HPC Users

General Purpose Computing on Graphics Processing Units

New General Purpose Graphics Processing Units (GPGPU) have come to market over the past year that enable users to perform computations that are normally executed by a compute core on a GPGPU. The best known GPGPUs are those from nVidia, and more recently ATI.

GPGPUs operate as parallel processors that are equivalent to multiple high performance cores, essentially a vector processor, and often speed up computationally intensive applications from 10 to 100 times compared to using a traditional compute core.

VPAC operates the 'Enrico' node, sponsored by Xenon Systems, with two Tesla C2050 GPUs having 64 bits computational capability and error correcting memory. Enrico enables VPAC users to both experiment and test GPGPU techniques using their own code written with either the Compute Unified Device Architecture (CUDA) or OpenCL, as well as running GPGPU enabled applications.

VPAC also provides researchers from Member institutions access to the specialised GPGPU cluster called MASSIVE, located in the Clayton precinct. See massive.org.au for more information on the facility, or to apply for a project account there through the VPAC share.

The nVidia CUDA Zone is an excellent first resource for researching applications that are GPGPU enabled using CUDA. VPAC staff also host the Melbourne area GPU Users group .

For further information on submitting jobs to VPACs GPU nodes visit the GPU user tutorials page.

Top of Page