thats an estimated max GFLOPs. you'll end up with about 10% of your estimate which it seems you are doing with your cards

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

Well I sure many people will disagreed, but in my observations I just found a very handy coincidence when estimating a CUDA card RAC. That will work only on old BOINC clients, like 6.6.20, wich I use on most of my machines.

At the moment, I ran this CUDA cards, wich have pretty much this RAC:

2 9600GT cards, that have an estimated 37GFlops and have a RAC of 3.7k
3 GTS 250 cards, that have an estimated 86GFlops and have a RAC of 8.6k
One GTX 295 (until tomorrow hehehe) card, that have an estimated 2x125 GFlops, and a RAC of 25k

Any one else can confirm this with other cards?
Its a nice coincidence, and maybe it can be used to have a close estimated RAC on new/unknown CUDA cards.

---