Those modified V12 builds were attempts at the introduction of Fermi to make something work, without a lot of clear reference available on Fermi compatibility. Since then there has been volumes of information, and we have universal fermi/preFermi apps in extended beta testing & ongoing development. Until some more refinements are made, you should continue using the stock fermi app.
There are no Cuda development material references to the 460's architecture (2.1 compute capability) and at least one obscure but annoying bug we found & isolated in Cuda 3.1 ... while Cuda 3.0 won't use all the capabilities of that card yet ... so we're in for a waiting game
Jason
