Originally Posted by netllama
The problem that you reported is a hardware limitation. On most GeForce cards, only the primary CRT controller can do dual link TMDS, which is what is needed for a DFP to do 1600x1200.
CRT's do not have the same limitations as DFPs in terms of specific modes.
I would have guessed that too, but DFP-1 is being reported as using the "Internal Single Link TMDS" and yet it still works. If it is a dual link TMDS it should be reporting a higher pixelclock than 135MHz as well. Also, when the LCD is the only output device and plugged into the primary controller, the EDID specified value works correctly.
I'm still a bit confused over this. Both the primary and secondary controllers are reporting using an Internal Single Link TMDS operating at 135MHz. If the LCD is plugged into the primary controller and a CRT is plugged into the secondary controller things act oddly (why would changing the CRT's modeline affect the DVI output?). If the LCD is plugged into the secondary controller, everything works as it should.
It seems like more is going on here than just a difference in the TMDS configuration on the card. There appears to be some relationship between the primary and secondary controllers and the primary and secondary displays as reported by the drivers. If they don't correctly match (IE the LCD is connected to the primary controller, but the drivers still choose the CRT as the primary display device) things start behaving in an undefined manner.
Edit: I didn't notice you said the primary controller is the one with dual TMDS. It was only when I plugged the LCD into the secondary controller that it worked.