Yes. Its a bit of a PITA. It works really well using SR-IOV. Ultimately I like this method much better than a simple PCIe passthrough of the device because you can share the iGPU with up to 7 VMs simultaneously AND you don't lose video for the host itself. But for now it is not mainlined into the kernel and has to be done using a DKMS repo and a bit of config (its this way though at least kernel 6.8, with Proxmox currently at 6.5). I'm not sure what the status of getting this mainlined is.
Use
this guide. Derek does a great job of detailing the steps and has clip-able commands so all you have to do is copy and paste each step.
His guide works whether you are using Secure Boot or not - but I prefer not having secure boot on if you are using DKMS modules because you have to authorize the module each time it re-compiles, which means each time the kernel gets updated in Proxmox. If secure boot was enabled when you installed proxmox you'll really want to follow the last step and pin the kernel. Authorizing the module requires console access at boot, which is a problem if your machine is stuffed headless into a server stack somewhere (ask me now I know this
).
If you DO NOT have secure boot enabled make sure you skip the steps where you import the module keys into MOK or you will find yourself stuck on reboot with the bootloader looking to approve the keys, but the Proxmox "non secure boot" install does not include the tool required to do it...which is kinda bad. The step you want to skip in a non secure boot install is "
mokutil --import /var/lib/dkms/mok.pub"