Well, i was a bit curious regarding vertex and fragment programs costs (regarding perfs)
I’m using a FX5200 and when enabling fragment and/or vertex program extensions i got something like 30% penalty… Ok this is not a scoop but… what about the hardware? Did Nvidia has kept both pieces of hardware for fixed pipelines functions & shaders?
In the negative then (fixed func emulated through shaders), why using fixed pipeline functions only is still faster than using vertex/fragment prg?
The fx5200 has both pipelines implemented in hardware but it has very poor pixel program processing power. Still I’d have thought it would not be as slow as you stated in vertex programs.
Yep, penalty was between 20%-30% (prior drivers).
Finally, Did Nvidia has planned to drop fixed pipe hw? is it already done with latest boards?
Does anyone know which ATI card is the first where vertex/fragmentprograms are not penalised? On my 9600 XT, fragmentprograms still seem to be treated alot differently than the ffp.
AFAIK all R3xx+ hardware don’t have a fixed function pipeline anymore, meaning everything goes through the programmable pipeline.
If there’s a cost then code better than you 
I tested it with a fragment shader that only stores a fragcolor and thats it. I also heard that the R3XX+ is not supposed to distinguish between a custom fragment shader and the ffp but for me, it wasnt the case ;(