3DMark06 or WayOffTheMark06?

F

Frosteh

2[H]4U
Joined
Nov 30, 2004
Messages
3,664
OK so last night I retired my Leadtek 7950GX2 which has served me well for about 10 months and slapped in a BFG 8800GTX OC.

First thing I ran was 3DMark06, now I know that 3DMark has been a bad benchmarking tool ever since measuing performance was no longer a simple matter of measuring memory bandwidth or fill rate (way back in 2001/2003 editions) but I never realised quite HOW bad it was until last night.

Overclocked 7950GX2 with a PC running a E6600 @ 3.0Ghz and 4Gb of RAM, scored a reasonable 8,000 3dmarks, and the same system running with a 8800GTX clocked in just under 11,000 3DMarks which doesn't seem like a massive leap. I was a little disapointed to see them numbers come out and then turned my attention towards all the Nvidia tech demos to see what new features it was capable of.

But then I ran Oblivion, and oh wow was I impressed.

Old settings:
1680 1050
Medium settings for most sliders
Grass/self shadowing off
HDR ON
0xAA
4xAF

30 FPS outside

New settings:
1680 1050
Max everything
HDR ON
16xTSAA
16xAF

40 FPS outside

Wow what a leap, putting most of the sliders on max would have been enough for me, it crippled my old card doing so, I think mostly the grass and shadows were doing it. Then theres Max AF at 16x, which is significantly better than before.

Now this is the kicker, TSAA

"what in the HELL!" is all I could say!

Now in my personal experience Transparency AA tends to scale in performance with the amount of surfaces which actually use transparent textures, and in Oblivion outdoors, thats a LOT, so it's really no supprise that on my old card when I enabled just 2x TSAA it came to a crawling halt, seriously 5 FPS max. Quite frankly after seeing such an increase in all other graphics (in game options and AF) I never thought that AA would even run at 16xAA much less 16xTSAA, I thought maybe at a stretch 2xTSAA might be possible, anyhow boy was I wrong.

So my sort of initial "oh..." (read with slight dissapointment) at my 3DMark06 score really was squashed and I was happily playing Oblivion for most of the rest of the night in what can only be described as "all its glory"

My point?

Well I knew 3DMark has been a bad benchmarking tool ever since 3DMark03 and not to rely on its results because between AMD/Nvidia cheating and its base innacuracy it's not worth it. But I didn't realise it was THIS BAD.

I've not got hard numbers to back this up with but I estimate that if my 7950GX2 could run oblivion in the settings the 8800GTX can that FPS would be well below 1fps, in fact probably well below 0.1FPS (one frame every 10 seconds) The performance increase or rather the additional settings the 8800GTX allows is way higher than if you look at 3DMark06 which looks meager in comparison.

Lets get something straight I always knew 3DMark was a bad tool to use and why it was the first thing I tried I dont know (convenience probably) but I'm truly supprised by just how much it's off the mark.

I dub thee WayOffTheMark06 or WOTMark06 ;)
 
Isnt that funny though....I did a simalar thing when i grabbed a 8600GTS on release day and sold my 7950Ko to my nieghbor because the Shader technology in the GTS blew my 7950 away at 1024x768@75Hz on a 22" widescreen ...I couldnt believe it and many others didnt because all the benches were saying how crummy it was LOL... I own a 8800Gts which blows that away too ...Whats funny is i read 3dmark time after time saying wtf when the 8600gts score was low but the way it plays is amazing for midclass card and the 8800GTS well is like owning a Stealth fighter compared to a biplane just shear power. I always use Games to justify my new purchase , Its the only real way to see what one is made of!
 
Fully agree, using games as benchmarks is definately sound practice, I always used to run 3DMark just out of curiosity and also to test relative performance of overclocks on just one video card, I knew it was never going to be accurate but the difference when looking at the 8800GTX vs 7950GX2 is just dissapointing, real life games show a much different picture, especially when high levels of AA and AF are used.

