The scores DO seem a little high.

It wasn't too long ago the [H] put up an article outlining the reasons and methodology behind their "Real" testing, as opposed to using canned benchmarks. As justification, it was shown how drivers can (seemingly) be optimized for specific benchmarks, rather than the game the player actually plays. By way of demonstration, they showed how the Crysis benchmarks were showing considerably higher framerates than could be reasonably expected within the game itself, to the point where a hardware setup that gives an acceptable benchmark may not actually yield a playable experience. Without getting into the gritty debate over whether that's true or not, the bottom line is that they were finding real world performance to fall short of the benchmarks. Whereas you seem to be getting numbers that are actually BETTER than the benchmarks, at the given settings.

So with all that in mind, I'm a little interested in what kind of a demo/situation you use to create you own numbers. Is there something distinguishing that would account for most of the difference? Just curious, now that it's been brought up.