LOL, using a percent to demonstrate 1 FPS increase is absurd. Showing the actual difference in FPS between cards would have been more clear. Simple to those who are not easily offended by the results of the new 260 so far
As far as 8xAA goes, it's what they used. I can only assume they did it to see if the new 260 offered better performance when more AA was used. So far, none of the results show 2 FPS or better between the old and new 260. To me, those results are simply within the margin of error and doesn't demonstrate any real advantage for current 260 owners.
Actual fps were shown. Difference in fps is misleading. As I said 1-2 fps when the baseline is 15 is "a lot", around 7-13%. You will find a hard time OKing most HD48xx cards so much, foe example.
On the other hand 1-2 fps when you are at 100fps is close to nothing. In the review with more than 30 different result pairs, fps's were low, thus the percentages were around 4-6%, always in favor of the new card (not in one of the pairs the old GTX260 won) AND they were consistent across the whole table.
- if there's one fps of difference AND sometimes it's in favor of one of the cards and other times in favor of the other = margin of error
- on the other hand if the difference is always in favor of one of the cards and the percentual difference across all the results is consistent = actual performance difference.
Nowadays graphics cards performance is very consistent across all ranges of fps (from very low up to very high), until a severe bottleneck is found. What I mean is that a 5% difference when on low fps, effectively means the card will have that same difference on high fps situations. You can see that on any review. With a little margin of error, and some few exceptions, the "rule" works almost always.
EDIT: BTW, no one is offended, and anyone with 2 neurons wasn't expecting much more of the new card. See, number of SP/TMU being the only difference, only a difference of 12% was expected in the best case scenario, AKA ROP demanding free scenarios. Link doesn't work now, but when AA was dissabled there were up to 8-9% differences. With 8xAA and high res, the difference was almost always above 3%. You can't ask for more.