Sign in with
Sign up | Sign in
Your question
Solved

Single Prime95 error?

Last response: in Overclocking
Share
May 29, 2010 2:40:43 AM

Last night I had my q6600 to approx 3.201ghz. I'd been slowly upping the voltage until one happily let me into/run windows. I ran Prime95 for at least 2 hours before getting my first error. It was on the very last core, which is kinda wierd unless it's a crap thread because despite Prime saying all 4 were @ 100% load, the last 2 were always around 3 degrees cooler (celcius) than the first two. It was only a rounding error (and the number was close to target). A google search showed this was likely just the voltage being a whisker too low. I upped the voltage a notch and tried again. This time after the same length of time I got a different error (same core). This was a FATAL ERROR: XXXXX expected answer did not equal XXXXX (the usual message, sum didn't come back expected answer).

Does this HAVE to mean I just can't push it that far, or could the voltage still be too low? I can't play too much more with voltage because under full load I'm getting up to 58-61 degrees (though after a cool night it's sitting in the background testing again under full load at 52-55 degrees).

I still have a couple of heat reduction options. I've ordered some compressed air/electrical cleaner and Artic Silver 5 thermal paste and I plan to rip out the Big Typhoon and give it a nice warm bath. I stupidly smoke at the the PC and it has some melted on smog.

I was only running small ffts and I ran memtest86 4.04 for 10 hours last night. 0 errors.

Anything I can do to also rule out North Bridge?

More about : single prime95 error

May 29, 2010 10:34:56 AM

Lithocardia said:
Last night I had my q6600 to approx 3.201ghz. I'd been slowly upping the voltage until one happily let me into/run windows. I ran Prime95 for at least 2 hours before getting my first error. It was on the very last core, which is kinda wierd unless it's a crap thread because despite Prime saying all 4 were @ 100% load, the last 2 were always around 3 degrees cooler (celcius) than the first two. It was only a rounding error (and the number was close to target). A google search showed this was likely just the voltage being a whisker too low. I upped the voltage a notch and tried again. This time after the same length of time I got a different error (same core). This was a FATAL ERROR: XXXXX expected answer did not equal XXXXX (the usual message, sum didn't come back expected answer).

Does this HAVE to mean I just can't push it that far, or could the voltage still be too low? I can't play too much more with voltage because under full load I'm getting up to 58-61 degrees (though after a cool night it's sitting in the background testing again under full load at 52-55 degrees).

I still have a couple of heat reduction options. I've ordered some compressed air/electrical cleaner and Artic Silver 5 thermal paste and I plan to rip out the Big Typhoon and give it a nice warm bath. I stupidly smoke at the the PC and it has some melted on smog.

I was only running small ffts and I ran memtest86 4.04 for 10 hours last night. 0 errors.

Anything I can do to also rule out North Bridge?


ur CPU voltage are low, Increase ur voltage to the next level and then test it with prime95, If the same error comes again with 0 load on the last core it means ur voltages are still low, try upping ur voltage to the next level again and then test, do this until u find it stable on prime95.
May 29, 2010 1:58:19 PM

What do you mean with a 0 load? Are you talking about the 'worker' stops after the error? All cores are under 100% load during the test. The core's 'worker' only stops (not sure if it returns to 0 or keeps doing normal background stuff) after the error.

Why are you so sure it's voltage? I'm not arguing, just trying to learn. For example, if I had the world's most awesome cooler and could set this to 1.5v (max according to Intel & BIOS) would you expect it to work up to something like 5ghz?
Related resources
May 29, 2010 2:38:38 PM

Lithocardia said:
What do you mean with a 0 load? Are you talking about the 'worker' stops after the error? All cores are under 100% load during the test. The core's 'worker' only stops (not sure if it returns to 0 or keeps doing normal background stuff) after the error.

Why are you so sure it's voltage? I'm not arguing, just trying to learn. For example, if I had the world's most awesome cooler and could set this to 1.5v (max according to Intel & BIOS) would you expect it to work up to something like 5ghz?


Ya am talking about the worker stops after the error, any particular worker will stop as soon as the error comes on any particular core and this happened because of low voltages so u need to give the next voltage level again in the BIOS and then re-run the prime, if it still happen again, then again go to BIOS and give the next level of voltage. Do it until u have prime95 stable.

No Air-cooler will let u go to 5 GHz, Max u can expect upto 4 GHz or 4.2 GHz on any world's top air-cooler.
May 29, 2010 3:29:53 PM

What I meant was, wouldn't the top limit of at least one core whack out before heat became the primary problem?
May 29, 2010 5:55:20 PM

Lithocardia said:
What I meant was, wouldn't the top limit of at least one core whack out before heat became the primary problem?


No
June 2, 2010 2:46:24 PM

Oh, in that case cool, ta.

I'm sitting at 3.2ghz with what I think is acceptable temps (65% max on one core under prime95 testing). To get that result stable I bumped up the voltage as you said. It sits in the mid-30s temp wise when doing normal tasks.

Best solution

a b à CPUs
a c 100 K Overclocking
June 2, 2010 3:48:38 PM
Share

It's definitely voltage - whether vtt or vcore however is hard to say. Vtt is more dependant on what base clock you're using. It can often be left at auto if your mobo had auto adjustment. Vcore however needs to start to be increased for stability at a certain point (ie: beyond the max efficiency area, you start to exponentially need more vcore for a little bit more speed).

I just got my system stable Prime95 for about 8.5hrs large FFT. I ran it overnight first, but there was an error after about 2 hrs. I upped the Vcore 1 notch (0.00625V). Ran again and got an error after about 3hrs. Upped it one more notch and it ran stable for the 8.5hrs. Also, those slight bumps in voltage resulted in almost no increase in temp. Tops 1C. My i5 750 is now running at 3.7ghz/4.2ghz turbo at 1.30625V Vcore, 1.20625V Vtt (ASUS P7P55D Pro mobo). Interestingly, on my old mobo (MSI P55GD65) I ran the same base clock and settings/speed but the voltages were 1.288V Vcore and 1.197V Vtt. So there's a good example why you can't just copy and paste other people's voltages (not that you were but in case someone reads this and tries to)
June 12, 2010 10:06:30 AM

Best answer selected by Lithocardia.
!