CSIRO seeks petaflop supercomputer
- 14 November, 2016 14:11
The CSIRO has invited tenders for a replacement of its BRAGG supercomputer cluster, with the organisation hoping to have a new petaflop-capable system up and running in the first half of next year.
BRAGG, used by the CSIRO’s Information Management and Technology team to deliver its Accelerated Computing Service, was first installed in 2012 and upgraded in 2014.
According to the CSIRO, it was one of the first supercomputer systems to combine CPUs and GPUs for processing. The BRAGG cluster is currently ranked 410 on the Top500 supercomputer list, capable of delivering peak performance of 472.5 teraflops.
The CSIRO said today that it would spend up to $4 million on the new system, including three years of support.
The new system will be based on SUSE Linux Enterprise Server, with each two-plus CPU compute node running Power or Intel x86 64-bit CPUs. Each node will have at least two GPUs running NVIDIA’s Pascal architecture.
The investment is “an integral part of our strategy working alongside national peak computing facilities to build Australian HPC capacity to accelerate great science and innovation,” CSIRO acting deputy CIO, scientific computing, Angus Macoustra, said in a statement.
The new system will be installed in the CSIRO’s Canberra data centre. Tender documents state that it should be capable of delivering at least 1 petaflop performance.