You haven't seen the recent benchmarks of the final product GTX 480 have you?
Straight from Nvidia, not speculations like you posted:
http://gizmodo.com/5487352/nvidia-gtx-480-takes-on-ati-hd-5870-in-benchmark-gauntlet
15% performance increase until the part of the uniengine uses a high level of hardware tessellation that no game known to man uses or will use. In hardware tesselation the GTX 480 wins by almost double the performance, in real life test scenarios around 16% max. Please note this is the UNIENGINE SYNTHETICS NOT A REAL GAME.
As I have mentioned even in one particular game (name escapes me) the max performance was 15%, with other games actually showing favor the the HD 5870 and others to the GTX 480.
GTX 480 is still well under an HD 5970's capabilities. It takes the single gpu card trophy, but looses the single card trophy which is what matters (remember GTX 295? Well HD 5970 is just that.).
Read this as well. Charlie receives various benchmarks from moles that are "friendly" to Nvidia thus received GTX 480 final product samples (meaning they are the same cards that will ship out to the public for market). Some saw 5% increase in performance average against the HD 5870. Notice there are NOT assumptions they are well verified benchmarks by Nvidia friendly hardware reviewers who cannot risk exposing real evidence for fear of been excluded as Anand and others have (as mentioned by Charlie)... Even if you say he's biased, you can't say he hasnt been right with everyone of his predictions starting from too optimistic of clock speeds, to fail yields, to heat/power consumption being incredibly high.
http://www.semiaccurate.com/2010/02/20/semiaccurate-gets-some-gtx480-scores/