![](https://blog.julienarnaud.fr/media/website/dsc_0069.jpg)
Machine learning desktop
I built a new computer to play with some machine learning.
![](https://blog.julienarnaud.fr/media/posts/7/IMG_20240206_201241.jpg)
It uses what seems to be the best scoring case for airflow and temperature : Fractal Torrent Compact.
The full setup is :
- Fractal Torrent Compact (2x 180mm fans)
- MSI PRO B650-S
- Ryzen 7900 (non-X, for its 65W TDP)
- 192 GB DDR5 (4x48GB Crucial)
- NVIDIA 4090 Founders Edition
- Seasonic 750W 80+ Platinium (with its peak efficiency at ~400W)
- Samsung 990 Pro 1 TB
- Samsung 970 Pro 1 TB (recycled from my previous 5800X desktop)
![](https://blog.julienarnaud.fr/media/posts/7/2024-02-06-19-26-19-075.jpg)
The 4090 is capped to 65% TDP as going any further quickly hits drastrictly diminishing returns (anything above 75% should be considered OC). When capped, the card only runs at about 60° ! That thing really is a crazy object and product.
![](https://blog.julienarnaud.fr/media/posts/7/Screenshot-from-2024-05-28-17-08-21.png)
During training the computer draws about 400W at the wall. Which sits perfectly in the sweet spot of the PSU.
![](https://blog.julienarnaud.fr/media/posts/7/cold1.png)
Since then I have replaced the CPU cooler by the same model, but in black.