Hi guys! I swear I have searched first!
What I am doing wrong?
I have a Dell R730, 2x Xeons 2698 v4, 40 physical cores in total, 128gb RAM, and 2 Tesla P40s, which are vGPU capable, indeed.
PVE 8.3.0
Drivers 16.4.
Win10 Enterprise IOT(not in IoT mode) and Win10 Pro for WS.
Using the Proxmox vGPU – v3 script (https://wvthoog.nl/proxmox-vgpu-v3/#comment-3354) script I was I was actually able to have BOTH of them working in vGPU pass-through. One was divided in several Q profiles for Windows VMs – which also worked without hassle, and one was given with its whole 24GB Q profile to another Win10 VM for the test – works without issues. So, both of them working as vGPUs on one physical machine is doable.
Problems and questions, that I need your kind help or suggestions with:
#1 I cannot get not matter which one of these GPUs (vGPU mode) working on Ubuntu or Fedora?
Gnome or KDE, doesn’t matter the method used – manually (as suggested from wvthoog.nl) with the drivers from there, or with the drivers from the google apis repository, or from the official linux repositories. It doesn’t matter which one of the GPUs in what Q Profile or configuration. I can see the devices in the VMs (lspci) and everything I am downloading the headers and everything, but
#2 nvidia-smi says "nvidia-smi has failed because it couldn't communicate with the nvidia driver"
#3 on one of the occasions I virtio console show only black screen, should I be able to see the vm through virtio, of vGPU works?
#4 BUT! the strangest thing is, after it fails with the installation it breaks the configuration in Proxmox and I CANNOT see the GPUs as mediated devices any more and cannot assign them to VMs any more. This is very strange, cause the only changes that I make are on the VMs????
I would appreciate any help or suggestion what to search and where to read!
Thank you!
What I am doing wrong?
I have a Dell R730, 2x Xeons 2698 v4, 40 physical cores in total, 128gb RAM, and 2 Tesla P40s, which are vGPU capable, indeed.
PVE 8.3.0
Drivers 16.4.
Win10 Enterprise IOT(not in IoT mode) and Win10 Pro for WS.
Using the Proxmox vGPU – v3 script (https://wvthoog.nl/proxmox-vgpu-v3/#comment-3354) script I was I was actually able to have BOTH of them working in vGPU pass-through. One was divided in several Q profiles for Windows VMs – which also worked without hassle, and one was given with its whole 24GB Q profile to another Win10 VM for the test – works without issues. So, both of them working as vGPUs on one physical machine is doable.
Problems and questions, that I need your kind help or suggestions with:
#1 I cannot get not matter which one of these GPUs (vGPU mode) working on Ubuntu or Fedora?
Gnome or KDE, doesn’t matter the method used – manually (as suggested from wvthoog.nl) with the drivers from there, or with the drivers from the google apis repository, or from the official linux repositories. It doesn’t matter which one of the GPUs in what Q Profile or configuration. I can see the devices in the VMs (lspci) and everything I am downloading the headers and everything, but
#2 nvidia-smi says "nvidia-smi has failed because it couldn't communicate with the nvidia driver"
#3 on one of the occasions I virtio console show only black screen, should I be able to see the vm through virtio, of vGPU works?
#4 BUT! the strangest thing is, after it fails with the installation it breaks the configuration in Proxmox and I CANNOT see the GPUs as mediated devices any more and cannot assign them to VMs any more. This is very strange, cause the only changes that I make are on the VMs????
I would appreciate any help or suggestion what to search and where to read!
Thank you!