It ran 4xMSAA last year, now it was FXAA (that would most likely allow single gtx580 to run it too). That removed memory bottleneck fermi had with the demo.. And i bet there has been a lot of other optimizations.
Basically, i cant see how it could be used other than to say, "it most likely is more powerful than fermi" but it cannot be used to even get idea how much.
Bookmarks