Cuda and ATI Stream technology diff


I like to know the difference between Nvidia Cuda and ATI Stream technology

Also which one is more matures

like cars (toyota vs honda): same function, but other car body, motor and internal electronic devices.

ATI hardware and nVIDIA hardware are different. You can’t say this or this is THE BEST.

Thank You.

I agree with your statement.

I been read ATI doc on there GPU technologies, to see if any one has run the same code examples test on both technologies using OpenCL

nVIDIA supports OpenCL, but i think for other multi processor hardware (Intel Xeon…) it is only a compromise (currently?).

I don’t know how good ATi supports OpenCL?!

CUDA is quite a bit more mature than Stream, and it has a (slightly) more abstract programming model which allows your code to be simplified a bit.

The nVidia OpenCL drivers are in beta right now, and I believe that ATI has released an OpenCL compiler for their platform as well.

The OpenCL API is based on the CUDA driver API, so you wouldn’t have too much trouble porting a CUDA-based program (again…if it’s using the Driver API) to OpenCL.

I don’t know any examples that run on both yet, though it would be interesting to compare them (not only for hardware comparisons, but for compiler efficiency as well).

After trying to use both, I would say that it is much easier to develop programs using the Nvidia Cuda approach. Here we have good documentation, good forum community, and easier programming model made by Nvidia.

ATI is using Brook+, which is not nearly as mature (preprocessor for it is not finished) and is more finicky about minor problems. However, for the same price, the ATI cards are much faster at certain things than NVidia cards (password cracking is a big one).

Nvidia Cards dont have a barrel rotate instruction, which makes it much slower than ATi cards of the same price.