Eh. Dislike dissecting people's posts sentence by sentence, but I couldn't resist, so here goes:
X800 has more pure fillrate ?
X800 ? 475 x 12 = 5700
6800GT ? 350x16= 5600
By 1.7%. It's a factor, but only one you should take into account if the cards come out equal in every other respect, and even then the color of the packaging might have more significance. A slight difference in architechture or driver efficiency or a slightly better overclock by a few MHz from either card will make it entirely irrelevant.
Sites like FS shouldn?t be running the 6800?s with the hacked down Brilinear.
Then sites shouldn't be running the X800's with their hacked down brilinear either (which is tough, seeing as ATi doesn't provide a means to turn it off). News flash: both ATi and nVidia have optimizations for both trilinear and anisotropic, and neither is noticeable to the naked eye without artificially coloring mipmaps or high magnification. Nothing you'd ever notice during a game, certainly less than the gain in FPS they provide. (Go
here. You can open up a lossless PNG image by clicking on the JPEGs. Open them in seperate windows/tabs with the image in the exact same position on each, then switch between them if you need convincing. There. Is. No. Difference.)
The only place the GT has an advantage over the X800pro is OGL. Problem is 90% + games are DX.
It's more like this: At OGL, the GeForces demolish the Radeons pretty much without exception. At DX they're even, with the win going either way depending on the game. (Assuming their fillrate is close as with the 6800GT vs. X800 Pro; with its 30% higher fillrate the X800XT will beat the 6800U pretty much always.) Although, the Radeons' weakness at OGL seems to be a driver issue, so it's possible that they'll catch up if ATi improves them, but I wouldn't count on it.
The X800pro is faster where it really counts -- DX9, AA/AF, and shader intensive games.
One out of four is valid half of the time: The GeForces suffer a disproportionately large performance loss at AF when they have their optimizations turned off and the Radeons don't. To be fair, they probably still lose a bit more than the Radeons, but it's not a huge difference. (If you'd care to back up the DX9, AA, and "shader intensive games" parts of that with any examples not from HardOCP, I'm willing to listen.)
where NV hasn?t had a chance to ?hack/cheat? reduce IQ somehow
:/
Both parties are guilty of this, whether intentionally or not (most likely not, I'd say they're just driver bugs). I won't cite examples from nVidia since you no doubt know them by heart. From ATi I will mention the disappearing shadows in Splinter Cell and Far Cry.
Since what you probably meant to imply was optimizations, I'll address that as well. It's funny, you see, because it's precisely the other way around: current games are not optimized for the GeForce 6 series, since they were developed long before they were released. I understand they have quite a bit of latent potential by optimizing to execute multiple instructions per clock (per pipe, per whatever makes up parts of the pipe, etc.), so it'll be interesting to see how much they gain once games *do* start optimizing for them.
I?m also betting that the X800pro could prove to be substantially faster in some of the new shader intensive games than the 6800GT (early indications) . So the better shader performance of the X800pro will probably win out
Link?
the X800 also has better AA -- a usable 6AA.
I'll give you that. Although I've heard that nVidia is preparing a 4xMS 2xSS form of 8x AA for a future driver which should considerably increase performance for that mode.
@ GeneralGrievous:
Actually a usuable 12x AA effective with temporal AA coming up.
Hmm. Is that possible? IIRC, temporal AA works by choosing reference points (or sample points? dunno what they're called) at random out of those available. Thus, it can simulate an already available mode of AA faster, such as 4xAA with 2xAA performance (with various caveats), but can't make up entirely new reference points to be able to do 12x AA like you mention. I don't know the specifics of this, though, so correct me if I'm wrong.