hope somebody can help with a kind of bizarre issue, only experienced on antergos/arch linux distros (yes, for testing I tried to build also a pure arch linux machine).
Tools used: virtual machine manager (virt-manager), libvirt, KVM, ovmf-git UEFI bios, Windows 10 standard ISO.
So after following a guide in how to setup my system (namely http://dominicm.com/gpu-passthrough-qemu-arch-linux/) everything works flawlessy (installing Windows 10 guest and configuring it, by using the Spice console), until I start to pass my dedicated GPU - an Nvidia GTX 1080.
When I try to boot it, virtual machine manager CPU graph shows that the usage skyrockets to 100% and I have no display output; if I’m patient for around 10-15 seconds, then I will have display output on passed gpu (I see the tianocore splash screen) and the system starts correctly with no performance loss.
I have seen tat this behavior is becoming worst is I assign more memory to the guest OS. Let’s say I have the 10-15 seconds cpu 100% usage/no display output if I assign 8GB ram, 25-30 if I assign 16+GB.
If I assign 2GB of RAM to the virtual machine,boot and display output, it’s quite immediately.
This lead me to think that some sort of bug (tianocore git UEFI bios? libvirt? kernel?).
As I have tested other linux distros, like Fedora 25, Linux Mint 18, Debian Stretch, I am experiencing this “issue” only on antergos/arch.
Any suggestions will be greatly appreciated.
My system info:
Asus X99-E WS
Intel i7 5930K @ Stock 3.5GHZ
32GB DDR-4 2666MHz RAM
Samsung SM951 M.2 SSD Disk
Linux GPU Nvidia GTX 670
VM dedicated GPU Nvidia GTX 1080
Thanks and regards