What license for vGPU in Linux VMs


I work for a company that develops a system that runs on Linux that does remote rendering for medical applications. I would like to be able to run it on ESXi with vGPU. It is completely unclear to me if I need a software license for this application and if I do what type I need. My application creates one or more OpenGL contexts per user for remote rendering, but it does this in a single VM. It does not use common RDI infrastructure like RDP, it is a proprietary system. My wish is to get as much of the native GPU power through to the VM. Can you advise?

For testing purposes I plan to buy a Tesla T4 and stick it in a test server that we already have running ESXi 6.5. If I need a license for the VM can I buy it directly from NVIDIA or where would I go?

I have browsed https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/solutions/resources/documents1/Virtual-GPU-Packaging-and-Licensing-Guide.pdf and NVIDIA® Virtual GPU Software Licensing Frequently Asked Questions. If my answer is in there, I failed to find it.

I really hope you can help me.

Best regards,

Hi Thomas,

To be honest I’m wondering why you couldn’t find the answer in the licensing guide. For sure you will need a license for your use case. It is required to have at least a vPC license or a vWS license if you need a profile with more FB or CUDA support.
NVIDIA doesn’t sell directly so you need to go to Distis or OEMs that are able to resell NV licenses.
BTW: T4 seems to be the right GPU :)


Thanks for your answer, Simon. I think part of the problem with the licensing guide for me is that it frustrates me how much of the advantage of vGPU that is chopped off by having to deal with a complicated license model. If you could take a humble wish back to the product managers it would be to drop this per-VM multiple types of clients licensing mess and just recoup the revenue loss by making the data center GPUs (even) more expensive.