C90S: 8600M GT/ 512 DDR2 3DMark06 results

All,

AS above C90S 3GB RAM Vista x64. I run the free version of 3Dmark06 and the tests seem to run at 1280x1024 res. I have listed my results below. Most of these tests had a reboot in between. I'm running the 174.93 driver.

475/950/400 3077
500/1000/416 3112
550/1100/440 1679
561/1122/416 3516
561/1122/430 3393
570//1140/416 didn't complete
580/1160/416 didn't complete
600/1200/416 didn't complete
570/1140/400 didn't complete

Other forums suggest that the maximum stable setting the 8600M GT 512 can be clocked to is 625/1250/482 do people agree with this ?
My question is why can't I get anywhere near this setting ?

Any comments or suggestions on the above would be welcome

 

not all gpu's are made equally. and neither are all laptops. maybe the most stable gpu (luck of the draw) matched with a laptop with excellent air flow, can be borderline stable if you have your fans blasting 100%.

I personally put my core clock at 540, and it doesn't like being above about 555 no matter what the temperature.

 

Even though every single 8600M GT is supposed to come out of the factory exactly identical to every other, they often do not. So one card that will run 615/1250/482 is more solid and well built than one that cant handle barely above stock speeds. Figuring out which you have and gambling with replacements is a pain, but if you want a card you can push to the limit, you may have to do that.

Hope that helps.

 

The highest I reach is 499 (530 rivatuner)/1350 (1450 rivatuner)/527

 

Your thermal compound has alot to do with it. Mine can reach 635/1400/535 with some arctic silver and my zalman cooler.

 

Also these tests seem to return different results for no reason ...

for example I have run the 561/1122/416 test 5 times and got 3227, 3247, 3201, 3291 and 3516. The 561/1122/430 returned 3161 and 3393

you would expect that testing computer hardware you would get similar results under the same circumstances. Or is this how the 3dmark software performs. ?

How does the resolution that the test is run at affect the result ? is 1024x768 a standard res for running these tests ?

Thx

 

It's 1280x1024

 

A computer running a 2008 Operating System is not a deterministic machine. There are many random elements at work, such as (for instance) one process getting more CPU clocks during a seemingly identical run, than in a previous run, due to the multi-threading algorithm (I don't have the exact technical terms right now and this may not be the best example but you get my point, i hope).

Therefore, there will be slight differences between different runs of a particular benchmark. In technical terms, the benchmark score will be a stochastic variable. The only way to get a reliable measure is to run many tests (at least 10) with the same settings and make an average, and compute a standard deviation and a confidence interval.

That being said, a 300 or 200 pts difference in score is 10% and a bit large. If 3D Mark results are so inaccurate it's not really a reliable benchmark. Maybe something was running in the background during most of the runs (like antivirus scan, Vista indexing, updates, defragmenting etc.etc. there are many possibilities and many are hidden and difficult to control)