(hypothetical) headless CUDA environments

This is a hypothetical question. I am curious to know if OpenGL is usually available on a CUDA cluster, and if so is it something people frequently use.

I have had access to some CPU clusters, and OpenGL was not available (which makes sense). But since CUDA and OpenGL are in some senses two different “ways of looking at” the GPU, it would make sense that OpenGL would be available.

Sorry if this question is inappropriate for this forum.

On linux, OpenGL is “available” on any NVIDIA GPU* (excepting Tesla K20c) that has a proper driver install. However “available” here might not mean what you think it does.

This may help illuminate, and note that there is a whitepaper linked at the end of the answer there:


On Windows, OpenGL is available on any NVIDIA GPU* that has a proper WDDM driver install. Not all NVIDIA GPUs support operation in WDDM mode, however.

*My comments apply to the usual PCIE and SXM2 GPUs that are commonly discussed in this forum. For TX1/TX2 or other products, consult the appropriate forum or other resources.