OpenCL Enabled Products NVIDIA hardware

Any timelines on the availablity of OpenCL enabled NVIDIA drivers?

AMD already has one:…tream_9270.html

Best Regards

What part of “OpenCL®-ready technology” and “Support for the OpenCL specification (when available)” on that page indicates that one can already download and use OpenCL on this card? It is just a marketing gag. If NVIDIA wanted to, they could go all the way back to the 8800 GTX and market them as “OpenCL-ready” too.

The most I can find about OpenCL/AMD on the web is that they demonstrated a CPU demo last month:

We had a OpenCL GPU demo on Linux back at SIGGRAPH Asia:

So, we’re working on them.

Goood to know! Thanks!


I have my own doubts on whether a OpenCL program would scale uniformly on all platforms?

Will the program get me optimal performance in all?

Any comments?

Thanks for the link. The same site talks about NVIDIA’s demo as well on the GPU :

That will offcourse depend on how good of a job the platform people are doing in implementing OpenCL…

Yeah, You are right. I also used to think so… But it might depend on the parallel algorithm as well…

It will be fun when the products comes out and people start using it. Thanks!


Request you to have a look @…/ThomasHL09.pdf

Check the “Conclusion” part and read the line that starts like “The surprising result…”

The paper tries to compare CPU, GPU, FPGAs in Random Number Generation and they found that performance pretty much depends on the algorithm used and it varied from platfom to platform…

I am more worried about such aspects with OpenCL.

Best Regards,

But the platform does not change between CUDA & OpenCL. Also they are very similar in programming technique (with OpenCL resembling the driver API quite a lot).

I mean, even the optimizer will be shared between CUDA & OpenCL on NVIDIA hardware (OpenCL will compile to PTX just like CUDA).

Off course when changing from a GPU to another OpenCL-supported platform you code is probably not optimal anymore, but there is no such thing as magic ;)

This is what I have been talking about. ANyway, I think nothing will be of help here… But if we write CUDA with mind, hopefully we wont have significant drop in performance on other platforms. We need to wait and see…