View Full Version : Nvidia Releases OpenCL Driver to Developers
trinibwoy
04-20-2009, 10:38 AM
SANTA CLARA, CA—APRIL 20, 2009—NVIDIA Corporation today announced the release of its OpenCL driver and software development kit (SDK) to developers participating in its OpenCL Early Access Program. NVIDIA is providing this release to solicit early feedback in advance of a beta release which will be made available to all GPU Computing Registered Developers in the coming months.
http://www.nvidia.com/object/io_1240224603372.html
Guess you need to be a registered developer to gain access but at least it's a sign that things are moving. Hopefully more OpenCL stuff will start coming down the pipe once the standard is fully ratified.
Eastcoasthandle
04-20-2009, 10:44 AM
My cpu physics (in Ghost Busters) thread is gone :shrug:. There was some decent discussion going on there too. I hope to see more developers use OpenCL in their current (and future) programs.
paulhamm
04-20-2009, 11:42 AM
So does this signal the beginning of the end for CUDA and PhysX?
Newblar
04-20-2009, 12:04 PM
So does this signal the beginning of the end for CUDA and PhysX?
hopefully, yes
bedlamite
04-20-2009, 12:24 PM
So does this signal the beginning of the end for CUDA
No. Because CUDA is whole technology connected with using GPU for calculations, not only C for CUDA extension. It's also upcoming Fortran derivative, and Open CL tools. So OpenCL is strongly connected CUDA and will be executed by CuDA on NV cards...
grimREEFER
04-20-2009, 12:30 PM
both opencl and directx 11 compute shader are both just handled through cuda iirc.
so all that money they invested in it is likely not to be a waste
trinibwoy
04-20-2009, 12:41 PM
So does this signal the beginning of the end for CUDA and PhysX?
No to both. CUDA is a lot more than just OpenCL. PhysX is a library, not a compute API. You can write PhysX in OpenCL, CUDA, C++ or whatever. Perl if you want.
Another reason for CUDA to stick around is that Nvidia won't wait for some standards body to determine their hardware capabilities. OpenCL is pretty much G80 level, I'm sure GT300 is going to have a lot more fancy bits that OpenCL won't expose and you'd need CUDA to access them.