New to OpenCL - questions on SDK, Driver etc ...

Accelerated Computing CUDA CUDA Programming and Performance
1

Apologies for cross posting with the CUDA on Linux forum, but these are some OpenCL specific questions - So here goes -

Here are my cards -
nVidia Corporation Quadro NVS 290 (rev a1)
nVidia Corporation G80 [Quadro FX 5600] (rev a2)
nVidia Corporation Quadro NVS 295 (rev a1)

And I am referring to this page - [url=“OpenCL | NVIDIA Developer”]http://developer.nvidia.com/object/opencl-download.html[/url]

(a) They are CUDA enabled cards. Does it automatically mean they are OpenCL enabled?
(B) What driver should I install on Linux x86_64? nVidia says 190.53, but I understand OpenCL is not supported under it? Should I go with the 195.30 beta? Should I go with the one listed on the webpage above 190.29 ? Which one works with the above cards and allows for OpenCL code?
© I am really new to OpenCL , so if this sounds silly, apologies! I still need to install CUDA toolkit, correct? I can have the CUDA toolkit side by side with the OpenCL Visual Profiler v1.0 Beta, and run both “CUDA SDK samples” and “GPU Computing SDK samples” at the same time, right?

Thanks!

2

CUDA enabled cards are automatically OpenCL enabled.

Get the newest drivers, 195 or newer preferably. You will need SDK 3.0 Beta to work with the 195+ series of driver (2.3 won’t work). You don’t need the CUDA toolkit but you can use it side by side with any OpenCL tools.

GPU Computing SDK contains everything CUDA SDK does and more (OpenCL and DirectCompute samples). You’ll only need the GPU Computing SDK.

You can get the new drivers and SDK here [url=“http://forums.nvidia.com/index.php?showtopic=149959”]http://forums.nvidia.com/index.php?showtopic=149959[/url]
For OpenCL development you’ll need “Linux 64 195.17” and “GPU Computing SDK for Linux”. I’m not entirely sure how the Linux versions are packaged, you might also want “CUDA Toolkit for [your distro]” but it’s not needed on Windows so I guess not.