Kudos for CSIRO’s supercomputing capability

GPUs are the computer hardware which lies at the heart of game consoles and, increasingly, supercomputers.

CSIRO’s Group Executive, Information Sciences, Dr Alex Zelinsky, said the announcement –at the International Supercomputing Conference in Hamburg, Germany – furthers CSIRO’s goal of being a world leader in the application of GPU technology to a broad range of scientific and industrial problems.

“To be involved in NVIDIA’s CUDA Research Center Program, which is designed for institutions that embrace GPU computing across multiple research fields, is a great honour for CSIRO,” Dr Zelinsky said.

“We’re excited to be in such good company. CUDA Research Centers include Johns Hopkins University (US) and Nanyang Technological University (Singapore).”

CSIRO is currently the only CUDA Research Center in the southern hemisphere.

CSIRO’s GPU cluster, with 256 GPUs, was the first of its kind in Australia and is one of the world’s fastest computers.

GPUs speed up data processing by allowing a computer to massively multi-task through parallel processing.

NVIDIA’s expertise in programmable GPUs have helped make supercomputing inexpensive and more widely accessible. The company is based in California.

NVIDIA’s Director of Research, Dr David Luebke, said the CUDA Research Center program is designed to recognise and encourage the use of GPUs for scientific and high performance computing.

“CSIRO will gain access to the latest developments in GPU computing and become part of a wider community of organisations with GPU facilities, sharing information and ideas,” Dr Luebke said.

Platform Leader of CSIRO Computational and Simulation Sciences, Dr John Taylor, said that, as one of the world’s most diverse research organisations, CSIRO is uniquely placed to put GPU technology through its paces.

“We’re using the GPU cluster to speed up projects like modelling nuclear analysers, running 3D X-ray and CT image reconstruction, measuring uncertainty in complex environmental models and understanding biomechanical processes like human swimming,” Dr Taylor said.

“And we’re getting results 6-200 times faster than before.”

The GPU cluster is in everyday use with over 100 CSIRO scientists trained to use it. It was installed in Canberra last November by Xenon Systems of Melbourne and runs Linux and Windows applications.