does cpu overclocking limit graphics overclocking?

JPoPS13

Junior Member
May 1, 2008
10
0
0
Topic^^
Just wondering because i had my e8400 clocked at 3.60GHz, and was able to run 3dmark06 with my 8800gts at 815 core, 2021 shader, and 1120 memory. Then i increased my cpu voltage and got it to 4.1GHz, now i can't get 3dmark to complete a test sucessfully unless my graphics are at 780, 1934, 1066. Is there something else in the bios i have to change in order for my graphics card to work at high speeds again? or is this just what happens?

ps: i dont think that it crashes because my cpu is unstable because it says "your display driver has stopped communicating and has recovered", which happens when i push my graphics card to high.
 

ShadowFlareX

Member
May 6, 2008
150
0
0
Originally posted by: JPoPS13
... ps: i dont think that it crashes because my cpu is unstable because it says "your display driver has stopped communicating and has recovered", which happens when i push my graphics card to high.

Using Vista I presume?

If you go back down to 3.6GHz with your 8800GTS @ 815core/1120mem, is still stable? Are you running the RAM faster at 4.1GHz? Post the rest of your hardware setup, so it becomes clearer.
 

Continuity28

Golden Member
Jul 2, 2005
1,653
0
76
You're not using 3dmark to test CPU stability are you? Did you go into 3dmark only after testing it with more appropriate tools?

It could be the CPU that's causing the driver to crash. You need to first know if the CPU is truly stable at that frequency.
 

JPoPS13

Junior Member
May 1, 2008
10
0
0
no i was using 3dmark, though i never did test my cpu for stability (any suggestions for a good one?) However, after a little bit of testing, i found that it was the shader clock frequency that was causing it to crash. I unlinked it from the core clock in riva tuner, and ran 3dmark sucessfully with 810 core, and 1100 mem, however i had to leave my shader clock 1960.
The first guy asked about my build so here it is:
Mobo - DFI LP DK P35
CPU - Core 2 duo e8400
Memory - OCZ 2 x 2 800MHz DDR2
grfx card - BFG 8800GTS 512Mb
phillips ide dvd burner, wdc 500gb hdd

update: im sure it is the shader clock now, i completed another 3dmark test with graphics @ 815/1960/1120. for some reason my nvidia drivers don't like the idea of a high shader and cpu clock...
 

ShadowFlareX

Member
May 6, 2008
150
0
0
You can use Prime95 or OCCT to test your CPU stability. The longer you run it, the better (like 24 hrs) just so you're sure it's completely stable.
 

imported_Woody

Senior member
Aug 29, 2004
294
0
0
4.1 GHz is pretty impressive overclock. I assume you didn't jump from 3.6GHz straight to 4.1? :p

You really need to test the three main components, RAM, Motherboard chipset, and CPU (not including grapics card) of your system separately to determine what is the max overclock each can handle. Don't try overclocking the graphics card until you have the rest of the system stable.

There are many more detailed tutorials but basically you downclock the other components while pushing the one you are testing gradually to it's limit. I use memtest 86 to test the RAM stability and Prime95 to test the CPU and chipset stability (be sure to make sure you run Prime95 on all cores).

Once you have all that figured out you can come up with an overall plan to run your system at a stable overclocked speed.

3DMark is primarily for testing graphics performance and comparing the results of different graphics cards or overclocked graphics cards on the same system. It is also a useful tool to test the stability of an overclocked graphics card but can be a little hazardous. I personally use the built in 3D testing feature of ATiTool for finding the limits of a graphics card (works very nicely on NVidia cards too) then confirming the results with 3DMark. ATiTool offers the advantage of a controlled environment and running the 3D app in a window with an auto off feature as soon as artifacts are encountered. It will also gradually to the overclocking and testing for you while you watch saving some effort. I also use game benchmarking like Unreal Tournament 3. You can download ut3bench.exe and run it with the game demo.
 

JPoPS13

Junior Member
May 1, 2008
10
0
0
Originally posted by: Woody
4.1 GHz is pretty impressive overclock. I assume you didn't jump from 3.6GHz straight to 4.1? :p
Haha no, i went 3.00 -> 3.60 -> 3.86 -> 3.91 -> 4.00 -> 4.05 -> 4.08 -> 4.10, and even tried 4.12, 4.16 and 4.20. I know this because i documented pcwizard's "mandelbrot computation benchmark" for each frequency i oc'd to. But i did download prime95 and have been testing my system.

 

OCGuy

Lifer
Jul 12, 2000
27,224
37
91
What Vcore are you using for the 4.1? (What cpu-z says, not what you select in BIOS)
 

zach0624

Senior member
Jul 13, 2007
535
0
0
Like others have said check for processor stability with prime 95. What may also be causing your problem is the gpu might be getting overheated because your cpu is generating more heat. What do you use to cool your system and what type of airflow do you get through your case. Also google a gpu temp program (I think ntune does it. not sure though) and download core temp or real temp and check you cpu temp under load and idle.

