I can see a scenario where we no longer need to dedicate a physical GPU to a VM with passthru (say to use a windows game/program which cannot run on Linux) and instead split the GPU between the host Linux and the VM.
I wonder if this can be done dynamically, ie. you do that only when you start the VM.
I could be wrong about this but I think it is always dynamic
I’m also really interested in virtio GPU as it seems fairly vendor independent. All it does is provide a virtual GPU that then passes the calls down to the GPU driver of the host
It might not
All this allows is for you to split your GPU into multiple virtual GPUs. The virtual GPUs then can be passed to a VM.
Ah. Well, at least now I know the problem domain, even if it isn’t relevant to me.
Thanks!
I can see a scenario where we no longer need to dedicate a physical GPU to a VM with passthru (say to use a windows game/program which cannot run on Linux) and instead split the GPU between the host Linux and the VM.
I wonder if this can be done dynamically, ie. you do that only when you start the VM.
I could be wrong about this but I think it is always dynamic
I’m also really interested in virtio GPU as it seems fairly vendor independent. All it does is provide a virtual GPU that then passes the calls down to the GPU driver of the host
Wonder how well this might work for AI stuff… Dedicate GPU segments to processing on stuff VMs
I’m more curious about Linux gaming and VDI honestly
It would be amazing if I could split a high end card between a few people.
I recall reading a post on Reddit where somebody was able to have two people playing from a single machine with dual video cards dedicated to VMs.
More funn!