I wanna complain about 3DMark 2001!

[rogue1979]

Here are several reasons Mad Onion does not get my respect.

1. 3DMark 2001 does not allow the Geforce 2 chipset to score on DirectX 8 tests. However, it will run DirectX 8 games and demos fine.

2. There is a 25% difference in scores from 1000MHz to 1500Mhz on the same system, but all frame rates in games show very little improvement on the faster system (less then 10%)

3. I can fire up my computer on a cold morning, take of the case cover and put a fan there, go to 2.2v and 1620Mhz, overclock my Geforce 2 Ti 500 to 300/540 (the heck with the artifacts) and get just one pass without crashing. My score improves from 5500 to 5800, even though a second attempt will send my computer and video card crashing hard. Gee, now I can send my bogus score to Mad Onion for bragging rights, big deal!

Maybe I got the whole thing wrong, but isn't the whole point of a video card benchmark test to compare the real world performance of different video cards in different systems? Why can't they make a benchmark that incorporates a wide variety of games both new and old, allow testing with FSAA and without, loop it 5-10 times so you can't squeeze out one bullsh!t score, and include an artifact tester. That would be a real score, and would let you compare the real world speed of older video cards against brand new ones.

Sometimes I read about benchmarks to figure out if I want to buy a new video card and try to make an educated decision. But without a valid standard test, the only way to be sure is to buy the new card and test against the old in your own system, kinda of an expensive way to find out.


Whaddya guys think, am I just a complaining fool?

 

[DClark]

Only the GeForce3 and Radeon 8500 cards can do the nature test because they're the only ones with complete hardware Dx8 capabilities. Your card may run a Dx8 game, but it will run it with pixel shaders turned off (pixel shaders can't be emulated in software). The original Radeon does have pixel shaders, but they're not Dx8 pixel shader, so they can't do the nature test either.

As for scores, it's a basic calculation. (Low detail framerates x 10) + (High detail framerates x 20) + (Nature scene framerates x 20) = your score. This weights the tests properly, as it's much easier to get an extra fps in the low detail tests than the high detail tests, and less powerful processors will struggle more than top end processors with the high detail test.

In regards to cold benches, that's fine for a top-end, balls-to-the-wall run, but I mainly use 3DMark2001 to test the effectiveness of new drivers as they come out - I'd never think that anyone with an extreme top score would constantly use his computer at that spec - My high score is over 3200 marks, but in the trim I use day to day, I score just over 2700 marks. I have a baseline bench which is P3 700@933 and my graphics card (Radeon LE) clocked stock. Off the top of my head, I know that the 7075 drivers gave me 2623 marks, the 7199s 2607, and the new 9005s 2711 marks. The baseline scores I run are labeled as baseline or default runs and also labeled as Radeon LE, so others know that it's a non-overclocked score using the Radeon LE. Others label their benchmarks with their clock speeds to help with sorting through the scores, but I think in 3DMark2002 should sort scores further, allowing an option for separating 32mb and 64mb cards of the same type, separating OSes, and possibly mentioning the clock speed of the card in the specs.

3DMark is just a tool to gauge performance, like SiSoft Sandra, Quake3, and Serious Sam. It's just a very popular tool because of its massive database of scores. Complaining about 3DMark2001 is kind of like complaining about a knife when somebody uses it to stab someone. The guys over at MadOnion have my respect in spades, for delivering an excellent benchmarking utility and housing a database with which to judge one's overall performance. Plus the easter egg makes a nice little diversion .

If you're interested in a graphics card, just search the web for in-depth reviews - cold hard numbers are fine, but many times hardware can't be judged only on the raw performance in one benchmark - the difference between 100fps and 95fps may be more than compensated by any additional features the "slower" graphics card has to offer.

 

[rogue1979]

I guess I will have to admit 3DMark 2001 is the best overall benchmark out there, but it is still very inaccurate and needs a huge improvement. The issues I have addressed should be easy for Mad Onion to fix, but I feel they will not be motivated to do so since I believe they are politically biased.

 

[merlocka]

The ORB scores, especially those which score the highest, should be taken with a grain of salt. The tool is best used to compare results on your system before and after changes/tweaks/driver revisions, etc...

3dM2K1 is a useful tool, as is any well written benchmark although the competition to see who can get the single highest score is silly IMO.

 

[MilkPowderR]

Im the WINNA in P3's

 

[Crapgame]

Im the wiener in 1400mhz range ATI7500 defaults.

I get less than 200 points difference between my hyped up and my actual what I play at settings, I guess I just ride the edge as it is.

Im sure everyone has at one time or anouther pushed just a little harder than normal cause that one guy was a few points higher. I admit I saw how close I was to the top and wham I was dropping my multi upping my FSB and giving the card an extra 2/2mhz before you could say "this isnt a game". I think everyone takes them with a grain of salt as they should, If your machine is tweaked to perfect stabillity and performance and youre still a few hundred short of everyone in your category dont sweat it. Odds are you wouldnt want to run your machine like that, I dont get artifacts and I could run as my highest score settings but 155FSB/cl2 on Kingston valueram 133 just doesnt make me feel all warm and fuzzy. It does it but darn it just dont feel right.

 

[Innoka]

I don't know if the current situation is any different, but 6 months ago I was finding quake III the most representative benchmark. It stresses the videocard, bus and processor in that order. Rogue, can you tell us what DX8 games you've been running on a GF2? And, did you use the software mod on your GF3 recently and what were the results?

 

[rogue1979]

My video card is a Gainward Geforce 2 Ti 500, resistor modded to a Quadro2 Pro, flashed to the Nvidia Geforce 2 Ti reference bios. The nature scene in 3DMark and a demo of Aquanox (which I didn't like) is all I have tried in the DirectX 8 category, but they ran and looked just fine. Heck, if you listen to game developers and Microsoft they will tell you that an older technology video card won't work. They give you installation warnings that if you don't have DirectX 8 installed on your OS the game won't run. I tested this theory with Mech Warrior 4 and Black Knight and it was a bunch of BS, they ran flawlessly with 7.0a.

I feel that Mad Onion is all a part of this "conspiracy" to make people feel the need to upgrade when it isn't necessary. Yes, all of us on this forum know the truth, but the average Joe walking down the street will fall prey to the propaganda out there and spend alot of money he doesn't have to. Look at Microshaft right now, all of a sudden there is a refresh rate problem using DirectX and Open GL in Windows XP. That is the biggest load of crap ever, they are leaving it up to the video card manufactures to fix the DirectX problem, and hoping that no solution for the Open GL is found. This means that game developers will quit supporting Open GL completely, leaving Linux sh!t out of luck for gaming support.

 

[Insidious]

<< The ORB scores, especially those which score the highest, should be taken with a grain of salt. The tool is best used to compare results on your system before and after changes/tweaks/driver revisions, etc...

3dM2K1 is a useful tool, as is any well written benchmark although the competition to see who can get the single highest score is silly IMO. >>



That just needed to be said again.............. KUDOS![/i] >>

 

[DARRIN]

merlocka - well said.