• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

3DMark score keeps changing.

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

crull

Member
Joined
Dec 19, 2002
For some reason my 3DMark2001 score keeps changing and not by a little bit. I alway run the test right at boot up using the same exact system settings. My 6600GT is wired with an external temperature probe and is always running pretty cool, same for the rest of my system such as CPU. My system voltages also seem to be fine. Here are my scores running the test 8 times, notice the difference in scores. I never had this with my older 4200. The score would change, but never by this much. Any ideas? Thanks.

17,925
17,414
17,594
17,678
17,827
17,924
17,572
17,612
 
crull said:
For some reason my 3DMark2001 score keeps changing and not by a little bit. I alway run the test right at boot up using the same exact system settings. My 6600GT is wired with an external temperature probe and is always running pretty cool, same for the rest of my system such as CPU. My system voltages also seem to be fine. Here are my scores running the test 8 times, notice the difference in scores. I never had this with my older 4200. The score would change, but never by this much. Any ideas? Thanks.

17,925
17,414
17,594
17,678
17,827
17,924
17,572
17,612


Crull, those changes are not uncommon. Also, as your system's architecture changes, and the components change, you should base your results more on the present rather than the past. No looking back. I'm sure you ran 3DM'01 more than a few times on your old set-up, and I'm sure you had a fluctuation in the scores, even if it didn't seem "as bad" at the time. I don't remember running htat bench and getting identical scores dating back to Slot A PIII's. You have to also keep in mind, Crull, when you run a bench, there are literally millions upon millions of variables to consider, which would constitute variations of your final score(s).

All in all, I'd say you have no worries, champ. :)
 
glock19owner said:
01 is system based benchmark...if ran without rebooting between runs...this is normal...

I always run it with clean system at boot up.

I understand you never get the exact score everytime, but 500 point difference doesn't make sense.
My old set up had score changes but always under 50 Points, never this much. The score is all over the place for some reason and I can't figure out why.
Maybe the test because it is so old was not designed with such high scores in mind, but that is what I am trying to figure out. I need someone else who is running a 17,000+ score to run the test a few or more times at boot up to confirm how much their score changes.

Thanks
 
crull said:
Maybe the test because it is so old was not designed with such high scores in mind

I wouldnt say that...my 27k score was done a little over 2 weeks ago with the system in my sig...
 
500 points is basically a 3% change right?

If your old 4200 was within 50 points I'd say that was more the anomaly than the 3% difference you may see now. :)

What was the score range for the 4200? Just to guess: 12000-13000?

Are you on a different BIOS now?
 
The low detail Car Chase test fluctuates crazily. 30-40 fps (300-400 3dmarks) isn't uncommon at all. 3DMark is rendered real-time, so some of the events can go differently at different times. In this case its usually the barrels getting thrown around. As ZX12 said, its weirder that you didn't get any fluctuation with the Ti4200. Fluctuations are a lot more normal than not.
 
glock19owner said:
I wouldnt say that...my 27k score was done a little over 2 weeks ago with the system in my sig...

What I meant was if you were to run the test a few times. You score might jump around like mine. Try running it a few times, re-boot each time and compare the scores.
 
ZX12Rguy said:
500 points is basically a 3% change right?

If your old 4200 was within 50 points I'd say that was more the anomaly than the 3% difference you may see now. :)

What was the score range for the 4200? Just to guess: 12000-13000?

Are you on a different BIOS now?

Same bios with the 4200. I used the 4200 with two different CPU's. With my first 2500+ Barton the score was around 13,150. I write my scores down to keep track when I try different drivers. That score would vary at the most maybe under 50 points. I could always tell if there was any problem because it was very consistant.

Then I moved up to a better overclocked 2600+ Mobile with the 4200 and my 4200 score was around 15,000 give or take 50 points.

I agree that it is only around 3%, but I still never had that much change in score with the 4200. I had the 4200 for over a year so I tested it a lot.

To see the score jump so much now makes it hard to get a baseline for video performance. So I can test new video, fsb and CPU overclock settings to see if they make a difference. I can't tell now because the score is all over the place. I don't want to use 2003 or 2005 they take too long. I might give Aquamark a try if it is more precise.
 
The way to accomplish this is to send all of your scores up to the OnlineResultBrowser and check out how all the invidual tests are doing. This is what partially makes 3dmark01 so useful.

Car Low measures GPU and CPU, highly variable
Car High measures CPU and memory performance
The Dragothic tests both are primarily GPU dependent, but also can be used to compare CPU performance when pitted against the same GPU
The Lobby tests test memory performance
Nature is a 99% GPU test

Everything but car low should very clearly scale depending on what variables you've changed. 3DMark2001 is very clearly a full system round tests, that's why some of us loonies still love it so much.

03 and 05 are the ticket for pure GPU performance. You can press escape after all the game tests, btw, makes them take about as long if not shorter than 3dmark01. Aquamark also seems very system-dependent.
 
Gautam said:
The way to accomplish this is to send all of your scores up to the OnlineResultBrowser and check out how all the invidual tests are doing. This is what partially makes 3dmark01 so useful.

Car Low measures GPU and CPU, highly variable
Car High measures CPU and memory performance
The Dragothic tests both are primarily GPU dependent, but also can be used to compare CPU performance when pitted against the same GPU
The Lobby tests test memory performance
Nature is a 99% GPU test

Everything but car low should very clearly scale depending on what variables you've changed. 3DMark2001 is very clearly a full system round tests, that's why some of us loonies still love it so much.

03 and 05 are the ticket for pure GPU performance. You can press escape after all the game tests, btw, makes them take about as long if not shorter than 3dmark01. Aquamark also seems very system-dependent.

Thanks that's some useful information.

I'm trying to get an idea of graphic performance changing the CPU speed, and memory bandwidth by raising and lowering the fsb and CPU multiplier. I'm at the point right now where I am trying to sqeeze every last bit of performance out of my system, but having that score jump around is making it difficult. So I guess I should use just the Car high test?

Are you saying I can skip some of the tests in 03 and 05 and will still get a quick score to gauge GPU performance? I didn't know you could skip some of those tests. I knew you could with 2001 though.
 
Yeah, when it says its loading the first CPU test on both 03 and 05, hit escape. In 03, press "Details...", and in 05, it should pop you with a score immediately.
 
Gautam said:
Yeah, when it says its loading the first CPU test on both 03 and 05, hit escape. In 03, press "Details...", and in 05, it should pop you with a score immediately.


Really good to know thanks again!
 
Back