KVM and Pass-through Questions

Hey guys I’m new and I’m also a noob at this stuff, but I have a few questions that I’m hoping that you fellas can answer.

Now this idea looks awesome as I eventually want to transition to having Linux as my host operating system and running my games and other windows stuff through a vm, with GPU pass through. So here is where my questions start…

  1. Now I assume since Linux is the host OS in this case, KVM is used? (kinda like in proxmox, as that is what I am messing around with atm, and yeah I am still learning this stuff so my knowledge is limited)
  2. Wendell mentioned something about input problems, so that’s regarding mouse and keyboard input and more importantly, latency?
  3. I’m happy to have two monitors and having the output of the GPU go to another monitor for windows, but in doing so will I be able to achieve 144Hz/FPS? (I assume I would since it would be coming directly through the GPU) OR is it limited to 60Hz/FPS due to virtualization limitations? Getting 144Hz would be good for those “twitch” shooters :smiley:

Cheers guys! :slight_smile:

Well, first off, Valve being hostile to VAC inside virtualization means CS:GO will not be possible unless you risk a VAC ban.

The latency issue stems from the USB passthrough of the input to the VM. You can bypass that completely if you use a USB controller card or use pci.stub to not allow the host OS to hook onto a select amount of USB ports on your motherboard to allow that to be passed through directly to the VM.

And yes, KVM and QEMU is primarily used.

I did a quick search, so from what I understand KVM allows for containers since its a kernel based vm (kinda like containers in proxmox) and QEMU actually allows for full on virtualization, is this right?

They actually looked into that issue as far as I know. Also why would you run CSGO virtualised under Windows if you can just run it natively in the first place :stuck_out_tongue:

My idea would be to keep all the games inside the windows vm, but then again there is that work around…

Hey Guys, one other thing… if I want to achieve a GPU pass through, using fedora, ubuntu or whatever as a host, am I allowed to allocate all my 4 CPU cores to both host and guest systems? (since I only have a 4 core CPU on a 1150 socket)

Edit: Also if i can allocate all 4 cores without something going wrong, can i do that for another guest vm, ie: running two guest vms at once

No, You can’t, You never want a guest to have more cores than your host if the host has fewer cores than guest overall performance will suffer

Thought so, thanks!

Looks like if I want to do pass through ill need to get a new mobo and cpu at the least, since there isn’t much of an option on 1150 socket.

Would this be the same if I were to have an i7 4c/8t since I only have an 1150 socket? Can I get away with dedicating 4 cores each vm?

you don’t “allocate cores” to your host machine. can you allocate all four cores to your VM? yes. and it’ll be fine so long as you’re doing nothing at all on your host machine …but then, you’re gaming on the vm, right? why would you be doing anything on the host anyway?

when I set up my son’s machine, we put together two configs. one with only two cores, for stuff like updates, downloading, etc., where we expected to be using the host while we waited, and one with all four cores allocated to the vm for gaming. (the 2-core version also skipped passthrough.) worked just fine.

Yeah sorry I was meant to say, use all cores on the guest (that the host has). Looks like I’ll need to upgrade then, probably wait for the zen refresh.