GPU Compatibility for Theano [NVidia GeForce 8800GT]

Question

I am currently working on an ML project on my personal computer that has an AMD graphics card. I have an old NVidia 8800GT card that I could plug in for CUDA accelerated convolution, but I haven't found if it is compatible with Theano. Googling has surprisingly been unsuccessful.

I know the 8800GT supports CUDA and I've done some CUDA work with it in the past, but is compatible with Theano? (or TensorFlow?)

Best, Joe

It's not clear that you'll get any benefit from 8800GT. I tried on similar price range GeForce 730 and it had about the same speed for matmul as my z420 workstation without acceleration — Yaroslav Bulatov, Apr 06 '16 at 16:34
8800GT is not supported by CUDA 7.0 or CUDA 7.5, and also not supported by cuDNN. — Robert Crovella, Apr 07 '16 at 02:08
Do you have a link to a compatibility listing? I might head out and buy a compatible mid-line GPU then. — joe smith, Apr 07 '16 at 15:17
There is a list of cuda-compatible GPUs here: https://developer.nvidia.com/cuda-gpus — abergeron, May 05 '16 at 19:40

score 1 · Answer 1 · answered May 05 '16 at 19:39

Theano has no specific requirements for cards other than "it works with cuda".

If you want to use the cuDNN layers or other specilized things, then you might need a more recent card and the requirements for those is specified in the documentation for those libraries.

GPU Compatibility for Theano [NVidia GeForce 8800GT]

1 Answers1