Quote Originally Posted by felix_w View Post
Is there somewhere an updated compatibility chart ? The link at gpugrid's page is blocked.
The chart in the first post is the most up to date they have, but there's also this by skgiven:

My opinion, listed by performance in each row, but also based on running cost. Some common (non oem) cards.

Highly Recommended:
Geforce GTX 590, 580, 570, 480, 560 Ti (448), 470, 465
Tesla* C2050/C2070 (Workstation), M2050/M2070, S2050 (Data Center)
Quadro Desktop* Quadro 6000, 5000, 4000, 2000, 2000D

Recommended:
Geforce GTX 295, 280, 275, 560Ti, 260-216 (55nm), 460, 550Ti, Tesla10, Tesla20, 460SE, GTS450, Quadro 600, 480M (CC2.0)*, Quadro Mobile* 5010M, 5000, 4000, 3000, 2000

Note on the GT240. This card presently works on most setups, but it does not perform as well with the more recent 6.14 and 6.15 apps. It is expected that CC1.1 and CC1.2 cards (such as the GT240) will no longer be usable whenever a CUDA 4.1 app is introduced. The GT240 has therefore been moved off the recommended list. Ditto for other CC1.2 cards such as the GT340.

The GT240 was a good card for the earlier 6.05 and 6.12 apps. These were replaced with the 6.13 app and then the 6.14 and 6.15 apps because Fermi cards would not work on the earlier apps. Apps tend to be upgraded to introduce support for new cards and improve performance for high end cards. The project can only support 2 apps.

*The Tesla and Quadro cards are very expensive compared to the GeForce cards, so these are only really applicable to business users.
Any mobile cards are recommended rather than highly recommended because they could cause heat issues with the notebook - it's up to you to determine if you want to use them or not. Only CC2.0 cards are Highly Recommended.

The Highly Recommended cards are most suitable for the long tasks. Other cards will also run them, but you should reconsider your system usage and Boinc cache settings.
Non-recommended cards should not be used with long tasks.