how can i use GPU when i dont have the NVIDIA graphics card?

I am interested in GPU Computation functionality of Parallel Computing Toolbox. I have an INTEL graphics card. When I attempt to use GPU Computation functionality by calling the GPUARRAY function I get the following error:

>>paralleldemo_gpu_devices

Error using gpuDevice (line 22) The CUDA driver was found, but the function "cuDriverGetVersion" was not found. This can occur with old versions of CUDA. The required CUDA version is: 4 or greater. The library name used was: nvcuda.dll.

is there a solution?

Answer 1

[Walter Roberson]

MATLAB does not currently support using GPU with non-NVIDIA cards.

You may wish to talk to a sales representative about the progress in integrating Accereye's Jacket product with the Parallel Computing Toolbox. (There is no benefit in asking that here: non-disclosure agreements are in place on the topic.)

Answer 2

[Marc]

Is this an integrated card? I was under the impression that Intel only makes integrated graphics cards. If so, I don't think you want to do GPU calcs on something that is integrated.

That said, I have to say, the GPC functionality is nice to play with using a NVIDIA card on my 15" MacBook Pro. Downloaded the right driver and had no problems with the Parallel Computing Toolbox.

Unless you are tied to a budget and a lot of software on Windows, I recommend the 15" MacBook Pros. The latest versions of Matlab for MacOS are very good compared to their Windows counterparts.

Not perfect but I am seeing less fungibility issues going from my work PC to my personal Macs...