The real lesson learned is that don't ever buy a card based on a "synthetic benchmark" when anticipating a "particular game." Not because the benchmark isn't indicative of the card's eventual performance, but because the game's release and testing should be your final decision factor. Too many people got 9700's and 9500's on the idea that it would help them with HL2 or DIII. Thos cards are both pretty much "outdated," and may not be your best choice if HL2 was released a month from now.

I must admit I'd a tad guilty of this as I have a stupid "Get Half Life II free" coupon included with my 9600, however it's proved excellent for Far Cry so far.

I remember feeling so giddy getting this card, then as the months have flied by, I am left with a card that covers about 1% of what I use it for.

The biggest value 3dmark03 and 2001se were to me was to troubleshoot why my score was so much lower than people with similar systems. Turned out to be vertical synch on and AA and AF running at 2x.

For predicting future performance of video cards is pretty moot. A game like Far Cry already does this and even the 6800 Ultra was not amazingly ahead of the 9800XT (8 fps at 1024). Sure it rocked in the old games, but it appears that even the 6800 Ultra may not be too advanced to handle the new games just yets. My point is not to compare cards or companies. However to show that in 3dmark03, the improvements were very impressive from the 6800 Ultra to the 9800XT (nearly doubled), but not in Far Cry that better mimics future hardware requirements. The fps was better for the 6800 Ultra, but not $500 worth in my opinion.

Overall, my point is when the benchmark is released any card you buy soon after will not be that good when the game comes out that the benchmark is supposed to mimic.

Why is 3DMark even relevant anymore? *FP

Sounds fishy, like we'll see more drivers with pattern-matching "optimization" routines. I'd be more confident in NVidia and ATI if they spent less effort on hitting the product cycle and more on better drivers and more efficient fast hardware.

With all the recent talk of procedural textures and (now) dynamically generated shader programs, it sounds like momentum is building for a move away from static textures. Maybe we will see the recent insane increase in game size slow down (or even reverse)?

i thought a 3Dmark unsually follows a DirectX release
since MS isn't release a new DirectX untill a while later, what will this 3Dmark be based upon?