NVidia have had similar dodgy tactics in the past, including blocking their GPUs from being used in a VM using IOMMU. For my use case (where I had Windows in a VM on a Linux host) this was a massive PITA.