View Single Post
Old 07-09-03, 08:04 PM   #57
Nv40's Avatar
Join Date: Aug 2002
Location: everywhere
Posts: 2,216

well i finished reading the whole article...

its excelent.. wow! good job Digitlife... here is a quote that sums well everything in the article ..

Let's sum up the results of the 3DMark03. It's clear that NVIDIA prefers to camouflage the speed drops in the shader operation using certain tricks that boost the speed in the shader tests. The 3DMark03 is the most popular for today. But ATI is not only keeping pace in using such cheats but also lies by promising to remove them. Moreover, the RADEON 9800 stands farther from the reference than the FX5900 in the Game4. There is no doubt that the methods that let spend less resources on one or another mathematical function should be welcome. And if the guys at FutureMark do certain things imperfectly, and NVIDIA offers a simple and more optimal way, that is good. But in this case there is another rule that everyone must stick to: a program must work as originally planned (though I wish the developers showed what must be in the end). It's unfair to replace key commands in favor of one or another video card. We thus measure not the video cards performance but smartness of programmers.
and i will change the end by this.. "but the smartness of Futuremark programmers".because i have no problems with smart optimizations that increase performance without any drop in IQ.

the article could be even better in my opinion if they replaced the word cheat by optimizations. because we dont know exactly what ANti-detector is desactivating if legal optimizations or cheats. neither we know is antidetector is equally efective on ATI or NVidia drivers.because even the programmer UNWINDER told that he was unable to block some aplication detection from ATI drivers ,but that are there. so i think that unless ANti-detector do the same JOb in both cards ,or that we can clearly know what it is desactivating ,it should never be taken as the -ultimate Proof -of cheating.but more as a reference when ATI and NVidia detect aplications and compare the IQ diferences if there are any. even FUturemark patch 330 (the ones that FM says put everything right on track) and the detonators 4403 drivers doesnt look 100% exactly as the reference rasterizer images. so things are even more complicated because FM is not doing well their job.

The problem is deeply studied. In short, today all testers actually measure a degree of smartness of the guys at ATI and NVIDIA as well as of FutureMark instead of estimating cards' performance in the 3DMark. FM turned out to be a prostitute between these two majors. If it's so simple to replace shaders and the developers keep silence, what can we call it yet? That concerns 3DMark2001. Strangely enough but they started protesting against cheats only in the 3DMark03. At the same time it turned out that NVIDIA wasn't a beta tester anymore as the company didn't want to pay through the nose every year.
so even Futuremark in being critizised here.
the image comparisons between ATI and NVidia and the references shots looks exactly as the references images ,even at 400% zoom shots. the diferences are only noticeable when analizing the images at the ultra-microscopic level with the DIrecX9 references images,with photoshop .

however what was a surprise to me is that in PS2.0 test and in game4 with or without optimizations the IQ of detonators 44.61 was even closest one -very close-to the directx9 references shots.what will be the ATI explantions now? that directx9 reference shots are not reliable

lest give up with 3dmarks ,enjoy it better as a demo,and compare better the performance in games.

Last edited by Nv40; 07-09-03 at 08:26 PM.
Nv40 is offline