Glitchy OC with 2600K - What the bleep is going on?

Right, maybe someone can help me because I am TOTALLY confused. Setting a voltage on "Auto", ALWAYS sets the voltage higher then a tweaked voltage right?... I mean if it sets it at 1.35V you can almost guaranteeto get it down to 1.3V with the same stability?

So... I try overclocking my 2600K, wanted to use the offset voltage and was experimenting with LLC.

I had:

LLC: Ultra High (75%)
Offset Mode: - 0.10 (I think)

Basically, I had to get a vCore of 1.37V with LLC at 75% to be LinX stable at 4.5GHz.

I thought 'holy cow, that's a lot of vCore compared to everyone else!'.

So, I then decided to boot into 4.5GHz on "Auto" voltage, No LLC and EVERYTHING else left on auto... low and beyond a LinX load vCore of 1.32V... Huh? Then I tried 4.6GHz... auto 1.36V.. HUH!?

So, long story short... ASUS BIOS (Or UEFI should I say) makes a voltage of 1.36V~ with 75% LLC FAIL at 4.5GHz but allows a nooby auto OC with 0% LLC and a vCore 1.32V to PASS.

Feel free to unconfuse me!

Does anyone have the slightest idea as to what in earth is going on?

Ah, nevermind the LLC thing, seems that 'auto' actually = Ultra High (75%). even so the volts don't add up.

when you changed the Vcore did you leave everything else on auto?

also, not sure about your setup but when I use auto voltages, (1156) vcore is variable dependent on load, so if your measuring in bios when there's low load, your voltage might be a little less than is needed for stability.

cant think of any other reason at the mo. hope this helps.

ScratchFive said:

when you changed the Vcore did you leave everything else on auto?

also, not sure about your setup but when I use auto voltages, (1156) vcore is variable dependent on load, so if your measuring in bios when there's low load, your voltage might be a little less than is needed for stability.

cant think of any other reason at the mo. hope this helps.

Click to expand...

I did leave everything on auto yeah, but I found out what the problem is now:

When using Offset voltage, quite a bit more vCore is required. For example I could clock to 4.6Ghz with 1.35V manually, which under Ultra high LLC would drop to 1.336V at full LinX load and never fell below that. At idle of course the voltage remained at 1.35V.

Using offset voltage however I needed to pump more vCore to attain the same results. What with varying vCore, to make sure that again, it never fell below 1.336V, I had to pump more to the offset resulting in peek voltages of 1.368V. So although I save voltage using offset it requires about 30mv more headroom. What is strange is that the manual voltage of 1.35V, fell to 1.336V under load and stayed there. Using Offset the voltage fluctuates at load from anywhere as low as 1.328V to 1.368V. ridiculous 40mv fluctuation is the price you pay for using Offset voltage. It's still more energy effcient but it sure as hell needs a BIOS patch to stabilise it. Either that or more detailed/strict settings to stop it from overshooting. It's a shame because I don't need more than 1.336V to maintain a stable 4.6GHz.

Wow this gets better and better. Offset Vid of 1.38V with Ultra High LLC at 4.5GHz = LinX crash after 3 rubs. Everything Auto = Offset Vid 1.37V with no LLC which results in a vDroop to 1.30V and it works fine.

The only thing I think it could possibly be is C3 and C6 states. I've noticed that LinX crashes not during a string but at the end, as the voltage changes. For that reason I think there isn't enough voltage to maintain C3 and C6 states and so it crashes. Using Auto it seems to switch those states off as the voltage doesn't drop past 1.012V.

At last! We are getting somewhere! I was right, C3 and C6 states were causing the random crashes. Changing them to "Auto" (I previously had them enabled) I am now at 4.6GHz with about 1.36V using offset (I think it's Ultra High LLC with -0.020), which droops to about 1.33 > 1.35 under load. I have now done 100 runs on LinX with AVX and it's 100% stable . I might try for 4.7GHz now. I will also lower the CPU PLL, I got it down to 1.4V LinX stable so I will try that again, and then see if I can lower the vCore after as most say by lowering PLL it allows for lower vCore.

I had to turn C3 and C6 states off for my OC.