GPU Overclocking-Crash Question.

When I overclock my 8800m GTS through Nvidia System Utils...I am encountering strange inconsistencies in crashes, that I can not correlate to anything specific...

What I mean is...if say, I set it to 625/950/1548, sometimes I can go hours and its fine, sometimes, it will crash in 5 mins..

And the crashes are unique as well, cpu just goes to black screen, but the system stays on underneath...no blue screens, file dumps or error messages when you reboot...

Now I have done ALOT of testing...and my GPU never gets above 65c even until full load, when it does, the FAN kicks on to high and it drops down to 60/61c...

My CPU under full load is between 50-55c (t9300) and the two temperature zones posted are 50/60 max for zone one and two respectively.

Nothing in my system peaks over 65c under the most extreme conditions...

So, temperature obviously is not the issue...

Same events will happen on smaller overclocks as well, no rhyme or reason...

Any Ideas?

 

Do you really want to kill your gpu?

 

that means you overclocked too much...

 

You may be encountering voltage instability; i.e.- even though you GPU temp is under control, your A/C adapter can only output so much power to the machine. So in theory, albeit you temp is sub 70C, you OC is still limited. Try undervolting you CPU with rmClock, and see if that helps.

 

Undervolting the CPU won't affect the GPU other than lowering the temperature.

What the OP has come across as said by amoney3 is his OC limit. The GPU is drawing power that does not exist and crashes. I would think the only way to OC further is to overvolt the GPU, but this not worth the risk...

Just wondering, what are the stock clocks?

 

you have overvolt your GPU like i did

 

well like I said...the strange thing is, sometimes I can play for hours at max load with no crash at all, and sometimes it will occur 5 mins in...

Could it be a Power Supply Issue, ie not exceeding clean, so when the gpu is drawing more power, it stresses the PSU?


The Stocks on it are 500/800/1250... for a test, yesterday I ran 3dmark06 yesterday about 10 times at 625/950/1512, and it never crashed, ad I peaked out at 10100ish...

But basically...since its not heat related, its power related...so either the OC limit is hit or the PSU just can't keep up....

 

Tavara...

Did you really over volt the GPU? If so, how and how did it affect your temps?

 

yes i did, with NiBiTor,my temp are same, just my fan kick more often

 

I would do some serious stability testing using ATiTool and 3DMark06 before you nuke your GPU.

 

From the title, I thought you had a specific question for me!

Anyways, have you tried using ATITool's artifact tester?
http://www.techpowerup.com/atitool/

 

Can't get ati tools to work on Vista 64, says the low lvl driver is unsigned

One thing I noticed today in doing some more testing that the crash will only happen after the gpu fan kicks on high for a bit...if the fans not running hard, then it will never crash...the fan kicks on at 65c, and lowers the gpu temp to about 60/61 then shuts off for a while...until the gpu reheats to 65c again...

It is just overall weird and random, I get no artifact, no system/driver crashes...just the black screen crash, with no predictable pattern...Like I said can go hours/days with no crash, then get them every 5-15 minutes in a session...

 

Sounds odd. Have you tried a different driver?

 

Been all of the map on drivers, some are MUCH more stable then overs overclocking, right now the new dox 182.50 seem to be the most stable and overclockable that I have found....