Hmm. It says "ATI x800 was the card of choice amongst many of the testers, as it ran roughly about 30% faster than Nvidia's best cards". That's a pretty ambiguous statement :/. Does it mean the X800XT is 30% faster than the 6800U and the X800 Pro is 30% faster than the 6800GT? Or just the former? I could easily imagine that being the case - on a PS2.0 codepath in Far Cry, the other DX9 game, the X800XTPE is 20-30% faster than the 6800U, courtesy of its much higher core clocks. On the other hand, the 6800GT is equal to better than the X800 Pro, and I see no real reason why this can't be the case in HL2, asides from Valve not wanting it to be (although I've heard that it'll utilize 3Dc, so that could be a reason; but I don't think that would make for a very huge difference, certainly not Doom 3 proportions).
So I don't really know what to make of the ATi being faster at HL2 thing. With the previous generation it's almost certainly true; GeforceFX is crap at DX9, so the R9x00s will no doubt outpace it by, err, a lot. On the other hand, asides from 3Dc, I see no technical reason why the GF6s couldn't perform just as well at HL2 as they do at Far Cry. All we can do is wait and see.