Did you know you can do this?

Sadly, integrated GPUs do not output any images. It rarely works. I have tried various iGPUs (ESXi and Proxmox). Passthrough does work in principle, but there is no image output. So for a headless sever setup which uses iGPU through passthrough and needs graphics acceleration (e.g. Plex) its nice.

you can also do sr-iov on the integrated gpu (10 or so vms) via a patched i915 integrated gpu driver if you don't want/need the resources of a standalone gpu. there's also a patched driver for pascal cards that allows for sr-iov as well, but I don't think you could fit one into one of those minisforum machines due to space.

Had to give thumb down for this. There's a lot of much more superior tutorials conserning this subject. Let alone that not a single word is mentioned about hardware requirements for passthrough to work as you can't do this on mmuch of the older hardware. Thing that's generally always mentioned in those other tutorials.

Great Video, but another question for USB passthrough: Its possible to passthrough mouse and keyboard in vm and switching them between my proxmox host and my workstation pc, without need to replugging them? I heard something about KVM Switch

So would the steps be the same here if I wanted to do this with a Radeon Pro v340? And would I be able to use an older Intel Mac Pro? I am 110% noob when it comes to VMs and I would like to tinker a bit and try to make 2 Mac-based workstations out of 1 if possible.

In a practical sense how much of the GPU's actual power is available to the VM? For example if you have a headless server with a GPU whacked in to perform AI engine calculations would this be on par with a non-VM'd (straight to baremetal) installation for the GPU performance? Probably an unfair question unless it's tested side by side on equivalent hardware I guess.
Also that's a big/little core design chip isn't it? If so how does Proxmox handle the big/little split for VCPU allocations?

Yeah please do a full walkthrough please i have an old macbook with no gpu and i don't wanna install fedora on it but want to move to fedora and thought my other computer running as server can pretty easily do it and I can move away from my slow macbook.
Full walkthrough……PRETTY PLEASE

The GT 1030 passed through to a VM is quite meaningless unless you are connecting a display directly to the GT 1030, since the GT 1030 does not have an encoder and trying to access the VM using any remote access software (like moonlight/sunshine or nomachine) would mean the vCPU ends up doing the video encoding work. So for the way it is done in this video, it works. But if you are going to remotely access the VM, you will end up with no hardware video encoding. I know… because I used to do this with a GT 1030 and learned the hard way.

sound not clear like normal……….

why does the windows virtual machine still show high ram usage even after virtio driver install

Pretty cool