GTX295 "dual" GPU handling in CUDA

does anyone know how the GTX295 “dual GPU” shows up in CUDA? Does it show up
as 2 separate devices that need to be handled separately? And do the 2 GPUs
reference the same device memory address space, or separate device address spaces?

E.g. is there a benefit of having 1 GTX295 vs. 2 separate lesser GTX units…?

I am speaking regarding CUDA numeric computing, not graphics display.

And, does this “SLI” technology have any CUDA implications, or is
that only for graphics?

Thanks all!

two separate devices, separate address spaces, no obvious benefit to single GPU beyond physical constraints, SLI does nothing for CUDA.

(can you tell I answer this question a lot)

Thanks! An additional related question: Do the GTX260M or GTX280M mobile GPU devices support double precision (compute capability 1.3?)?

I couldn’t find this on the spec sheet in this web site. Actually I am just looking for any mobile GPU device that supports CUDA double prec.

I don’t think any mobile devices currently support double precision.