Originally posted by jbirney
Can not happen. I agree real games is the best but forward looking? What happens when the game engine gets complete overhauled? Based on your statement I could look at how UT preformed and make a prediction based off that of how a card will work in UT2k3? Will a video card today runs JK2, SOF2, RtCW will give me an idea how it will perform in DOOM3? Heck you can not even look at the same engine and predict how different titles will react. Case in point. Frame rates for cards in Q3 do not carry over to RtCW, JK2, SOF2. For example the 8500 looses big time in standard Q3 benches to the GF4 ti4200. But its right on the heels if not ahead of a GF4 ti4200 in the other Q3 based games (RtCW, JK2, SOF2, ect). So even from one game to the next you can not really predict how things are going to run.
Synthetic benches give us a guess how cards might perform in future games. A good review will have real life bench and a few synthetic thrown in. You should never base a review off one or the other.
That analogy applies exactly the same with 3dmark03, if the radeond9700 beats the geforceFX in 3dmark03 (or even if it was the other way around) means nothing to doom3 because their 3d engine is completely different.
Even games that share the same engine have different results. As you mentioned the radeon 8500 was slower than the ti4200 (and even the geforce3) in quake 3, but almost meet the ti4200 when running MOF and in sometimes even beat it in RTCW.
I think that benchmarking has become much more complicated, since its becoming much more difficult to have apples to apples comparison when you have so many different variables.