CUDA 3.2 on RHEL 5.6

OS: RHEL 5.x (originally 5.4, currently 5.6)
Kernel: 2.6.18-164, 2.6.18-238
GPU: Tesla s1070
Driver: currently 190.53
CUDA: currently 2.3


I have a system that was originally installed with RHEL 5.4 (kernel 2.6.18-164). Driver 190.53 and CUDA toolkit / sdk 2.3 appeared to work fine with this kernel.

A user has requested I update to CUDA 3.x for Matlab PCT support. In addition, since installation, this system was updated by yum to RHEL 5.6 with kernel 2.6.18-238. From what I can see, release notes for CUDA 3.1 don’t mention kernel requirements. CUDA 3.2 release notes mention support for RHEL 5.5 with kernel 2.6.18-194.

Here’s my question. Given that yum makes rollback to RHEL 5.5 difficult (if not impossible), is there any reason not to try CUDA 3.2 with RHEL 5.6 / 2.6.18-238? Any caveats? If I take the try-it-and-see approach, which driver is best - the devdriver available on the page with CUDA downloads, or the most recent GA driver release available on the driver downloads page?

Thanks in advance for any advice.

AFAIK there’s no reason why it won’t be fine. The newest r260 or r270 driver should be fine.

kvalentine, You might was to point your user to Jacket if they are interested in GPU computing in MATLAB. We’re happy to provide an intro demo to MATLAB GPU computing too if that would help.