Yes, it is feasible to utilize the integrated GPU for virtualization via GPU passthrough.
Yes, it is feasible to utilize the integrated GPU for virtualization via GPU passthrough.
From a practical perspective, running advanced graphics through a hypervisor in KVM-style configurations isn't logical. The host operating system typically manages the VMs and might serve as a central file server among them.
I can only mention that it might be feasible in theory, but I haven’t heard anyone actually attempt or succeed with it. The main challenge lies with the host; when you try to connect a GPU to a VM, it depends on the type of GPUs it supports rather than the category. Essentially, you're routing through a PCI_e slot, and whatever you plug in there will dictate what the VM receives. Even if the iGPU uses the same connections as a PCIe slot, it won’t be recognized for a VM. That said, it’s not impossible—there might be a Linux driver or add-on that could access that slot and forward it to a VM. In practice, it’s possible, but we need someone to actually write the code.
Also, I realized recently that certain devices can be routed using their device IDs. If the iGPU has a unique identifier, it might be feasible to pass it along. Whether it works or not will still be uncertain.