I am running ver 181.51 (and have run 181.20) drivers from laptopvideo2go as well as dell's 179.50 drivers on my quadro FX2700m and boinc 6.6.20. The GPU is crunching as designed. But, I noticed that boinc is reporting:
4/22/2009 11:18:45 AM CUDA device: Quadro FX 2700M (driver version 18151, CUDA version 1.1, 512MB, est. 22GFLOPS)


So why is boinc reporting CUDA 1.1 instead of 2.1?
AFAIK, the driver I am using is the 2.1 driver (or at least the 2.0 driver).

Since this is being reported by boinc and not seti or gpugrid, the compiler and drivers for the science application is not coming into play.

What do I need to do to take advantage of cuda 2.1?

So why is boinc reporting CUDA 1.1 instead of 2.1?

This is the maximum CUDA capability (compliancy) of your card, as detected by the drivers. You can still install later drivers, but they won't magically add the capabilities to your card.

Compare it to DirectX. You can run a DirectX7 compliant card very well with DirectX9 installed, but the card won't use all of Dx9's capabilities, only those of Dx7.

---

That makes sense, except that I can't find anything like that in the nvidia documentation I have looked at. The website lists CUDA enabled cards and does not distinguish between cards which are capable of the various versions. Additionally the specifications for the card state "CUDA enabled" while OpenGL and DirectX all list a version number. The CUDA 2.0 docs point to the list of CUDA enabled cards and do not mention a shortened list of cards.
My card has Compute Capability of 1.1, which is different than software drivers of CUDA 2.1.