The Fermi Thread - Part 3

[highoctane]

Well you can subtract 851-479 = 372W more for the extra card, assuming nothing else was changed. That's at the wall of course... taking a 0.85 efficiency, that's 316W (at 0.80 efficiency, its 298W). So either way, the cards are drawing WAY over the 250W Nvidia advertised.

Holy moly!

Kind of an eye opener thats for sure, too bad TSMC doesn't have a high-k metal gate process on 40nm, 28nm should be high-k metal gate or at least be an option with TSMC later this year.

I'm wondering how much of a role leakage may be contributing with the high transistor count.