Hi all,
I am running a PVE enviorment with an i5-10400, 64gb ddr4, 1x Radeon rx 6600, gtx 1080ti (GPU in queston), and storage.
I have been running a Windows VM with the 6600 for almost a year now with no issues. I set up IOMMU and blacklisted the drivers and everything works great. I recently purchased a 1080ti because I was discussing my server with him and he got to talking about the AI docker container he runs using Ollama. I thought it sounded really neat and wanted to set one up of my own. I tried with the 6600 however Ollama does not support it. I went out bought the 1080ti along with a larger PSU that would support running two GPUS. I plugged everything and when I booted up my PVE enviorment I saw that it recognized the 1080ti and it was also in its own IOMMU group so I would not have to worry about splitting a group. I tried setting up a new Ubuntu server VM so that I can install Docker and Ollama on it and was able to spin up the server fine without the GPU enabled. I then turned off the VM, added the 1080ti as a pci device (making sure to enable PCIE), turned off the virtual display output and then attempted to start the VM. it appeared to be loading for all of one second until I got an error saying
. I have been using chaptGPT for past 2 hours trying to get things working because I cannot find any articles related to what I am trying to do. Everything I find regarding GPU passthrough is for Windows or some other VM that has a graphical output. Since I am just using the GPU for AI I do not need the VM to have a GUI.
I was trying to follow https://nopresearcher.github.io/Proxmox-GPU-Passthrough-Ubuntu/ however I get stuck when they say "Add the repository for the nvidia driver. I tried using the standard repos however, those drivers did not work." because I cannot boot into the VM to install the drivers.
If anyone has any ideas I would be super grateful.
I am running a PVE enviorment with an i5-10400, 64gb ddr4, 1x Radeon rx 6600, gtx 1080ti (GPU in queston), and storage.
I have been running a Windows VM with the 6600 for almost a year now with no issues. I set up IOMMU and blacklisted the drivers and everything works great. I recently purchased a 1080ti because I was discussing my server with him and he got to talking about the AI docker container he runs using Ollama. I thought it sounded really neat and wanted to set one up of my own. I tried with the 6600 however Ollama does not support it. I went out bought the 1080ti along with a larger PSU that would support running two GPUS. I plugged everything and when I booted up my PVE enviorment I saw that it recognized the 1080ti and it was also in its own IOMMU group so I would not have to worry about splitting a group. I tried setting up a new Ubuntu server VM so that I can install Docker and Ollama on it and was able to spin up the server fine without the GPU enabled. I then turned off the VM, added the 1080ti as a pci device (making sure to enable PCIE), turned off the virtual display output and then attempted to start the VM. it appeared to be loading for all of one second until I got an error saying
Code:
error: start failed: QEMU exited with code 1I was trying to follow https://nopresearcher.github.io/Proxmox-GPU-Passthrough-Ubuntu/ however I get stuck when they say "Add the repository for the nvidia driver. I tried using the standard repos however, those drivers did not work." because I cannot boot into the VM to install the drivers.
If anyone has any ideas I would be super grateful.