paper was prepared on the basis of the author’s report at the International Conference on Parallel Computing Technologies (PaVT-2009; http://agora.guru.ru/pavt).

New computational technologies

Downloads

Published

2009-04-06

Issue

Vol. 10 (2009): Issue 1.

Section

Section 2. Programming

Author Biographies

M.V. Kudryavtsev

Research Center for Electronic Computing
• Head of Sector

D.V. Moshkin

Research Center for Electronic Computing
• Head of Sector

M.A. Polunin

CCJSC «TRONIK»
• Technical Director

L.K. Aisymont

Research Center for Electronic Computing
• Head of Department

References

  1. Слуцкин А., Эйсымонт Л. Российский суперкомпьютер с глобально адресуемой памятью // Открытые системы. 2007. № 9. 42-51.
  2. Фролов А., Семeнов А., Корж А., Эйсымонт Л. Программа создания перспективных суперкомпьютеров // Открытые системы. 2007. № 9. 20-29.
  3. Митрофанов В., Слуцкин А., Эйсымонт Л. Современное состояние и перспективы развития суперкомпьютерных технологий для стратегически важных задач // Материалы конференции «Перспективы развития высокопроизводительных вычислительных архитектур». № 1. М.: ИТМиВТ им. С. А. Лебедева РАН, 2008.
  4. Abts D. et al. The Cray BlackWidow: a highly scalable vector multiprocessor // Proceedings of the 2007 ACM/IEEE Conference on Supercomputing (SC’07). New York, 2007.
  5. Scott S., Abts D., Kim J., Dally W.J. The BlackWidow high-radix clos network // International Symposium on Computer Architecture (ISCA). Boston, MA, USA. 2006. 16-28.
  6. Brightwell R., Pedretti K.T., Underwood K.D., Hudson T. SeaStar Interconnect: Balanced Bandwidth for Scalable Performance // IEEE MICRO. 26, Issue 3. 2006. 41-57.
  7. Konecny P. Introducing the Cray XMT // Proc. Cray User Group Meeting (CUG 2007). Seattle, WA. 2007.
  8. CRAY/MTA Principles of Operation // Cray Inc. 2005.
  9. http://www.cs.virginia.edu/stream/FTP/Code/stream.c
  10. Strohmaier E., Shan H. Apex-Map: A Global Data Access Benchmark to Analyze HPC Systems and Parallel Programming Paradigms // Proceedings of the 2005 ACM/IEEE Conference on Supercomputing (SC’05). Washington, DC, 2005.
  11. http://www-unix.mcs.anl.gov/mpi/mpptest/
  12. Кудрявцев М., Мошкин Д., Полунин М., Эйсымонт Л. Суперкластеры - между прошлым и будущим // Открытые системы. 2008. № 8. 40-47.