Quote Originally Posted by GoldenTiger View Post
What I mean is, that even though the numbers vary ~5% or so (which is in margin of error for manual run-throughs and various setups), the pattern is the same: game A performs roughly X% better, game B performs roughly Y% better, etc. across all the reviews. For example, Far Cry 2 is a clear, large winner across reviews; Crysis is a very small bit better across them. We can assume that if a review fits inside of the pattern on how its games are performing, that it's pretty close if not spot-on. If for instance some review showed Crysis as 30% better when most of them are showing it as around 3-5%, we'd know something was up.
It ends up 11% different to anand/hexus, that's a fairly substantial margin of error there.

the only cause for this which someone has mentioned is they ran at 4ghz, which if this is the case, means the fermie is better at higher speeds.

But to blanket say 'it's margin of error' is taking the piss a little.