we really all are going to make it, aren’t we? 2×3090 setup.

“`html A British user has successfully run a local AI setup using the club-3090 model, achieving impressive performance with high prompt processing…

By AI Maestro May 14, 2026 1 min read
we really all are going to make it, aren’t we? 2×3090 setup.

“`html

  • A British user has successfully run a local AI setup using the club-3090 model, achieving impressive performance with high prompt processing speed (PP/s) and throughput (TP/s).
  • The user reports that this setup is now running on Ubuntu as a dual-boot alongside Windows for better system integration. They are seeing rates of over 4000 PP/s and 113 TP/s without the use of NVLink, indicating significant improvements in local AI performance.

“`

### Takeaways
– **Improved Local AI Performance**: Users can now run high-performing AI models locally with impressive throughput.
– **Compatibility Across Operating Systems**: The user successfully ran this on Ubuntu alongside their existing Windows setup, demonstrating its versatility for different environments.
– **Future Potential**: With these results, the user is considering moving to a more powerful hardware configuration to explore whether smaller models can reach frontier intelligence in the near future.

Scroll to Top