What makes me concerned is that they benchmarked their 480 against the 5870... And in a self-made benchmark, which obviously means it's cherry-picked and slanted. Given that the 480 is gonna probably run about $600-750, (given the fact that they're paying for 12 GDDR5 chips plus one of the most massive GPUs ever made) they're gonna have to have it comparable to the 5970, not the 5870, in order to justify the card's existence.
And as others have mentioned, the fact it's so late out the gate leaves it in quite a vulnerable position; if the benchmarks suggest the 5870 is within striking distance of the 480, AMD will surely be able to push out a refreshed 5890. Even with the 280, which found itself under fire before it even hit the market, had to endure for half a year before nVidia was able to boost their lineup with a revision. And as both G92 and GT206 have shown, it'll be impossible to use straight GF100 to make a dual-GPU board. And while GT206 was able to come out in only 6 months, that was mostly due to the fact that GT200 came out using the already-antiquidated 65nm process; this time, it's already at 40nm, so the 470 and 480 will have to hold the fort until someone OTHER than Intel masters 32nm.
Of course, at least we can all take comfort in knowing that tesselation is "The Future" and the "Killer Ap" for Direct3D 11... Just like it was for D3D 10.1, 10.0, 9.0c, 9.0b, 9.0, 8.1, and even D3D 8. Just shows that the more things change, the more they stay the same. (and ATi/AMD's been JUST as guilty of this as Nvidia)