Unfortunately thats the exact attitude team Hard feels is the problem.
Over the last few years they have documented quite thoroughly the relative uselessness of benchmarks reflecting real world performance. With things such as companies posting drivers optimized to boost a 3dmark score but does absolutely nothing for game visual quality or fps.
Sure there is still a small place for benchmarks, but if you think just b/c one grfx card has a better 3Dmark score means it is somehow better than another then I have some of Jack's magic beans to sell you. Hard has extensively tested game performance compared to benchmark performance and the data doesnt lie, benchmarks are pretty rubbish.
Sure a 3Dmark score of 1000 on an Radeon 9200 is not gunna be as powerful as a 10,000 from a Radeon HD 3750 (for example) and this might help a noob not understanding that the larger model number in this case doesnt mean faster, but in general a higher model number means a stronger card, that is if the price doesnt give it away. But comparing a HD3750's 10k marks to a NV 8800 GTX that scores only 8500 marks doesnt mean crap when game settings can be set higher on the NV card, runner a faster frame rate, and even on similar quality settings the NV card renders a better image. In this case (which prevails the industry right now) 3Dmark means shit. (Disclaimer the card names/versions and scores were just pulled out my ass just to illustrate the point. Unlike where I had to pull names/numbers out my ass, Hard has years of documentation of precisely what I have put here.)
If you didnt know this fact, or are somehow disgruntled that I might have the audacity to post this, go over to Hard and educate yourself, or have fun living in the bliss of ignorance.
It doesnt change the fact that benchmark scores dont represent anything but an artificially generated number and doesnt give any significant reference between gameplay quality on cards running games like crysis, cod4 or WOW.