I based my saying mostly on XBit Labs' review on ASUS Extreme N7800 GTX TOP, versus stock speeds X1800 XT.
The Extreme N7800 is over-clocked at 486/1350.
In Battlefield 2 at 1600 x 1200, with 4 x FSAA and 16 x Aniso:
ASUS Top - 72.8
X1800 XT - 85.7
In Far Cry -Research-, at 1600 x 1200 with 4 x FSAA and 16 x Aniso:
ASUS Top - 78.1
X1800 XT - 83.1
In F.E.A.R. Multiplayer Beta, at 1024 x 768, with 4 x FSAA and 16 x Aniso:
ASUS Top - 33 (max)
X1800 XT - 41 (max)
Etc.
No, of course, the X1800 XT isn't faster everywhere, in every game.
Is it a perfect card ? Hell no, far from that. Is any of the GeForce 7 series GPU perfect ? Nope.
Does nVidia offers perform better in OpenGL ? Sure thing, as usual.
Are the best known games rendered in OpenGL ? So far, no. Doom III ? Please ...
Am I saying that because it is more expensive that it's worth to buy a X1800 XT anyway, just because it's just a little faster in most situations ? Hell no.
What I'm saying is that, technically speaking, the X1800 XT is a faster solution. Not everywhere ... yes, I know. Not everytime, yes, I know. But the fact that it performs better just by 1 frames-per-second means it is, technically, faster.
And to be honest, to see that the GTX 256 MB version needs to be over-clocked to perform as well or just a little more than the X1800 XT shouldn't be taken into consideration.
Because no web-sites so far reviewed extensively a X1800 XT over-clocked at 690+ / 1600+ ...
P.S/ Here's the article in question from XBit Labs:
http://www.xbitlabs.com/articles/video/display/asus-en7800gtx.html
And I am sorry to turn this into a potential debate. It wasn't my intention. I thought most people, including the most-informed Rollo would know all this by now