Bring on Crysis :D
 
3Dmark is an excellent benchmarking tool and fun to watch.

if you're using it for the wrong purposes and to achieve ends it cannot that's your own damn problem.
 
I only ever use 3DMark to see if my system is running up to snuff. If I get a score similar to rigs with a similar configuration I know the build is solid in that I don't have any weird driver or hardware issues. I also use it occassionally to check and see if updated video drivers are performing like they should.
 
I tried running 3dmark and got a horrible score.

I played fear and hit 100 FPS. Screw 3dmark. If my games run well, I dont care about a freaking score..
 
i see your point OP but i dont agree to a certain extent. when i get a new rig i usually run 3dmark to test my old vs new. i dont care how high or low the score is, i only want to see the difference between my 5k score computer and my potential 3k+ upgrade, IE i just want to see the performance boost. i dont care about the score, just the difference between old and new. as a gaming benchmark i agree, its horrible.
 
Most of your points made so far are valid, but lets give credit where credit is due. For most of the 6 and 7 series video cards 3DMark06 is a great tool, but does not seem to be the best for the new 8 series cards. With that in mind, I must say "Hats off to Futuremark" who understands this situation and is currently working on the next installment of 3DMark that will do a better job with newer hardware.


XFX Support
 
Excuse me? How do you incorrectly use a canned benchmark, isn't that the whole point?

Same system (except video card), same OS and the latest version of their benchmarking software doesn't show a realistic difference between the 2 video cards, it shows a relativly small increase in speed to what you actually get in games like Oblivion.
 
real world > synthetic. no doubt. but pissing on 3dmark is looking a gift horse in the mouth. it arbitrates as any good diagnostic should. it is not the final stamp on whether system x can run game y, but...
when i looked at the latest vga charts on tomshardware last, it took only a cursory glance to discover that the 'overall games fps' average of 7 unique 3d intensive games scaled almost identically with the '3dmark 06 [v 1.0.2.]' results. reckon this is only coincidental?
 
I also get around 10,200 in 06, but I dont use it as a benchmark, I use it as a tweak and system setup utility, and for that it seems ok..

but as the OP stated scores in the 3dmark series cant be transfered over to real world gaming..
 
ojibewa said:
but as the OP stated scores in the 3dmark series cant be transfered over to real world gaming..
Click to expand...

Not always no.

But that still doesnt give anyone reason to crap all over it entirely. Just because you cant use a thermometer to measure atmospheric pressure doesnt mean you should decry thermometers. They have their valid uses.
 
I remember in mid-2005 when I got my 6800GT and installed it to play BF2. I noticed that it barely performed better than my previous FX5600. I asked some people and after running 3dmark05 for the first time I was able to see that my performance was half what it should have been. After notifying people on a particular forum they advised me to update/install my chipset drivers again.

3dMark helped.

I remember when I had my 7900GT go bad. I couldn't play any demanding games at my 22" monitor's 1680x1050 res. at max quality. It was choppy and unplayable in Battlefield games. the card scored an 11,777 in 05'. When that card died I got a 7950GT to replace it. The 7950 made gaming at 1680x1050 possible. It was honestly a decent step forward for me. The 7950GT scored only a couple hundred points more, but was also clocked higher.

3dMark was useless.


Its not completely useless, and its not the only thing people should rely on when considering performance in actual games.
 
It's a benchmarking tool, it's supposed to give you a score for hardware which is comparable to other hardware to give you a relative comparison, if it can't do that then it's not doing it's job.

The only use I can fathom is for overclocking cards, using it to stress and search for artifacts and to see how thg score increases with clock increases. But it's certainly not marketed like that.
 
Benchmark: A standard set by the best existing practice, product or service. A standard by which something can be measured or judged. Benchmarking is the process of comparing performance against that of others in an effort to identify areas for improvement.

AKA

Something that serves as a standard by which others may be measured or judged.


3dmark06 performance does not convert into real world performance... as flawed as that is however, that's not it's purpose.
 
You must log in or register to reply here.
Back
Top