Special function unit performance in future GeForce

Accelerated Computing CUDA CUDA Programming and Performance
1

I develop an GPU software of which sin function is a hot spot.
So, a growth of SFU performace of GeForceis a matter of concern.

I have two question on performance of SPU.

1.Is SPU performance of GeForce same as that of Tesla like single precision unit,
or worse as double precision unit?

2.In regard to double precision operation,
NVIDIA make GTX 680 worse than GTX 580.
How about Special Function Unit?

Thanks,

2

P.S.
As a developer, I purchased GeForce TITAN, but not every user has such a monster GeForce.

3

Don’t know anything about future GPUs, but your other questions are answered in the Programming Guide.