MrGaribaldi
Regular
Yeah, too bad they burned that bridge. But since they are working with ARM chips for Tegra, it could be another option.
As for powerful enough, I agree this would probably not work as a gaming solution, but for a of HPC/GPGPU it could work. The ARM core running a "Master" kernel, launching and coordinating CUDA kernels, to avoid the long round trip to the CPU and back. Think a bit like the CELL solution, and then being able to plug in as many of these as you want, sort of a distributed solution in a single system.
EDIT: This could also be interesting in a perf/Watt perspective.
As for powerful enough, I agree this would probably not work as a gaming solution, but for a of HPC/GPGPU it could work. The ARM core running a "Master" kernel, launching and coordinating CUDA kernels, to avoid the long round trip to the CPU and back. Think a bit like the CELL solution, and then being able to plug in as many of these as you want, sort of a distributed solution in a single system.
EDIT: This could also be interesting in a perf/Watt perspective.