http://bugs.winehq.org/show_bug.cgi?id=7284
--- Comment #48 from Tobias Jakobi liquid.acid@gmx.net 2009-05-04 09:32:38 --- I don't see where this information from RCC can help us here.
I thought that we agreed that the problem is how NaN are mapped when written to the color output register. As far as I can see rcp doesn't clamp his result. That's in agreement with the D3D8 docs.