Is it possible to use the integrated GPU for virtualization through GPU passthrough?

[YamiYukiSenpai]

My desktop has a GeForce GTX 1060 6GB and since my Intel GPU is not being used at all, is it possible to use my iGPU for virtualization?

 

I'm planning on setting up a Windows 10 VM using QEMU (or using my already-setup Windows 10 VM in VirtualBox).

[Drak3]

If you're asking about using the iGPU for the host OS, yes.

 

If you're asking about using the iGPU for the VM, most likely not.

[YamiYukiSenpai]

13 minutes ago, Drak3 said:

If you're asking about using the iGPU for the VM, most likely not.

How come?

[Drak3]

Just now, YamiYukiSenpai said:

How come?

From a use case stand point, it doesn't make much sense to have the hypervisor using high end graphics in KVM style setups. The host OS is usually responsible for hosting the VM(s) and maybe acting as a file server between VM(s).

[Windows7ge]

I can only say that it may theoretically be possible but I've not heard anyone ever try or successfully use it.

 

The primary issue is in the host when you want to passthough a GPU to a VM it doesn't do it based on the category of GPUs it does it based on the motherboards PCI_e slots. So what you're doing is passing though a PCI_e slot to the VM. What you plug into that slot will determine what the VM actually gets.

 

So with this in mind even if the iGPU is tied into the processor using the same lines/protocols as PCI_e it is not a slot that can be given to a VM.

 

However that means it's not entirely impossible that a linux driver/add-on could exist that accesses the "slot" the iGPU is in and could pass that though to a VM. So I'm saying it's possible it could exist. It's practical it's just a matter of has anybody written the code to make it work.

 

Edit:

I just remembered that specific devices can be passed though by using device-id's. If the iGPU has it's own id (and it should) it may be possible to pass it though. Weather or not it will work is another ordeal all together.