you don't even have to, you can do a single GPU passthrough to VM and integrate it seamlessly.