Originally Posted by Ascii Aficionado
2 hours during my initial test, no more than 10 minutes for subsequent tests. judging by how everything's performed it's been "stable" every time. Though I'm aware most other people test significantly longer. What's important to note is that when the x124 BSOD does occur every 2-3 months it only occurs when the system is idle
, which is what I find odd. I have literally never had a BSOD while doing anything on the system except for one time about 6 months ago.
What does indicate that it's degrading is that after a BSOD does occur that Vcore is no longer sufficient to run P95 for more than a single minute, and I have to give it a .010 V increase then everything's back to being fine.
I tried the Offset voltage and it went up to 1.429 V during P95 so I killed it within seconds as I thought those are the dangerous levels.
Also, I was under the impression that using some others settings such as power saving that my idle voltage was significantly lower, which it apparently isn't.
So, I was using an Offset Voltage of + 0.0000 which is the default one, what's making it jump up to 1.429 V ? Which is also what people warned me about when I first started to try my OC.
The reason for 1.429V is because this offset voltage is an Automatic one. Try +0.005V and increase it from there.
Idle BSODs usually occur when the idle voltage is too low in order to maintain stability. Instability is also stressful for hardware, so this may be why degradation could be occurring. 2 hours of Prime95 is nowhere near enough to determine whether or not a system is stable, so I'm willing to say that it hasn't been stable yet to date even though it usually seems like it's stable. I mean if it were truly stable, then these problems wouldn't (or shouldn't) be happening, especially with your system. It should be smooth sailing.
A minimum of 12 hours of Prime95 is pretty much required in order to begin determining whether or not a system is stable, especially at bigger overclocks like this one. I will always recommend 24 hours because it takes about 24 hours for Prime95 v27.7 build 2 to get through all 82 of its FFTs in the Blend test.
Getting back to the idle voltage being too low for stability, this is also usually caused by the idle voltage being lower than the full-load voltage, and this is usually caused by using way too high of a Load-Line Calibration setting. Using too high of an LLC setting can also be dangerous on lower quality motherboards because of the micro voltage spikes that occur when load starts on the CPU and when load stops (it's kind of like a spring or a rubber band or a guitar string getting plucked - it spikes and then settles down). Too high of an LLC setting is also much more stressful on the VRM which also results in higher VRM temperatures. So, it may be possible that your VRM is degrading. It's hard to tell at this point.
So, I recommend lowering the LLC a notch or two and then adjusting your core voltage in order to compensate. It's safer to have the idle voltage being slightly higher than your full-load voltage. This also helps avoid idle BSODs entirely. Speaking of which, if you decide to use an Offset voltage, then disable C3 and C6 because otherwise you will get idle BSODs or lock-ups.
Ultimately, I do recommend working on making your system so stable that it can run Prime95's Blend test for an absolute minimum of 12 hours. I say that if you don't want to do that or you can't, then you may end up having worse problems down the road, like data corruption of irreplaceable data. It's like lightning: you can't predict when or where it will strike.
Edited by TwoCables - 9/30/12 at 2:29pm