GPUs In The CS Linux Clients

The CS Lab Linux clients contain Nvidia GPUs with support for CUDA and OpenCL.

These systems include:

• masters1 through masters16 (Masters' Lab)

• ugrad1 through ugrad24 (Ugrad Labs)

However, our "compute server" Linux clients, gradx, gradz, ugradx, and ugradz (that are only remotely available through ssh or scp) do not have CUDA-capable GPUs in them. You'll notice that CUDA software is available on those machines merely because we share the same configuration amongst all our Linux client systems, whether they have the same hardware or not.

Details About Our GPUs

To see specific details about our GPU, log into any of our labs' Linux computers referenced above and run:

nvidia-smi

How To Use The GPUs

We do not have specific info on the "how-tos" of using our GPUs, because... how you use our GPUs will very much depend on which library or framework you're using and what you want to do with them. Generally, CS courses that make use of the GPUs will cover how to make use of them in the course materials.

We recommend that you consult documentation for the specific GPU library you want to use. For example if you plan to use CUDA with our GPUs (and we believe several CS students do), you can check CUDA's own documentation.  Another common framework is PyTorch (which itself uses CUDA, but also provides a lot of other functionality). Our systems also support OpenCL, in case you want to write code that is portable across more GPU architectures than just the Nvidia GPUs we have. And your instructor might be able to assist you as well.