Proxmox 8.3 installer has support for BTRFS mirror as a "technical preview", but out of the box the Proxmox installer does not actually create a fault-tolerant installation. In this post we go over how to sync the bootloaders using proxmox-boot-tool
The fan controller I bought from Amazon works, but I hate it's fan curves. What it needs is an over-engineered PID controller. In this project we'll be building a fan controller for my NVIDIA Tesla K20x using ESPHome.
I've had a little homelab for decades and I've been using Proxmox for many years now. I had a cluster, but I wanted to rebuild it so I migrated everything to a standalone server and then everything paused there. I want to pick up on that project, but this time I want to document it along the way.
Debian 12, on which Proxmox 8 is based, has the correct driver in the non-free repository. However, that driver doesn't support the CUDA Toolkit in the non-free repository so it must be installed via run file downloaded from NVIDIA's website. In this post we'll go over installing the NVIDIA Tesla Driver 470 and CUDA Toolkit 11.4.
Proxmox 8.3 installer has support for BTRFS mirror as a "technical preview", but out of the box the Proxmox installer does not actually create a fault-tolerant installation. In this post we go over how to sync the bootloaders using proxmox-boot-tool
Debian 12, on which Proxmox 8 is based, has the correct driver in the non-free repository. However, that driver doesn't support the CUDA Toolkit in the non-free repository so it must be installed via run file downloaded from NVIDIA's website. In this post we'll go over installing the NVIDIA Tesla Driver 470 and CUDA Toolkit 11.4.
I've had a little homelab for decades and I've been using Proxmox for many years now. I had a cluster, but I wanted to rebuild it so I migrated everything to a standalone server and then everything paused there. I want to pick up on that project, but this time I want to document it along the way.
By default, Stable Diffusion only works with CUDA 3.7 and above. If this project is successful, we will be able to run Stable Diffusion on an NVIDIA graphics card that only supports CUDA 3.5. Even using older graphics hardware is faster than CPU-only image generation.
