Honestly, want to say HAH! to Nvidia, but its hard to tell, maybe timedemo's(all to often from Nvidia supported games) are biased towards sections that are best on Nvidia hardware, but then maybe these guys found the only sections to run well on AMD cards.
Its very hard to tell and the inherant inconsistancy in real world gaming makes performance difficult to measure, but then, maybe more realistic compared to a benchmark/timedemo.
Thing is, I think its been clear since day one, the 480gtx isn't worth minimum 33% more than a 5870, never has been and never will be. A 10% shift in performance either way won't change that, the price difference was higher than 33% to start with anyway.
However, the most important thing TO ME, in terms of the actual competition as opposed to how well these cards do for end users, largely because I'm very interested in the future of the tech and the market, is the fact that at £400 Nvidia is not making anywhere near the profit AMD is at £300 for a 5870, and likewise Nvidia is almost certainly making a loss on 470gtx's and probably 465gtx's though its very hard to know how many are even being sold of each and what yields are per wafer which changes things dramatically.
Long term Fermi, profits/sales/reputation wise is a freaking disaster, though again I'll point out, if the 480gtx was £250, I'd probably buy one, even with the silly power usage, of course if it was that price, AMD would drop the 5870 to £250 and at the same price, I'd take the lower power usage and easier to manage card which becomes even more manageable in crossfire/trifire should I want to.
What I've been saying since DAY ONE of Fermi though is, drivers are UP TO DATE and mature, theres not a huge amount of performance to gain. Fermi is NOT NEW and was NOT NEW at launch, it was a 6 month old product with the driver team working BEFORE they had hardware, because thats what driver teams do, they had hardware as early as what, July last year, they don't need 50k cards to write a driver and check the driver works or work on optimisations, they don't need a single one to improve the driver, they need a single one to check real world non theoretical performance and find real world shortcuts to better performance. They've had exactly as long as AMD to make a driver for their top card.
AMD's architecture, by design, is FAR more complex to get the performance out of so realistically(as with 10.4) theres more scope for driver/performance increases on AMD hardware than Nvidia at this point, though fixes for new games/broken sli/broken xfire and the likes really don't count for performance improvements, they are performance fixes.
