3Dmark11 does not make sense ? low score

[gripen90]

I am trying to figure out if my score is ok or if something is really wrong, because some results just don't make sense to me

Can some one enlighten me if CPU speed has great impact on the Physics Test and Combined Test ? Mine seem so oddly low when I compare the overall results to a system a friend has.

My system is Intel Core i7 870 2.93GHz, 16GB DDR3 1333Mhz ram CL9, 3x GTX 580 3-way SLi, Asus P7P55 WS SUPERCOMPUTER, Win 7 64bit Sp1, Forceware 275,33.
Nothing overclocked

http://3dmark.com/3dm11/1364547

3DMark Score P11995
---------------------

Graphics Score 17061
------------------------

Physics Score 6807
-----------------------

Combined Score 5758
-----------------------

GraphicsTest1 74 FPS
-----------------------

GraphicsTest2 77 FPS
----------------------

GraphicsTest3 109 FPS
-----------------------

GraphicsTest4 53 FPS
----------------------

PhysicsTest 21 FPS

--------------------------
CombinedTest 26 FPS


My friends PC scores this with Intel Core i 7 960@4.2Ghz, GTX 570 OC SLi, 6GB DDR3 1333mhz, Asus Rampage III Extreme , Win 7 64bit sp1, Forceware 275,36
http://3dmark.com/3dm11/1349511

3DMark Score P11926
----------------------------------

Graphics Score 13259
---------------------------------

Physics Score 9682
---------------------------------

Combined Score 8484
----------------------------------

GraphicsTest1 58 FPS
----------------------------------

GraphicsTest2 60 FPS
---------------------------------

GraphicsTest3 85 FPS
---------------------------------

GraphicsTest4 41 FPS
----------------------------------

PhysicsTest 30 FPS
----------------------------------

CombinedTest 39 FPS


We pretty much get near the same Perfomance score ! and I can see that since the physics and combinered score count so much in the overall performance score you actually get pretty misleading results

Compare link
http://3dmark.com/compare/3dm11/1349511/3dm11/1364547

Comparing my results with my friend he has a much better Physics and Combined scores with GTX 570 OC SLi and the OC'ed Core i7 960, but if you then compare with the graphics scores/ Game 1 to 4 tests he has 13'259 point then I have 17'061 ... but still his system comes out as the overall better one !.

To me this is very misleading ... and I really don't get it.

Is something plain wrong with my system ? 3Dmark Vantage give proper results and so does Unigine Heaven DX 11 2.5 Benchmark.

[Usual Gubbins]

seems the trend in results reflects the two machines, he has a much faster cpu clock and grade of cpu compared to yours and you have a much higher grade graphics system compared to your friends. Would be much better in trying to compare a system to yours that has at least the same cpu/speed or the same graphics setup for a proper compare. You are effectively comparing an audi to a bmw, for example both do the same thing but do it differently, you need to compare the same car but with a different engine would be better.

Chris.

[gripen90]

seems the trend in results reflects the two machines, he has a much faster cpu clock and grade of cpu compared to yours and you have a much higher grade graphics system compared to your friends. Would be much better in trying to compare a system to yours that has at least the same cpu/speed or the same graphics setup for a proper compare. You are effectively comparing an audi to a bmw, for example both do the same thing but do it differently, you need to compare the same car but with a different engine would be better.

Chris.

Still that doesn't change the fact you get very misleading results. Cars only drive at one type of "speed" no matter the engine.

Here's my friends setup at the CPU stock clock at 3.6GHz vs mine at 3.2GHz (due to Turbo Core enables).
Still his performance score is very very close to mine and his Physics and Combined scores are way higher than mine.
http://3dmark.com/compare/3dm11/1349423/3dm11/1364547

I also on the ORB found a near similar setup as mine (3x GTX 580, Core i7 940) it also gets higher Combined and Physics Score ... but lower Graphics score.
http://3dmark.com/3dm11/455419

compare http://3dmark.com/compare/3dm11/1364547/3dm11/455419

[codydewey]

Its just the speed your CPU is running at. Lots of users here that could help you in overclocking that processor, then you should be dusting your friends score.
Good luck. Post a thread in the CPU section for help overclocking your I7.

[gripen90]

Its just the speed your CPU is running at. Lots of users here that could help you in overclocking that processor, then you should be dusting your friends score.
Good luck. Post a thread in the CPU section for help overclocking your I7.

Well I'm not interested in setting records etc, but just understanding the "logic" behind these results which I evidently now see there's not much of.
Overclocking I'm not really fond of since stability and data integrity means a lot to me.

[Jarnis]

You get a lot better graphics score, he gets a lot better CPU-related score (1.2 GHz difference is huge) and in overall score those pretty much even out.

The only somewhat-oddball thing is the combined test score on your system but that may be due to driver issues. It seems that the compute shader code in test 3 sometimes runs into issues with SLI/Crossfire setups with 3 or more cards, probably due to the video drivers being not very well tuned for that scenario. Considering that 3DMark 11 is currently probably the only application that even tries to fully stress multi-GPU setups with simultaneous graphics rendering *and* compute shader code even in triple/quad SLI/Crossfire, driver issues like this are not too surprising.

So... assuming the drivers get fixed in the future, I would expect your combined score to creep up a bit - even if it will still be held back by the CPU when comparing the systems you compare.

[Meaker]

Crank it up to a higher setting and your cards will stretch their legs.

There is only so much they can do at 1280x720 with your current CPU.

Also are you running mismatched ram and only at 533mhz (1066DDR)? Or is that the detector going funny.

[gripen90]

Crank it up to a higher setting and your cards will stretch their legs.

There is only so much they can do at 1280x720 with your current CPU.

Also are you running mismatched ram and only at 533mhz (1066DDR)? Or is that the detector going funny.

Well I have two sets of these RAM http://www.corsair.com/memory/xms-cl...m2a1333c9.html but for some strange reason the SPD of one of the sets has 533mhz in the SPD eventhough they run at 667Mhz just like the other pair.
BIOS, Speccy and CPU-Z show them all run 667Mhz (DDR3 1333Mhz)
So the detector is both right and wrong [rolleyes]

[Meaker]

Lol well just double checking heh.

[gripen90]

You get a lot better graphics score, he gets a lot better CPU-related score (1.2 GHz difference is huge) and in overall score those pretty much even out.

The only somewhat-oddball thing is the combined test score on your system but that may be due to driver issues. It seems that the compute shader code in test 3 sometimes runs into issues with SLI/Crossfire setups with 3 or more cards, probably due to the video drivers being not very well tuned for that scenario. Considering that 3DMark 11 is currently probably the only application that even tries to fully stress multi-GPU setups with simultaneous graphics rendering *and* compute shader code even in triple/quad SLI/Crossfire, driver issues like this are not too surprising.

So... assuming the drivers get fixed in the future, I would expect your combined score to creep up a bit - even if it will still be held back by the CPU when comparing the systems you compare.

Ok that explains a fair bit about the problematics !.
I did another comparision with a person with same motherboard, a Core i7 750 and 3x GTX 580 too. The CPU is overclocked to 4GHz and the GTX 580's slightly OC'ed also.
I can see the 4GHz has a slight invoke one the physics and combined test, but still they make out an important significance in total. So If running stock he'd land around my results too ?. The core i5 750 is fairly comparable with core i7 870 eventhough it lacks hyperthreading, but I'm not sure how significant a factor HT is on 3Dmark11.
http://3dmark.com/compare/3dm11/321778/3dm11/1364547
Anyhow thanks to all replies making me about wiser about how 3Dmark11 works.[thumbup]