GPU Computing

rainwilds

Distinguished
Oct 17, 2010
80
0
18,630
Hi all,

I have done a little research lately trying to educate myself on what is meant by 'computing' in GPU-Z where they have check-boxes for OpenCL, CUDA, PhysX and DirectCompute. I have a few questions for the knowledgeable out there:

1. Do both Nvidia and AMD support OpenCL and DirectCompute?
2. Does AMD have a similar physics engine to Nivida?
3. How do workstation cards differ to gaming cards for computing?
4. Lastly, what is CUDA?

Thanks
 
Solution
Their new cards(GTX 4xx and HD 5xxx series and up) support both OpenCL and DirectCompute as they are DX 11 compatible

AMD has one, its just not used very much, GPU accelerated PhysX is fairly rare itself.

Workstation cards have ECC memory for more reliable calculations, often times more memory so they can call on the CPU less, and better drivers that are more optimized to handle large calculation based tasks vs rendering based tasks.

CUDA is nVidias custom version of C/C++ that lets them take some tasks which are heavily parallel and thus poorly suited for heavily serial CPUs and offload them to the GPU, AMDs equivalent is called Stream, but again, isnt used nearly as often.
Their new cards(GTX 4xx and HD 5xxx series and up) support both OpenCL and DirectCompute as they are DX 11 compatible

AMD has one, its just not used very much, GPU accelerated PhysX is fairly rare itself.

Workstation cards have ECC memory for more reliable calculations, often times more memory so they can call on the CPU less, and better drivers that are more optimized to handle large calculation based tasks vs rendering based tasks.

CUDA is nVidias custom version of C/C++ that lets them take some tasks which are heavily parallel and thus poorly suited for heavily serial CPUs and offload them to the GPU, AMDs equivalent is called Stream, but again, isnt used nearly as often.
 
Solution