3570k & Z77, not stable at 45 even with 1.376V

[tracerbullet]

Yep, I do remember that thread, some of my settings research was from it. I may (meaning probably will) continue to mess with it, and RAM could be a good thing to try.

What I've reminded myself for now, and - it may actually work - is that even if I just stick this chip at 43, at something like 1.300 which looks do-able, that's more than a 25% overclock at a voltage that I'm comfortable using and know I can cool. Phrased a little differently that's actually pretty good and something I can live with. And with that I can probably put back all the C-states, the other voltages to stock, and so on "just in case" it turns out something functions better that way (sleep mode or anything).

 

[coffeejunkee]

Yeah, just forget about turbo boost and think 3.4 basespeed, you get almost 1GHz overclock! (ok, might not be that easy to forget).

Vectronic mentions playing around with VTT and PLL but he also overclocks his ram. I'm using Asus P8Z77M Pro and I don't even have VTT option so I'm not sure what it does.

What I do know about voltages: PLL is the clock generator voltage and usually can be lowered while still staying stable. This lowers heat which in turn allows vcore to be lower and still be stable at the same clock. But it's something you work on after getting a regular stable oc.

Internal PLL overvoltage is only needed for booting Windows at high clocks, but after that it's back to regular vcore adjustment.

Changing PCH voltage makes not much sense to me, it controls the southbridge and you're not overclocking that. Ok, so maybe an overclocked cpu puts a little more stress on it somehow, but you already tried increasing it and it didn't help.

VCCSA: system agent + memory controller, increase when running lots of overclocked ram. You didn't change this but you did try running ram at 1333. Maybe it was still too much, so I said try with just 2 sticks, but I don't expect much from it.

 

[tracerbullet]

^^^ Awesome info, thanks.

Hmm, this may need a different thread, I haven't seen it asked: How does one go about figuring out a *minimum* voltage needed? I.e. at idle. I'm thinking of more play time, but in the other direction. Say I stick with a 43 multiplier at the high end, and at that high end I now know how to find out what voltage is required to get it there, and how to check that it's stable.

But changing around LLC, Offset, Turbo Boost - has an effect not only on the top end but lower as well. Idle voltage is going to be in the neighborhood of 1.000 Volt, but if you wanted to see how low it could go, what would you do?

Trying to be more scientific than just not booting or getting a BSOD, is there a test method? I'm thinking of this:

1) Hard setting the idle multiplier (instead of 43 or 45, using 16 or 19)
2) Hard setting a voltage of say 1.000
3) Doing the same tests i.e. Prime95

If it "passes", I know that as long as I have an idle voltage of at least 1.000, things (LLC, Offset, Turbo combination) are fine. Then just lather, rinse, and repeat for other voltages until problems occur.

 

[Kenmitch]

Overclocking can be fun but also makes a person want to pull their hair out at times. I remember trying to get my 2550k to boot windows at 5.7ghz but it crapped out at 56x on my p67 ext4 gen3 motherboard. Wasn't scared of vcore as it's got the optional warranty from intel anyways....Very tough chip it looks like. I was more scared of blowing the caps off my motherboard

So far you've tried the forced overclock style with c-states disabled and fixed vcore only?

Have you tried with c-states enabled and using offset vcore yet? Most of the time works better anyways. At a certain point c3/c6 being disabled helps with random bsods during light loads.

 

[tracerbullet]

I did, actually. I have a lot of that documented in the earlier posts but while it made sense to me it may not have been clear.

I tried low offset with high LLC, high offset with low LLC, and hard set voltages. In all cases, around the 1.376 mark, 45 "ran" but gave WHEA errors every minute or so. Very little difference in any of the approaches (based on admittedly very little testing). But it seemed clear that regardless of how I got there, it was going to take a lot of voltage to ever get stable.

Playing with it still, of course, I did find that the C3/C6 / Package C made a difference after all. It didn't have a noticeable effect up high, but as you mentioned it made a difference down low. When not running Prime95, but instead coming back here or elsewhere just browsing around, with those 3 turned on I'd continue to get WHEA errors a couple times an hour. With all 3 turned off, I've gone several hours now without seeing any.

It kind of seems that it's commonly mentioned 45 is attainable, and that 46, 47 get tough with lots more voltage needed, additional BIOS changes, etc. Looks like for me that 43 is easy, and 45 and higher gets tough. Didn't spend much time to see where 44 falls in there. Probably won't.

As mentioned above, further messing around will probably be trying the low end, minimum idle voltages.