Clocking i2500k and voltage spikes

strumpusplunket · 14 Dec 2011 at 23:07

Had my i2500k/Asus P8Z68-V/gen3 setup for a couple of weeks now and still really sussing out the new methods for overclocking, taking things slowly and so on. Still, one thing has had me a bit concerned.

Basically, I started clocking using offset voltage (so the vcore could drop at idle) and left it on auto. I worked my way to 4.2GHz in smallish steps and all seemed well with idle and load voltages and temperatures.

But then I noticed something odd. Whilst full load volts were okay (1.30v load), there were also weird spikes just as the system went from idle to load. Even if there wasn't time for the clocks to ramp up, the voltage still would. At 4.2GHz it would spike at 1.36v! At 4.3, 1.368v! I didn't want to push much further.

I'm now using high LLC and manual negative offset voltages (-0.015 currently). This has taken the spikes right down to 1.336v, which is okay, but load volts are now higher and I clearly can't push the negative offset much further before the idle voltage drops too low.

Anyone else noticed this weird idle-to-load brief vcore spike? Is this a potential reason so many SB chips died? Is it something to be concerned about?

I wouldn't even have noticed but for coincidentally glancing at CPU-Z at the right time. Anyone found any other solutions?

meths · 14 Dec 2011 at 23:41

It's probably to do with LLC, especially if you've got it on the highest setting.

Try taking it down a notch and see what happens.

Also 1.368v is okay if your temps are fine. IIRC max core voltage according to Intel is 1.52v.

strumpusplunket · 14 Dec 2011 at 23:55

Yes, 1.368v wasn't worrying me in the sense of thinking it would damage my CPU in itself.

It was more that this was happening only at 4.2GHz, so at 4.5GHz plus... well, it's not an ideal situation to be pushing, I guess.

Oh, and sorry for the ambiguity (trying to say too much at once I think...

), but actually I didn't have LLC on at all for these tests.

I've only just started experimenting with LLC enabled (just tried a couple of IBT runs at lowish clock speeds before my first post).

LLC pushes all the voltages up, including the idle-to-load spike. I'm currently experimenting with negative offsets to bring this down, and it has come down somewhat, but now my idle voltage is 0.976v, which is probably on the border of stable idle vcore so I don't know how much lower I can go there.

I'm being cautious, I know, but I was curious if anyone else saw these brief idle-to-load spikes (oh, these spikes are also there on loading Windows and last until all the startup apps are loaded).

Oh, also is they're going to be anything to worry about if the spikes get up to 1.4v plus. I mean, it's only for a split second, but it happens every time the system goes from idle to load, so I don't know about the long term effects...

bedazzled · 15 Dec 2011 at 01:28

My 2500k is runing at 4.4 GHz with 1.192 vcore under full load, idle at 0.936v on an Asus p8z68-v pro board

LLC = regular
offset mode
-0.045
everything else on default settings

set LLC to regular and increase the neg offset until windows fails to load, then reduce it a notch or two. Test under full load and if prime95 fails increase LLC a notch and test again. Once stable slowly increase the neg offset again until prime fails then decrease it a notch.

strumpusplunket · 15 Dec 2011 at 08:32

Yeah, like I say, I have been experimenting with negative offset voltages and different levels of LLC.

However, it would seem your chip is a very nice clocker compared to mine. I need 1.304-1.312v under load to be able to Prime at 4.4GHz. This with offset -0.015 and High LLC. If the vcore drops any further I get BSODs.

It does take care of my spiking problem, though. It now only spikes to 1.336v, which doesn't trouble me and leaves some headroom.

However... just had a BSOD after an hour and a bit of Prime at 4.5GHz using high LLC and -0.020 offset (which gives me 1.312-1.336v load - it seems to BSOD on dropping to 1.312v).

So, it looks like I won't get much further with this chip, I don't think - not if I want to keep these voltage spikes under control and have vcore dial down at idle, anyway. Reckon once I hit 4.5 or 4.6GHz, the voltage spikes will be up in the mid-to-high 1.3x range once again even with the negative offset, which is where I want to limit it. Ideally I'm looking to keep vcore consistently under 1.36v or thereabouts, and I'm not comfortable with periodic, albeit very brief, spikes to that level.

Mostly I think I'm surprised that I've not seen more written or discussed about these periodic transitional voltage spikes with offset voltages... I was hypothesising to myself that perhaps some instances in the early days of people frying their SB chips was due to the voltages looking okay at idle and under load, but these transitional spikes going unnoticed (I only just noticed them by accident/coincidence) and passing a dangerous enough voltage surge to degrade the processor?