Really, I think that you give the GTX 400 series a bit TOO much benefit of the doubt. Especially when it comes to the GTX 470, given that it often LOSES to the 5850, seemingly more than it wins, it makes it a head-scratcher at $60US more; how could it get an honorable mention for slightly inferior performance with a ~20% higher price? And when the 470 does win, it tends to be a very narrow margin; compared to an often-wide margin when it loses.
The GTX 480 isn't quite as bad off; it USUALLY wins over the 5870, but then again, it can be quite questionable to ask for the nearly $100US pemium over the Radeon; certainly a more questionable premium than jumping straight to the 5970, which at times seems to offer better price/performance than the 480.
Similarly, it makes it another head-scratcher that you went out of your way to list the 470/480 on separate, higher tiers than the 5850/5870. With the former of each pair, the 470 most certainly CANNOT be argued to be outright superior; at best it is really on a par. As for the latter, while the 480 shows itself to be typically stronger than the 5870, It's relatively small margin most of the time.
[citation][nom]neiroatopelcc[/nom]Question about graphics.....Do you think we'll see non-flat floors in a game any time soon? tesselation is for non important stuff right? and bump mapping doesn't properly work on floors - even the crysis 2 screenshots very clearly show a flat floor with objects placed on it. I'm soo looking forward to the day where the geometry doesn't feature painted on grass or shader grass on flat surfaces (that was good in oblivion, but that's ages ago).[/citation]
First off, Oblivion's grass wasn't all that complex an implementation; it simply was a procedural-generation routine that on-the-fly populates the ground with grass "clumps," with the chance, color/texture, and size of them appearing determined by the ground texture(s) used for that vertex.
As far as the floor surfaces, what they look like is largely determined by the programmer; if one wanted, one could readily apply paralax mapping (or even a more complex shader) to ground surfaces, it's just that they typically don't. I honestly don't quite know why, (perhaps Cleeve would) but my two hyptheosis are that it could be a design choice to distinguish "walkable" terrain, or that it could be a limitation of the terrain-heightmap subsystem. (possibly calculating it over very large surfaces?)