Review 003: Effects of CPU frequency on 3DMark 06 Scores (SLI)

Review 003 - Effects of CPU frequency on 3DMark 06 Scores (SLI)

Components

Processor: Intel Core 2 Extreme QX9650 C0 L744B174
Motherboard: Asus Striker 2 Formula 780i Bios 1401
Ram: 2 x 2GB DDR2 800 Crucial Ballistix Red Tracers
GPU: Asus 9800GTX 512MB x 2
HDD: VelociRaptor 300GB
PSU: Silverstone Zeus ST85ZF 850W
Casing: Silverstone TJ09-BW

CPU Cooling: Swiftech H20 120 Compact TX-2 (Silverstone FM121)
GPU Cooling: Thermalright HR-03GT TX-2 (Zalman ZM-F3RL) x 2

OS: Windows Vista Ultimate 64 Bit
Video driver: 175.16
Software: 3DMark 06 Advanced Edition
Software: CPU-Z 1.45

Voltages

All voltages are set in bios.

Ram: 1.92V
1.2V HT: 1.23V
NB: 1.52V
SB: 1.5V
PLL: 1.52V
VTT: 1.24V

CPU @ 3.0 Ghz: 1.1V
CPU @ 3.2 - 4.0 Ghz: 1.34V
CPU @ 4.2 Ghz: 1.4V
CPU @ 4.4 Ghz: 1.45V

Overclocking

All overclocking are done in the bios.
Only the CPU is being overclocked.
GPU at stock frequency.

3.0 Ghz - 333 x 9
3.2 Ghz - 400 x 8
3.4 Ghz - 400 x 8.5
3.6 Ghz - 400 x 9
3.8 Ghz - 400 x 9.5
4.0 Ghz - 400 x 10
4.2 Ghz - 400 x 10.5
4.4 Ghz - 400 x 11

Ram is at 800 Mhz 4 4 4 12 Timings.

Review 003 - Effects of CPU frequency on 3DMark 06 Scores (SLI)

- This review is on the effects of CPU frequency on 3DMark 06 Scores.

- To allow for a more accurate results, 3DMark 06 will be run three times for the each set of CPU frequency.

- Each indiviual test will be carried out after a fresh restart of the system and allow a 5-10 min waiting time for the memory to be cached before 3DMark 06 is run.

QX9650 @ 3.0 Ghz + 9800GTX SLI @ Stock










QX9650 @ 3.2 Ghz + 9800GTX SLI @ Stock

QX9650 @ 3.4 Ghz + 9800GTX SLI @ Stock










QX9650 @ 3.6 Ghz + 9800GTX SLI @ Stock

QX9650 @ 3.8 Ghz + 9800GTX SLI @ Stock










QX9650 @ 4.0 Ghz + 9800GTX SLI @ Stock

QX9650 @ 4.2 Ghz + 9800GTX SLI @ Stock










QX9650 @ 4.4 Ghz + 9800GTX SLI @ Stock

Overall Results







Graphs of SM 2.0, SM 3.0, CPU and 3DMark 06

All graphs are plotted using the average values from the three tests.

SM 2.0 Graph







SM 3.0 Graph







CPU Graph







3DMark 06 Graph







Analysis of the graphs and tables

- SM 2.0, SM 3.0, CPU and 3DMark 06 scores increases linearly with an increase in CPU frequency.

- There seems to be a greater fluctuations in all the scores obtained from this SLI system as compared to the scores obtained from a single gpu system.

Graphs of SLI and Single GPU system

All graphs are plotted using the average values from the three tests.

Graphs of single card system are plotted using the results obtained from Review 001.

SM 2.0 Graph







SM 3.0 Graph







CPU Graph







3DMark 06 Graph







Conclusion

- As expected, 3DMark 06 scores fluctuates alot more with a SLI system. Meaning, the scores obtained from running a 3DMark 06 benchmark with the same settings, varies.

- Hence, to obtain a best score for SLI systems, one needs to run the 3DMark 06 benchmarks multiple times at the best settings.

- Based on 3DMark 06 benchmarks, to take advantage of SLI system and increase 3DMark 06 scores, one need to overclock its CPU heavily.

- As we see from the various graphs, an increase of CPU frequency will always result in an increase in all the scores. Hence, it seems that the limiting factor of the scores is the CPU frequency.

- Comparing a single GPU system and a SLI system, we can clearly see that a SLI system beats a single GPU system flat.

- However, one important point to note is that one needs to overclock its CPU frequency heavily before the differences can be felt.

- To verify my statements and confirm my results are accurate, I went to check up on the 3DMark 06 scores online.

- Sure enough, CPU like QX9650 clocked to a frequency of 5 Ghz and above see a 9.5K SM 2.0 score and a 11K SM 3.0 score. This proves that the limiting factor of the 3DMark 06 scores is the CPU frequency.

Reserved..

Reserved...

hmmm performance scales very well with the increase in processor speed sia...

edit: so nowaday's dual GPU setups are actually bottlenecked by the CPU?

ravener said:

hmmm performance scales very well with the increase in processor speed sia...

edit: so nowaday's dual GPU setups are actually bottlenecked by the CPU?

Click to expand...

gpu scaling vs clock frequency nia

my gpu score in 3dmark for 3,2ghz is the same comparing E6750 and Q9450

ravener said:

hmmm performance scales very well with the increase in processor speed sia...

edit: so nowaday's dual GPU setups are actually bottlenecked by the CPU?

Click to expand...

varies with diff benchmarks...

3dmark vantage totally diff..

no oc..single

P 5K + scores..

no oc..sli

P 9K + scores..

oc...

P10K + scores..

1 thing to note..newer gpus score better in vantage..

seems like vantage is designed for newer gpus onli..haha..

will be doing on vantage..soon..

Another piece of great work.

wow nice review and nice scaling

Yo, alpha lol.. fancy seeing you here....

Anyway, it is a nice review and it help me a lot in understanding what are the bottlenecks in my system right now.. I just need that water cooling soon

great OC n review.

ER said:

Another piece of great work.

Click to expand...

thnx bro..

eliteneon said:

wow nice review and nice scaling

Click to expand...

brandon1987 said:

Yo, alpha lol.. fancy seeing you here....

Anyway, it is a nice review and it help me a lot in understanding what are the bottlenecks in my system right now.. I just need that water cooling soon

Click to expand...

hmmm..i need to do on vantage and in game fps to find out if its realli due to cpu, drivers or the software itself..

xBlackHawkx said:

great OC n review.

Click to expand...

yoz ppl..juz did a single card on vantage..

and its totally diff from 3dmark 06..

vantage benchmark..

qx9650 + single 9800gtx

gpu scores
cpu at 3ghz
4992
4997
4982

cpu at 3.2ghz
5002
4992
4992

cpu at 3.4ghz
4992
5001
4993

cpu at 3.6ghz
5014
4994
5018

cpu at 3.8ghz
4998
4985
4997

cpu at 4ghz
4995

haha..unlike 3dmark06, cpu frequency doesnt affect the gpu scores much..

gonna try vantage again with lower frequncy like 2ghz...

upzzz