We attended a GPGPU programming course at the Center for High Performance Computing this week. The course focused on new developments in OpenCL and included some very interesting presentations. We had a also had a brief discussion on the new CUDA 4.0 standard which has some very powerful features.
Thanks to the South African National Space Assosiaction for organizing and also to the CHPC for hosting the venue and providing lunch.
While our cluster does not yet incorporate GPU capability we are looking
at implementing a proof of concept, so if there's anyone out
there with a valid use-case, please contact us.
Update: apparently there is a test server with n x GPU cards heading our way where 1<n<=4
Update: apparently there is a test server with n x GPU cards heading our way where 1<n<=4