Anand states that a lot of NV30 info around is inaccurate

or does the NV28 really have 20 more million transistors?

 

There's that, too. The fact that NV18 was supposed to gain 30 mil is what's throwing me off--that seems much more wrong than adding 20 mil on top of the NV25. Both seem uncharacteristic of nV, though, especially given the minute model number increments.

Maybe those numbers are also missing an "effective" qualifier. Either that or it's straight up PEBKAC.

 

Demo: I don't disagree with you. However, I just don't see the point. I can't imagine that a competitor would base any business decisions on PRish statements that carry no weight (i.e. I doubt some comment about 128bit bus being overkill is going to mean anything to NVidia's competitors--I doubt stage whispering "boy, tonight's a real bad night to attack the fortress" is what Sun Tzu meant when he talked about misdirection. Looks alot more like the Monty Python bunny than the Trojan Horse to me Then again, I've met some frighteningly moronic people in positions of corporate power, so I'll sit out this hand