Supercomputer from GPUs
CSIRO’s graphics processing unit (GPU) cluster is now Australia’s ‘greenest’ supercomputer ranking 11th on an internationally recognised list of the world’s 500 fastest and most energy efficient supercomputers – the Green500 List.
Announced today at the international SC10 supercomputing conference in New Orleans, the Green500 List is a ranking of the TOP500 supercomputers by energy efficiency (performance speed per Watt of energy consumed).
It highlights the growing power consumption of the world’s fastest computers and encourages owners to reduce their carbon footprint by using technology that improves energy efficiency.
CSIRO’s GPU cluster ranked number 145 on the Top500 list earlier this week, performing the Linpack benchmark at 52.55 Teraflops in double precision, with an energy efficiency of 555.5 MegaFlops/Watt.
“We knew we had a fantastic computational facility in our GPU cluster, but we are particularly delighted to near the top of the Green500 List,” said Group Executive Information Sciences CSIRO, Dr Alex Zelinsky.
“GPU computing really fits well with our e-Research strategy and has proven to be a great success for CSIRO over the past year. This new high performance computing technology has accelerated our scientific research and put us on the world stage alongside China, Europe and the United States.”
CSIRO’s GPU cluster, recently upgraded by Xenon Systems, combines Intel Central Processing Units (CPUs) with 64 NVIDIA Tesla S2050 graphics processing units (GPUs) – an industrial version of the graphics cards found in game consoles and computers.
GPU-based supercomputers are twice as energy efficient as regular CPU supercomputers, completing calculations around 10-100X faster than CPUs. GPUs are also much cheaper to purchase and occupy half the rack space which reduces cooling and data centre costs.
CSIRO’s Computational and Simulation Science leader, Dr John Taylor, said energy efficiency and value for money were big draw cards for CSIRO.
“We were able to build a faster, greener supercomputer at a fraction of the cost using GPUs.
“We’re solving big national research problems at CSIRO, particularly in climate and environmental science, so it’s important that we walk the talk when it comes to sustainability,” Dr Taylor said.
CSIRO scientists using the GPU cluster have already seen dramatic performance speed-ups on a range of research projects from 3D image reconstruction, analysing genetic data from wheat breeding experiments and modelling interactions between nutrients and plankton in the oceans.
“Making the Green500 List and seeing the speed ups scientists are achieving has certainly validated our decision to explore this technology and, one year on, it’s really paying off for CSIRO,” Dr Taylor said.
CSIRO is also a joint partner in other high performance supercomputers featured on the TOP500 list including: the National Computational Infrastructure (NCI) facility #51, iVEC system in Western Australia #87 and the Bureau of Meteorology/CSIRO Sun Blade System #161.
CSIRO became an NVIDIA CUDA Research Centre in June 2010.
Have your say...
The approval of your comment is at the discretion of this article's publisher. Write your comment with the following in mind to ensure the highest likelihood of it being approved:
- No promotional undertones
- No use of profanity
- Good spelling, grammar and layout
- Check punctuation, language and missing words
- No use of aggression
- No unsubstantiated claims
We reserve the right to remove comments at our discretion.
Your name is used alongside Comments.