nvidia shader compiler details

Status
Not open for further replies.
IST said:
How is Quack a cheat when they fixed it and the performence was the same? Edit: IIRC, wasn't Quack originally a bug fix of some kind for the R100? Edit2: Forgot a word. Fixed.
The application detection was there because Quake3 was a benchmark: the game was optimized for specifically. This became painfully obvious when a bug in ATI's Quake3-specific code caused decreased image quality.

This had the exact same effect that many people are currently complaining about with the NV3x: benchmarked performance is lower than real-game performance (particularly if the games you play aren't one of a few specific benchmark games...the suite of which was less numorous at that time).
 
You are naive (to say the least) if you think nvidia's benchmark optimisations started with nv3x. I think we’ve heard comments here about 3dfx complaining about it.

Other companies have to do similar things because nvidia have been forging the way in these types of things for years beforehand. They moan about ATI "cheating" now, but if they were then they probably wouldn't have a benchmark to stand on.
 
You are naive (to say the least) if you think nvidia's benchmark optimisations started with nv3x. I think we’ve heard comments here about 3dfx complaining about it.

Of course they were cheating before the NV3x. Need we bring up the 40.72 drivers? They introduced the 3DMark2001 Nature Clip-Plane cheat, which increased the Natures scores on all DX8 hardware by 25 to 40 percent.
 
Chalnoth, I was under the impression that the "optimization" that caused the Quack thing was in the drivers way before the R7x00 and R8x00 were even released. I've pretty much gotten the picture that it was a fix for an earlier board that broke with the later models.
 
Status
Not open for further replies.
Back
Top