- Joined
- May 2, 2013
- Messages
- 170 (0.04/day)
@newtekie1
Do you have a proof of the GTX960 being better at MSAA?
With regard to your stand on CPU "physics", well... good luck letting the DX11 runtime manage 10 synchronous threads. It's simply not possible with the current or rather previous gen APIs.
I'd say that AMD's HSA chips are quite a good option for heavy simulation and AI processing. The GPU in these chips can address the whole virtual memory space and communication between CPU and GPU is done through hardware and without any context switches. All you need to do is add pointers in your code where you want execution to switch from or to either the CPU or the GPU. AMDs HSA chips seem like the most sensible platform for these kinda workloads.
Do you have a proof of the GTX960 being better at MSAA?
With regard to your stand on CPU "physics", well... good luck letting the DX11 runtime manage 10 synchronous threads. It's simply not possible with the current or rather previous gen APIs.
I'd say that AMD's HSA chips are quite a good option for heavy simulation and AI processing. The GPU in these chips can address the whole virtual memory space and communication between CPU and GPU is done through hardware and without any context switches. All you need to do is add pointers in your code where you want execution to switch from or to either the CPU or the GPU. AMDs HSA chips seem like the most sensible platform for these kinda workloads.