GeForce 8600M GT overclock
September 3, 2008
New updated result with higher result is posted here.
Truly, I am big overclocker enthusiast. Was, at least. First overclock experience was on Celeron 466 Mhz.
Afterward more, but this is not about my past.
Recently i received a new notebook from work. Brand new Dell XPS M1530. After reinstalling the VIsta for few times (some driver problems, that wifi problem, thought the driver was old, nope. Until i did not hang the wifi router, i was not satisfied. After I have realized, that the router was not so nicely set up. But ok, now everything is good).
The Notebook was up and running. Needed to make some tests. Played few evenings in Bioshock, Half- Life2, and so on. Have become bored. What else could be done with ‘the new toy’? Of course- overclock!
So here is the configuration
DELL XPS M1530
What`s inside:
|
CPU |
Intel Core 2 Duo T8100 2.1GHz |
|
Memory |
2048MB DDR2 667MHz |
|
HDD |
160GB 5400RPM |
|
Video |
nVIDIA GeForce Go 8600M GT 256MB |
|
Display |
15.4” 1680×1050 WSXGA+ |
|
OS |
Windows Vista Business SP1 |
And some screenshots from CPUID:
The OS is not optimized for clocking at all (yeah, with Vista and its Aero). Running from the AC- adapter, so the notebook does not clock down the video card, even accidentally. Later i will tell some strange thing about the video card.
Of course, this review is not so big as it could be, but here is overclocking potential tested.
The standard benchmark i used: 3dMark 2006. Screen is with all default values.
I decided to clock the memory first.
For that purpose I used Riva Tuner. Screen from my desktop.
Notice the difference between what is showed on Hardware Monitor and System tweak. The Core speed differes around 10 and 20 Mhz.
Started from 25 Mhz step, but it took too much time, so in the end i was taking bigger steps like 50 Mhz.
|
Memory |
750 MHz |
Stable |
|
|
775 MHz |
Stable |
|
|
800 MHz |
Stable |
|
|
850 MHz |
Stable |
|
|
900 MHz |
Stable |
|
|
950 MHz |
Stable |
|
|
1000 MHz |
Non Stable |
|
|
975 MHz |
Stable |
At 975 Mhz the memory was stable and took a benchmark.
Interesting thing, when memory was clocked to frequency, it could not operate normally. The frequency of whatever( core, memory) went down, to the ‘low power 3d’. And when you tried` v to start the benchmark, the frequency went down again! And it was impossible to change something, moving the frequency in Riva Tuner- video card did not react at all. Only restart helped always.
Don`t know why so, perhaps not to damage the card. Sometimes, when first tests were running nicely, at full speed, at the end the card lowered its frequency. Because of that 3dMark score sometimes was around 3,5k. Maybe now the main priority is image quality, not speed(meaning no horizontal lines anymore, when memory speed is too high?).Interesting…
Screen, showing the memory speed.
3dMark score with the memory on 975 Mhz.
With the memory on 975 Mhz,3dMark06 benchmark showed us 4272 (3dMarks?Parrots?).
+200 in 3dMark06, not so much at all. +275 Mhz.
Now we clear with memory, let us make some tests with core and shader clock.
I did not change the core/ shader speed ratio.
|
Shader |
Core |
Stability |
|
1080 MHz |
540 MHz |
Stable |
|
1116 MHz |
558 MHz |
Stable |
|
1188 MHz |
594 MHz |
Stable |
|
1242 MHz |
621 MHz |
Failed, another attempt. Works, 4583 3dMarks |
Again some interesting stuff. When the speed was adjusted to bigger value, than 621 Mhz (570 Mhz showed Riva Tuner), the speed did not increase. Same situation was with memory, maximum 1000 Mhz, but it was awfully unstable and the frequency was always dropped automatically. Some hardware protection?
When i was clocking the first time core and shader speed, it was made in one session. So the 1241/621 Mhz was unstable. It was already a bed time, so i went to sleep.
Next day, first attempt – 1241/621 Mhz. Passed! 4538 3dMarks!
Final Test
The memory was stress tested, core also. What do we need?
Complex benchmark. Now I run the whole video card clocked. Will see what happenes.
Here is some additional data:
|
Memory: |
950 MHz |
|
Core Clock: |
580 MHz |
|
Shader clock: |
1161 Mhz |
70 degrees Celsius.
Final set-up
Final Score
Result – 4648 3dMarks.
Good or bad? Lets use some math…
Memory:
Default: 600 Mhz
Clocked: 950 Mhz
Speed: + 58%
3dmark default: 4025
3dmark clocked memory: 4272
Increase: + 6,1%
Core/shader:
Default: 475/ 975 Mhz
Clocked: 580/ 1161 Mhz
Speed: + 22,1%/ +19%
3dmark default: 4025
3dmark clocked memory: 4538
Increase: + 12,7%
Full overclock:
3dmark default: 4025
3dmark clocked memory: 4648
Increase: + 15,5%
Conclusion
What to we have in the end? Increased efficiency by 15%? Not to much. I am concerned, that it is possible to squeeze out some more 200 points in 3dMark06, but is not it worth of it. We should not forget this is notebook, we are carrying it with us, and the battery has reasons to recharge extra fast with overclocked video.
My opinion- the overclock is not worth of it!
P.S. In time, perhaps, i will make some test, but only the PhysX stuff activated on this video card, i heard this is possible…







