This is the steps I took to get GVT-g working on my machine.

# nano /etc/default/grub

Find the line that starts with GRUB_CMDLINE_LINUX_DEFAULT and make that line to look like this:

GRUB_CMDLINE_LINUX_DEFAULT="quiet intel_iommu=on i915.enable_gvt=1"

Update the grub config:

# update-grub

After that add the following modules to **/etc/modules**:

vfio
vfio_iommu_type1
vfio_pci
vfio_virqfd
kvmgt

Regenerate initramfs:

# update-initramfs -u -k all

Reboot the server.

After reboot check if you get any output with this command again:

# ls /sys/bus/pci/devices/0000\:00\:02.0/mdev_supported_types/

Should show something like:
i915-GVTg_V5_4 i915-GVTg_V5_8

If it does, now you can add a new PCI device to your vm from the gui and the GPU will show as a mediated device.

I think that you need to run a Q35 vm machine with ovmf bios to get it working properly. I can be wrong on this one, but anyway, that is what I'm running.

Also make sure all virtualization options is enabled in your bios such VT-d and that virtualization is enabled, but that have you already probably done.

引用链接：

https://forum.proxmox.com/threads/intel-gvt-g-questions.82165/

https://192.168.8.35:8006/pve-docs/chapter-qm.html#_mediated_devices_vgpu_gvt_g