PCI passthrough of P6 card to single vSphere VM

A little confused as to what is required to both make this work and to be in compliance…

I have a need to create a Windows server that has full access to a physical GPU. Rather than purchasing a dedicated Windows server with a GPU, we want to use the VMware/Cisco blade infrastructure already in place for our datacenter environment.

We just want to pass through the Tesla card to an instance of Windows Server and use the Tesla drivers in Windows, same as if it were a physical machine using the same card. No virtual desktops, virtual applications, vGPU etc in use. Is a Grid license needed for this? When I contacted support, all I get is "vGPU requires a license" and a reference to the vGPU documentation.

Thanks in advance


Even in Passthrough, you still need a license for a Tesla GPU unless the GPU is running in Compute mode. What’s the workload you’ll be running?


Tesla driver = no license
vGPU driver = license

So simple…