Skip to content

Instantly share code, notes, and snippets.

@presidentbeef
Last active January 11, 2026 02:29
Show Gist options
  • Select an option

  • Save presidentbeef/d538245f5372b29542cab5e7485ed588 to your computer and use it in GitHub Desktop.

Select an option

Save presidentbeef/d538245f5372b29542cab5e7485ed588 to your computer and use it in GitHub Desktop.
Using RTX 5070 on Framework 16 with Mageia
  • Do not try to configure the nvidia card, only the main AMD card
  • Install the Bumblebee package
  • Use -open nvidia driviers
  • Reboot
  • optirun glxspheres64 to test
  • If you get primus: fatal: failed to load PRIMUS_LOAD_GLOBAL - change /etc/bumblebee/bumblebee.conf to use Bridge=virtualgl (and restart with sudo systemctl restart bumblebeed)

Ollama:

  • Install nvidia-current-cuda-opencl package
  • Test with nvidia-smi

If you have Ollama set up as a service:

  • sudo systemctl stop ollama
  • Edit /etc/systemd/system/ollama.service - ExecStart=primusrun /usr/local/bin/ollama serve
  • sudo systemctl daemon-reload
  • sudo systemctl start ollama

Otherwise: primus run ollama serve

To check:

  • Start a model (e.g. ollama run mistral)
  • Run ollama ps - should show 100% GPU
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment