SLI for CUDA? Is it possible?


Tried the forum search, and nothing came up (perhaps it was broken at the time, kept giving me errors.) My apologies if this has already been answered.

Is it possible to use SLI with CUDA? If so, how does it work? Or is it more headaches than it’s worth?

In light of Tom’s Hardware reporting that 2 GTX 460 cards in SLI blow away a single GTX 480 for all gaming benchmarks, it seems like a relevant question to ask.



SLI is ignored by CUDA. You see two CUDA devices and have to operate them from different host threads as usual.