PVE 8.4.1 host crashes when passing Nvidia K80 to VM

KNOW

New Member
May 27, 2025
1
0
1
Hello, I am having trouble getting an Nvidia K80 passed through to a vm.
when I launch the VM the host locks up and I need to reset the machine. I have another GPU (780Ti) in the system and when I pass that through everything works okay.
if I pass one or both of the GPUs on the K80 it just locks everything up (remotely) if I go to the physical machine I can still appear to do things in the shell however if I for example sudo reboot the machine will hang instead of rebooting so I am not sure if I can actually do anything to the machine after the VM kills the host.

Here is the VM config - I have tried this with and without the pci-hole64-size argument. I got it from a thread here where someone was trying to pass multiple video cards to one machine. it did not change anything for me

agent: 1,fstrim_cloned_disks=1
args: -global q35-pcihost.pci-hole64-size=2048G
boot: order=scsi0;ide2;net0
cores: 5
cpu: x86-64-v2-AES
hostpci0: 05:00.0;06:00.0,pcie=1,rombar=1
ide2: local:iso/ubuntu-24.04.2-live-server-amd64.iso,media=cdrom,size=3137758K
machine: q35
memory: 4096
meta: creation-qemu=9.2.0,ctime=1748320147
name: Ollama
net0: virtio=BC:24:11:65:3F:64,bridge=vmbr0,firewall=1
numa: 0
ostype: l26
scsi0: LVM1:vm-100-disk-0,iothread=1,size=90G
scsihw: virtio-scsi-single
smbios1: uuid=2c59dde8-5314-42cf-a4c8-6fd0a61bafe1
sockets: 1
vmgenid: d7448684-4109-44a0-86e0-a5bdeccd44f5
The host machine is an Asus TUF gaming B550 plus, AMD Ryzen 5 5600X, 64GB of ddr4 3000, and a few 250GB ssds and an older 1200w psu.
Proxmox 8.4.1, I have the "no subscription" repository enabled and the host machines software is up to date. the guest VM is ubuntu server 24.04.2

I see there are some people who have passed through a few K80s successfully, but this is an old GPU and I wonder if the software I am running is too new for this to work right.
 
Last edited: