SOLVED Proper motherboard voltages/timings

[mixin12s]

Page Fault BSOD

Okay, so I tried 250MHz again, and again upon stress testing it bsod'd on me giving me a 'page file in non paged area', which from research online happens when it attempts
to move information from the paging file to the unpaged (reserved for essential system functions) area of the RAM. I then tried 245MHz, and during stress testing it gave me
a bsod that read something like "a clock was not returned to a processor core within the required interval" or something to that effect. So, I'm assuming if it is possible to OC
it any higher I have to adjust the RAM timings again? Thinking I should just shut up and be happy with 3.6GHz on the processor... but the 3.75GHz oc that ups the HT Link
back to 2000MHz and the RAM back to 666.7MHz just looks so appealing... lol. I suppose though, a fried motherboard looks way less appealing. I did not however try upping
the voltage anymore on the RAM... QuietIce, you mentioned starting at the default +.05, currently I am at 1.52v (default 1.5v), could upping that more help maybe with the
bsod's I got? EDIT: Upon further research online the bsod I got "clock interrupt was not returned to a secondary processor in the required interval" is most associated with a
bad processor core... others have found if they go in manually and lower the setting on the bad core that they get no more blue screens, but being I'm already oc'd I doubt
the core is bad, it just would appear that one of the cores does not want to go above 3.6GHz, which is more of an overclock than I thought I'd get. So I guess, up the RAM
voltage a bit more and try again, or accept the 3.6GHz as all the higher she's going to go and leave it be? If what I read online is correct however, and it is 1 of the cores that
caused the bsod, then I probably don't want to risk any further damage to said core.

 

[mixin12s]

Finally done

Well... I tried upping the voltage of the RAM to 1.55v, and the processor to 1.45v to go to 245MHz and 250MHz on the FSB... and it still, I'm assuming would've given me the same bsod as earlier, as during stress testing one core stopped with a hardware failure immediately, and I quickly shut Prime95 down to prevent the bsod. Being one of the workers/cores shut down right before the last 2 bsod's, and shut down during this stress testing, it would lead me to believe what I suspected from my earlier research online to be accurate, that one of the cores just won't operate that high. So, I backed it back down to 240 and so far she's stable. Going to run Prime95 blend tests on it tonight for an extended period of time and see what happens although so far I haven't had any problems at those settings. Currently the CPU voltage is 1.42v, NB voltage 1.205v, and RAM voltage is at 1.53v. Thanks again for all the help Overclockers... you guys are great!

 

[QuietIce]

Good luck with the test!

 

[mixin12s]

core

Well, I'm pretty sure that core shutting down on me with those bsod's was the voltage now, and not the core. It did the same thing with the 240MHz FSB overclock of 3.6GHz on the processor on 1.42v, so I upped it to 1.44, and it made it further into the stress testing, but the pc shut down due to the overheat protection I had set in CoreTemp, which I had set at 53C. I then lowered it to 1.43v, and so far it's going through stress testing fine, no core failures as of yet, and temps hovering from 50-53C. I don't want it to go over 55C on the processor core temps, correct?

 

[trents]

Your assumptions are correct. 55 c max. Should be okay up to that or maybe even a tad more in some cases. Won't hurt anything as the CPUs are rated by AMD to be safe at 60+ c, not necessarily stable but safe. Is 1.43 what it set to in bios or is that what the software monitoring programs are saying under load?

 

[mixin12s]

OC

Well, I had to back it off some more, as further into the stress testing the one core still shut down further into it. I hope I didn't damage that core at all, but my temps were never too high, nor does it do it under stress testing at normal clocks, although it does appear to be the same core every time. Regardless, I backed the clock off to one I know made it through the stress testing for an extended length of time and went to 235MHz FSB/3.535GHz proc. Also, I had the voltage at 1.44v in the BIOS, then backed it to 1.43 when it shut down due to the heat protection. With it at 1.43 temps were hovering around 50C. Only reason I assumed the core shut down again was cause I woke up to a bsod on my pc. When I upped the voltage prior to going to sleep and waking up to the bsod, it had been doing the stress test for an hour before I fell asleep, so it seemed the voltage helped it make it further, but it still failed eventually.

 

[trents]

Do you have ACC (Advance Clock Calibration) in bios? It gets most press as the "core unlocking tool" but its original purpose was to be able to tune individual CPU cores for the very thing you're running into. If the same core is failing every time I would look into that. No one knows exactly how it does it but somehow it shores up lagging cores. One caution though, if you enable ACC you will lose the ability to monitor core temps. That's no big deal of you know what the differential is between CPU (socket) temp and core temp. You can figure out the core temp just by doing the math. HWMonitor is a good tool for this because it displays both core temps and CPU temp.

 

[mixin12s]

Core

I think I may have ACC in the BIOS... however with the voltage at 1.45v and running at 3.6GHz like it was, I was hovering up near 55C. I did back the clock off to 235MHz FSB for a processor speed of 3.525GHz and put the voltage back down to 1.42v on the cpu. Cinebench only gave me .18 higher of a processor score, and only 2fps higher on the openGL test for 3.6GHz from the 3.525GHz score. Being 3.52 meets more than my needs for now in a processor, I don't see the value in adding the extra heat to my processor by having to have extra voltage, to achieve that little bit extra. Thanks for the advice though, I'll keep that in mind should I ever decide I want to try to take it further.

 

[trents]

That's just it. No one is sure how the ACC does its thing. Its not clear whether it adds juice or just changes a timing or what. It may or may not produce more heat. At any rate, you are probably about at the top of the OC for that CPU anyway.

 

[mixin12s]

heat

I was referring more to the added heat of upping the voltage from the 1.42v I have it at on the current overclock to 1.45v. When I was running at 1.45v before with that overclock my temps were 53-55C so to turn on that feature without being able to monitor the temps, and then upping the voltage back to 1.45v, to get at most .75GHz more doesn't seem worth the risk and the chances I'd be taking. Also, currently I can still have all 5 of my case fans on low which with how much I run my pc will probably save a lot on the electric bill and the life of the fans, at 3.6 I had all the case fans on high. Idk, the only time it fails at 3.6GHz is when stress testing with 100% on all four cores... it makes it through benchmarks and runs everything fine. Think if it made it through stress testing fine with just 3 workers running it'd be fine? Never am I doing anything that would come close to pulling 100% from all 4 cores except benchmark/stress testing, and even then it makes it through the benchmark tests fine.

 

[trents]

Think if it made it through stress testing fine with just 3 workers running it'd be fine?

Try it. National security doesn't depend on it, does it?

 

[mixin12s]

Interesting... I'm going to try the higher overclock again without spread spectrum enabled. I just read having that enabled on an oc'd processor can cause the processor to lock up.

And... it didn't work. It still bsod'd on me. I think it's the processor voltage as when I up it to 1.45v from 1.43v it makes it further into the stress testing before crashing on me. The bsod this time said nothing about a clock interval anymore either. It just said it had a hardware error that it was unable to correct. Anyway, I think I'm finally done playing now, lol. Just wanted to thank you guys again for all the help.

 

[trents]

If it went longer into the stress testing with a small voltage bump that's a good indicator you need more vcore. I've heard that stuff about spread spectrum too but it never made a difference for me. What is your CPU-NB set to?

 

[mixin12s]

CPU-nb

I had the vCore at 1.45v(1.40v default), and the CPU-nb is at 1.205v (1.185v default). RAM I have at 1.55v (1.50v default).