HPC Smackdown: GP-GPU vs the Latest and Greatest CPUs
The 21st century equivalent of Frazier vs. Ali?
In this corner we have the GP-GPU (General Purpose Graphical Processing Unit) talking trash about TeraFLOPS and how the way forward for HPC is SIMD parallelism. In the other corner is the traditional CPU sporting the new multi-core look. Can the “do it all” CPU still deliver the performance? Are all FLOPS created equal? What about the tag team partners of OpenCL, Cuda, Brook vs the Fortran and C? Who is going to win this number smashing match?