I wonder if anybody could advise me if there is a simple way to get nvidia-installer to do a “minimal” install for CUDA only (presumably the kernel module, the cuda.h header file, the cuda shared library and symbolic links), on a system which is running X but on a non-nvidia card? I had a look in the advanced options but couldn’t see a way to do quite this.
I have a simultaneous cuda/ati stream system working (see this thread at the ATI Stream forum) with the ATI card doing the (accelerated) graphics and the cuda card working using the script provided in the nvidia release notes but I put the nvidia driver stuff in place “manually”. It’d be great if one could just type “./nvidia-installer --cudaonly” or similar, particularly when it comes to updating drivers etc and indeed in just being confident it is all there correctly.
Thanks for any advice,