> GPU passthrough is also doable pretty easily on NVidia nowadays.
By actively working against Nvidia who could break it again at any time if they wanted to:
> Starting with QEMU 2.5.0 and libvirt 1.3.3, the vendor_id for the hypervisor can be spoofed, which is enough to fool the Nvidia drivers into loading anyway.
If you already have Nvidia, fine, but to me this reads as a strong reason to not buy Nvidia if you can help it.
Oh, don't get me wrong, I surely don't want to encourage you to buy Nvidia.
To be fair here, AMD also has some gripes with VFIO: Namely, the reset bug on Navi (which I personally didn't experience, but read about quite a few times) and disabled vGPU support on their smaller cards, which is, as far as I know, only a software solution and not really something that would steal their business customers either.
Still, I'm rooting for AMD, if only for the fact that they're the reason it doesn't take six CPU generations any more to have a 50% performance bump.
By actively working against Nvidia who could break it again at any time if they wanted to:
> Starting with QEMU 2.5.0 and libvirt 1.3.3, the vendor_id for the hypervisor can be spoofed, which is enough to fool the Nvidia drivers into loading anyway.
If you already have Nvidia, fine, but to me this reads as a strong reason to not buy Nvidia if you can help it.