Quote Originally Posted by saaya View Post
idk... do you think that makes a big diff?
doubt those transistors were used a lot to begin with, otherwise nvidia wouldnt have cut them off, cause they want more perf, not less with gf110...
lopping off 200M transistors is probably a lot simpler than what they actually did to fermi. maybe they found a way to do the same thing with less logic? i dont really know and there is very little information available that can offer insight into that subject.furthermore, the number of gates is not a good measurement of area because speed affects transistor size. if fermi were designed for half of the original clockspeed it could be half the size and an even smaller fraction of the power. there are so many other things i could list that they might have changed but they would only go unanswered.

i have a hunch that the power circuit was improved. the ammeter and current limiter allows them to use a more efficient power circuit because peak current will be much lower. intel did something similar with montecito. it has 64 clockspeeds which saved them a lot of power when performance wasnt needed.
not used a lot = dont add to power consumption...
with leakage it does not matter if the transistors are being used or not. the power is still being consumed. dynamic power involves switching which is affected by activity. i would bet that the majority of g100 and gf110's power consumption comes from leakage, and more so than other chips.