BTW 4.1ghz is amazing.
 

KhadgarTWN

Junior Member
May 23, 2008
5
0
0
Originally posted by: JPoPS13
Topic^^
Just wondering because i had my e8400 clocked at 3.60GHz, and was able to run 3dmark06 with my 8800gts at 815 core, 2021 shader, and 1120 memory. Then i increased my cpu voltage and got it to 4.1GHz, now i can't get 3dmark to complete a test sucessfully unless my graphics are at 780, 1934, 1066. Is there something else in the bios i have to change in order for my graphics card to work at high speeds again? or is this just what happens?

ps: i dont think that it crashes because my cpu is unstable because it says "your display driver has stopped communicating and has recovered", which happens when i push my graphics card to high.

Theoretically, it won't, but technically it will, depend on your PSU and heat sinking devices
but CPU and GPU were supplied by +12 V mainly from PSU
while you OCed them , the
Power need (Pn) = [Power unOCed (Pi) x (Fo)/(Fi) x [(Vo)/(Vi)]^2] x (1+k)
where Fo = Freq OCed, Fi= Freq unOCed, Vo = Voltage adjusted, Vi=Voltage not-adjusted
when Fo/Fi is slightly > 1 (which means slightly OCed), k can be neglected
but when Fo/Fi is >1.5, or 2 (which means heavily OCed), k became larger

That means, an default E8400 is less then 35W, but an 4.1G E8400, especially with V adjusted can be over 100W
(Anand has similar test, plz find yourself if someone is interested in.)

Therefore, by default freq, your CPU+GPU may below 135W (35+100)
but while markably OCed, it can over 250W which may compromise the stability.
 

z1ggy

Lifer
May 17, 2008
10,010
66
91
I think that you should run OCCT for 24 hours and let us know what the results are. Before I was having issues with my E6750 at 2.8 Ghz running OCCT for more than 6 mins! But I increased my NB voltage by about .1v and my Vcore by 20mv and I become stable for 24 hours. Little things like that can make a big dif. Also make sure you run something like HWmonitor.. or speedfan to make sure temps don't get dangerously hot. I hope you have an amazing case airflow getting that chip up to 4.1. Very impressive.
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
Originally posted by: KhadgarTWN
Originally posted by: JPoPS13
Topic^^
Just wondering because i had my e8400 clocked at 3.60GHz, and was able to run 3dmark06 with my 8800gts at 815 core, 2021 shader, and 1120 memory. Then i increased my cpu voltage and got it to 4.1GHz, now i can't get 3dmark to complete a test sucessfully unless my graphics are at 780, 1934, 1066. Is there something else in the bios i have to change in order for my graphics card to work at high speeds again? or is this just what happens?

ps: i dont think that it crashes because my cpu is unstable because it says "your display driver has stopped communicating and has recovered", which happens when i push my graphics card to high.

Theoretically, it won't, but technically it will, depend on your PSU and heat sinking devices
but CPU and GPU were supplied by +12 V mainly from PSU
while you OCed them , the
Power need (Pn) = [Power unOCed (Pi) x (Fo)/(Fi) x [(Vo)/(Vi)]^2] x (1+k)
where Fo = Freq OCed, Fi= Freq unOCed, Vo = Voltage adjusted, Vi=Voltage not-adjusted
when Fo/Fi is slightly > 1 (which means slightly OCed), k can be neglected
but when Fo/Fi is >1.5, or 2 (which means heavily OCed), k became larger

That means, an default E8400 is less then 35W, but an 4.1G E8400, especially with V adjusted can be over 100W
(Anand has similar test, plz find yourself if someone is interested in.)

Therefore, by default freq, your CPU+GPU may below 135W (35+100)
but while markably OCed, it can over 250W which may compromise the stability.

generally cpu does little of the 3d graphics work which is offloaded to the gpu...In my opinion it would have to be (cpu ocing) extremely flaky already...

The 2 components that seem to effect it more when ocing is the system memory, and the PSU. As mentioned above.....Video cards are a huge draw in power in our current day systems....These two components are used heavily by the video card and system under 3d graphics load...

What POwer supply do you have? What are your rails? Single 12v rail? Double or quad? What is the efficiency rated as....

All our huge things. the more you are ocing your board, ram and cpu, the more you adding to the load needed under operation. Hence why we stress test and test under load....


Also should be noted that some recommend on certain boards when ocing the system to get stability with the graphics system they had to nudge the pci-express voltage up +.1v, perhaps nudge the NB chipset up a spot, as well as set the mhz to 100-105 manually in the bios....I have seen it work personally...