I don't think this evolve (devolve?) into a war. I am seeing anand taking the high road on this one (am I using the saying correctly?).
Hard made some interesting observation about how supposedly both AMD and nVidia cheat in the canned benchmarks in crysis. But that doesn't exactly refute the point that relative performance remains very similar. Which was dereks whole point, which they failed to refute.
They also compared the correctness of the "canned" benchmark to their "natural playtime" benchmark of running through a single level, but it is a well known fact that there are huge differences between performance in different levels using the same cards. For all we know had they played through the entire game rather then one level it might have matched the canned benchmark (percentage wise) more closely then their one level benchmark.
I can never really understand hard's results.. how big a difference does it make when they bump one out of ten options from medium to high on the better card to get 2.5 fps less but a "better quality picture". How much faster (if its faster), in percent, is that card then? (so that I could calculate its relative value per dollar by comparing how much more expensive it is, in percent)
I do like reading various sources though, without hardocp i wouldn't have known about the crysis benchmark cheating... And there is value to be gleaned from seeing the results of all testing methods... this means I can expect much lower performance in crysis engine based games with the GTS 512MB I just ordered... so I learned something from reading that article...
It is something to note though (perhaps in an article), that both AMD and nVidia cheat so aggressively in crysis benchmarking is deplorable and misleading to people buying those cards based on the canned benchmarks results they read online. The relative power of cards remains similar (should I go with a 3870 or an 8800GT is answered well by anands methods) but the hard method gives a better idea of what playable resolution might be at (the anand result of 30fps at 1600x1200 is due to both companies cheating, neither is playable at anything above 1280x1024).
Anand trys and succeeds in showing relative performance of cards compared to eachother. (I can tell, in percents, how much faster is a 8800GTS 512MB then a 8800GT).
Hard trys and succeeds in showing realistic and subjective playable performance of specific cards with specific games, but without substantial comparison between the cards. (I can tell how fast the 8800GT will be in crysis, I can't tell how it will compare to the 3870)