====== Power delivery ======
Tesla M60 uses 8 pin EPS connector with "Molex Mini-Fit" plug. **It is different than PCIe 6+2/8 pin plug!** It's actually the same as 8 pin CPU connector on motherboard.
**Do not force PCIe 6+2 plug into Tesla M60!** Look what might happen if you do: https://www.reddit.com/r/homelab/comments/uqfq69/psa_nvidia_tesla_cards_do_not_use_the_same_power/
====== Installing drivers on Debian 12 ======
> Based on https://docs.nvidia.com/datacenter/tesla/driver-installation-guide/debian.html
> Installed and tested as of 23-05-2026
===== Install kernel headers =====
# apt install linux-headers-$(uname -r)
===== Enable contrib repository =====
In ''/etc/apt/sources.list'' add ''contrib'' keyword to debian repository, update cache with ''apt update''.
===== Network repository enablement =====
> Replaced ''${distro}'' with ''debian12''
# wget https://developer.download.nvidia.com/compute/cuda/repos/debian12/x86_64/cuda-keyring_1.1-1_all.deb
# dpkg -i cuda-keyring_1.1-1_all.deb
# apt update
===== Pin Tesla driver version =====
> To find which version you should use, go to https://www.nvidia.com/en-us/drivers/ and select proper entries in drop-down menus, to select your nvidia tesla model. Then, click on find, and note the supported version. In my case, it's ''580.159.04''
> If you can't install the package with specified version, try looking at what packages are available using ''apt search nvidia-driver-pinning-*''
**This step is not optional!** By default, apt will install newest driver (with highest version number), which might not be compatible! In such case, remove those packages with ''apt purge nvidia-driver-cuda nvidia-kernel-dkms'', pin version, and install those packages again.
# apt install nvidia-driver-pinning-580.159.04
===== Install drivers =====
> This command install proprietary kernel modules for compute-only drivers. If you want to use other options, see https://docs.nvidia.com/datacenter/tesla/driver-installation-guide/debian.html#driver-installation
# apt -V install nvidia-driver-cuda nvidia-kernel-dkms
===== Finish =====
Your Tesla should now be visible in ''nvidia-smi''.
Sat May 23 18:20:54 2026
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.159.04 Driver Version: 580.159.04 CUDA Version: 13.0 |
+-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 Tesla M60 Off | 00000000:30:00.0 Off | Off |
| N/A 67C P0 42W / 150W | 0MiB / 8192MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
| 1 Tesla M60 Off | 00000000:31:00.0 Off | Off |
| N/A 66C P0 40W / 150W | 0MiB / 8192MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| No running processes found |
+-----------------------------------------------------------------------------------------+