Proxmox passthrough GPU into windows VM not working

[Sir-robin10]

I've been trying for a day now to get the Nvidia QUADRO P2000 to be usable in my windows server 2019 vm, but nothing seems to work. I still keep getting the 43-error in windows devices.

This are the settings I'm using:



This is the .conf file in the qemu directory:



This configuration however, gives me the Coe 43....




What am I doing wrong here?

I tried the CPU args, but that didn't work and so on....


I followed this guide: https://mathiashueber.com/fighting-error-43-nvidia-gpu-virtual-machine/

But it also didn't work out.... Anyone who maybe can help me? For linux would be great too! (ubuntu)

 

[Stefan_R]

The code 43 workarounds found online should never be needed on PVE, as we apply them automatically for Windows VMs anyway. That aside, code 43 just indicates a general unspecified error in the NVIDIA driver, so it doesn't really narrow the root causes down.

Make sure to carefully read through our PCI passthrough wiki page, especially the part about extracting and specifying a 'romfile', that can help in such situations.

 

[mbosma]

My fix for nvidia quadro cards is to export the bios and pass the bios file through to the vm.
Also simply running the vm as omvf instead of bios (requires reinstall) might work.

 

[Sir-robin10]

Okay, so I have it working on a windows server 2019 VM, works like a charm, but is it then possible to use the GPU (without using it in the VM) in a container? Or is that simply not possible at all? (It would be really sad)

 

[mbosma]

You should be able to use the gpu in a container when it's not passed through to a vm.
https://www.passbe.com/2020/02/19/gpu-nvidia-passthrough-on-proxmox-lxc-container/

 

[Okeur]

I had this issue and solved it by passing pcie=1 to the conf file... Could not explain why but it did the trick.

 

[Veeh]

try also with the flag: x-vga=1

 

[Sir-robin10]

I have followed this tutorial:
https://www.passbe.com/2020/02/19/gpu-nvidia-passthrough-on-proxmox-lxc-container/

HOWEVER, everyting works just fine, except, when I get to the reboot phase of the tutorial, after I get back into the node cli, there is an error already :§


I execute `ls -al /dev/nvidia*`, but this gives me the following error:

`ls: cannot access '/dev/nvidia*': No such file or directory`

Now, I have tried to install the drives and so on, but I never can access the GPU?? On the host not and not on the container...

root@node02:~# nvidia-smi
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Is what I get.

Further usefull information

root@node02:~# modprobe nvidia-uvm
modprobe: FATAL: Module nvidia-current not found in directory /lib/modules/5.3.18-2-pve
modprobe: ERROR: ../libkmod/libkmod-module.c:979 command_do() Error running install command for nvidia
modprobe: ERROR: could not insert 'nvidia': Operation not permitted
modprobe: FATAL: Module nvidia-current-uvm not found in directory /lib/modules/5.3.18-2-pve
modprobe: ERROR: ../libkmod/libkmod-module.c:979 command_do() Error running install command for nvidia_uvm
modprobe: ERROR: could not insert 'nvidia_uvm': Operation not permitted
root@node02:~# lspci -k | grep -A 2 -E "(VGA|3D)"
26:00.0 VGA compatible controller: NVIDIA Corporation GP106GL (rev a1)
Subsystem: NVIDIA Corporation GP106GL [Quadro P2000]
Kernel modules: nvidia

 

[dr.dad]

I have followed this tutorial:
https://www.passbe.com/2020/02/19/gpu-nvidia-passthrough-on-proxmox-lxc-container/

HOWEVER, everyting works just fine, except, when I get to the reboot phase of the tutorial, after I get back into the node cli, there is an error already :§


I execute `ls -al /dev/nvidia*`, but this gives me the following error:

`ls: cannot access '/dev/nvidia*': No such file or directory`

Now, I have tried to install the drives and so on, but I never can access the GPU?? On the host not and not on the container...

root@node02:~# nvidia-smi
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Is what I get.

Further usefull information

root@node02:~# modprobe nvidia-uvm
modprobe: FATAL: Module nvidia-current not found in directory /lib/modules/5.3.18-2-pve
modprobe: ERROR: ../libkmod/libkmod-module.c:979 command_do() Error running install command for nvidia
modprobe: ERROR: could not insert 'nvidia': Operation not permitted
modprobe: FATAL: Module nvidia-current-uvm not found in directory /lib/modules/5.3.18-2-pve
modprobe: ERROR: ../libkmod/libkmod-module.c:979 command_do() Error running install command for nvidia_uvm
modprobe: ERROR: could not insert 'nvidia_uvm': Operation not permitted
root@node02:~# lspci -k | grep -A 2 -E "(VGA|3D)"
26:00.0 VGA compatible controller: NVIDIA Corporation GP106GL (rev a1)
Subsystem: NVIDIA Corporation GP106GL [Quadro P2000]
Kernel modules: nvidia

did you find a resolution for this?

 

[Sir-robin10]

did you find a resolution for this?

I have!

The main issue here was that the GPU slots needed to be "separated" thus I mean that the iommu-groups must be separated. I bought an more expensive MOBO with more iommu groups and it works fine now.

 

[FluxSine]

This is what I am looking to do. So you were able to get a P2000 to pass through to a Windows 2019 server VM with no issues?
Any advice for this setup?
I was planning on using SuperMicro 6028U-TR4T and throwing the P2000 in there. Thoughts?