Originally posted by StealthHawk
no, you are missing the point. 3dmark2001 was great becauseof the database so you could compare systems.
why is this true? because 3dmark2001 was a system benchmark that stressed the memory subsystem, the video subsystem, and the CPU.
now, this is NOT true of 3dmark03. it is a video card benchmark, nothing more, nothing less. the CPU makes such an insignificant contribution to the final score that comparing systems leads to incorrect and erroneous conclusions. other people have already posted this. i don't want to repeat them.
i mean, we have already had people say that some guy with a 1GHz system and a DX9 card scored 2-3x higher than someone with a DX8 card and a 2.4GHz system or something. and obviously the latter system is faster in reality, right?
edit: and if you still don't believe me, have a look at this thread http://www.nvnews.net/vbulletin/sho...=&threadid=7456 even the CPU test of 3dmark03 shows ridiculous results.
no I am afraid you don't understand the point at all. Future DX8/9 games are going to be much more video card dependent and less CPU intensive. You state 3DMark03 is a video card benchmark and not a system benchmark. This is exactly correct and exactly how it should be. If the nature test, which used DX8/9 features can run smoothly on an Athlon T-Bird 1.4 and a Pentium 4 2.8 then that should tell you something. It means that finally all those 110+ million transistors in the latest GPU's are doing what they were designed to do. I understand that the CPU still has to do some physics/AI etc. but that computational load is must less intense than shifting around billions of pixels. 3DMark03 is designed properly for the future of gaming and gaming cards.
If you want to benchmark and get and idea of how a system will play games from the last 3 or 4 years then use 3Dmark2001. If all you want to do is play DX7 games all day then stick with a GeForce3/4 or Radeon 8500.
You are using today's games as a reference when talking about 3DMark03 which is just plain wrong. Read the 3DMar03 white paper it explains all of this and a lot more. I believe they have researched and studied the situation very well, what they have come up with makes sense.
EDIT: and I just wanted to add that it is unreasonable and actually crazy to expect 3DMark03 to use the same methodology as 3DMk2001. Why? Because 2001 is based on a set of criteria for a much much slower generation of cards with much much less custom imagery processing. So what do you expect your new GeForceFX to do, push the pixels around in the exact same way as your GeForce 4 only at a higher clock rate? So why bother packing in all those shaders if they still rely on the CPU to do all the work anyway?