It keeps up here:
HEXUS.net - Review :: NVIDIA GeForce GTX 470 1,280MB review. Fermi done right? : Page - 6/15
Benchmark Results: Battlefield: Bad Company 2 (DirectX 10/11) : GeForce GTX 480 And 470: From Fermi And GF100 To Actual Cards!
I think what a lot of people don't realise is, though this probably doesn't affect most of them anyway, but all these benchmarks that are being put on the GTX480/470 are games that don't really stress the card where it counts. Put it in Heaven and you can see that.
The 5000 series is designed for games of today, not games of tomorrow. In raw pixel pushing power the 5000 series will win, but you put a true DX11 game with proper tessellation etc and you will see more of a slow down with 5000 series cards than with GTX400s, unless the calculations and pipelines are improved for DX11 within the 5000 series. Nvidia did gamble on it though, and the pricetag vs performance for tomorrows games isn't worth much to a lot of people. Metro2033 is the only DX11 game out atm that uses DX11 with proper tessellation. A few benchmarks explained they had to disable the "Advanced DOF" graphic setting because it brought every card apart from the GTX400s to an unplayable state.
And even with the temps, people just jump to conclusions. Most of the temps you see at 92C etc are ran in Furmark and let the drivers control the speeds. Two points here:
The drivers default to trying to keep the card quiet, instead of as cool as possible, but engineers have said the heat won't be a problem.
Furmark is more the worst case scenario for temps. Even ATI have said it's inaccurate at reporting temperatures and it made their cards look like they run really hot. If you use manual fan control, and don't mind a bit of extra noise, you will get better temperatures in just normal gaming.
Point is really though, temperatures aside, in the long run the GTX400s will pull ahead.