[Official] Ivy Bridge Owners Thread - Page 220

I have been keeping a spreadsheet with the OC attempts I have done today. I have omitted any redundant stuff, so if something doesn't work on the sheet, you can assume that the voltages below it don't work either. I have them in mine being that I picked a starting point and if I didn't get it stable, I increased until I got a boot and those are in the sheet.
Spreadsheet (Click to show)
I have it color coded for ones that passed. The ones in black failed. Note that a TON of sets ran fine for the original 10-15 minute tests I was running but they ALL crashed at 25 minutes of Prime95. The only times Prime95 failed, it gave me a "Prime95 has stopped responding" and then I had to exit the program. ALL of these occurred at 25 minutes too. I was using a custom blend with 90 percent of available RAM and running the FFTs for 2 minutes each.

For the max voltage, the lowest number is the minimum voltage I saw on load, the middle is the most common voltage I saw, and the highest is the absolute max. The voltage fluctuated a bunch at load, is this normal? I noticed that WHEA errors coincided with the times that the voltage would be at the low end of the load values. Is there any way to make it fluctuate less? Note that I started doing this later on in trials, the earlier ones just show max or the range.

For the times, if it was stable, I wrote the time it was stable for, and the time I ran each FFT.

I stopped at 45 for a multiplier because I didn't want to push too much voltage through this chip, as I would like it to last a few years. What is a safe voltage to have it on load for it to last that long?

Where would you guys go from where I am now? That last run actually just finished and ran 30 minutes without problems.

I know there are tons of questions in this post, but I would really really appreciate it if you could answer them all

Thanks again,
Big Al

I'm on my phone...but
Low load voltage and you get whea errors but load voltage is stable. You need less LLC. This will raise the low load voltage.

Safe voltage is generally up to 1.35 on air but more so if temps stay below 90 during stress run you are ok.

I am at 46 multi, medium LLC, with offset 0.05, 1.26 ~ roughly load voltage for a 4.6 oc just for an idea.
My temps max about 75 during a stress run.

The higher the frequency the higher the vid voltage will be so offset settings will change as you clock higher.

Thanks for the help. I don't think I have gotten one WHEA error at idle, and when I get a WHEA error on load, I increase the Vcore until they disappear.


So you're saying there is no real way to make the vcore fluctuate less at load?

+rep!

I think it's normal for vcore to fluctuate at load. For example, my current settings: 4400, -0.020; LLC Medium (25%); PLL Auto. I only tested prime95 for around 12min (5min per FFT) and the load vcore is 1.200/1.208/1.192/1.184/1.176. But it'll generally settle on two (very subjective) voltages (see my bold voltages), so I use them as the vcore for comparison if I change any settings.

(Btw, my vcore also fluctuates a lot at idle but I don't know if my background programs or services are affecting them or not, so I generally just write down the two lowest vcore I see for data.)
Edited by Teiji - 6/21/12 at 8:28pm

Got ya. Thanks a ton for the help!

Is it possible that WHEA errors can be caused by my memory being unstable?

I was getting repeatable WHEA errors on a bunch of settings after I lowered the voltage and VTT on my RAM. Now that I upped it again, the WHEA errors are gone. I only ask because nearly everyone has said that WHEA errors are caused by a low offset (which in my case the offset didn't help.)

The only way I know of to get rid of load voltage fluctuation is to use manual vcore as opposed to offset. I personally prefer it that way, but that is because I'm coming from LGA775 prior to this, and overclocking was all direct volts.

WHEA errors can be caused by a multitude of things, they are simply a precursor to a BSOD, except Windows was able to recover from the fault. In your case, perhaps your WHEA's were a result of too low of a voltage on the memory.

Cool beans. Thanks a ton for the help +rep.

Is there a way to do manual vcore, yet have the voltage and multi drop off at idle?

just keep cstates on etc
Im using manual with all c states think i finally got mine stable

So does your chip lower it's voltage and clocks on idle?