Lots of questions OCing Skylake 6700K on MSI M7

[Mordachai]

Hey - haven't been back in a while, really since I got my 2500K stable at 44x for several years (damn fine chip!).

Now I've taken the plunge to 6700K + Z170, and I'm glad I waited, as this seems like a goodly jump and the OCing is easier than ever!

But I've got some things that are messing wit me mind:

When I run any stress tests, my package temps keep doing a see-saw pattern, jumping ~20c over the duration (up & down, up & down, ...) Not just jumping up and then plateauing or slowly creeping up from some plateau value - rather jumping up, and then see-sawing. My cooling is a Captain Watercooler 120. I've played with the fan controls for its fan, so that it cranks up the cooling fan sooner... but essentially I see the same pattern (maybe the overall temp is better, but that see-saw effect is plain as day).



As you can see, nearly 20c jump up & down in temps under constant load.


The second oddity is the vcore. According to XTU (pictured above) the vcore is around 1.2290v - this is suspiciously constant, and disagrees with both the MSI Command Center and CPU-Z. During the above test, I had locked it at 1.36v in msicc - and cpu-z agrees - it is currently 1.360 +- .008v.



QUESTIONS:
1. Have any of you experienced these see-saw temp patterns?
2. Any guesses on what's causing it?
3. Does it strike you as a hint that the water cooler isn't functioning properly? Should I be concerned?

4. Is XTU just known to be totally inaccurate for the vcore? Or is that value measuring some other obscure value?
5. What is the current consensus on "24/7 max safe vcore"?
6. If I set the vcore in BIOS or msi CC, I assume it no longer varies with load (as VID does)? (Is that a problem - I liked the old vOffset so that you used VID +/- offest)?

I find it bizarre that Intel's XTU gets the vcore wrong... but I see no other voltage monitoring value in its settings / options that makes better sense (and vcore is shown by default).

I have turned on charting the various throttling values - in case throttling was responsible for the see-saw. However, the chart showed flat-lines on those, so seemed either not to be triggered, or i don't know what to look for.

Thanks for any advice / insight you have to offer. Fun to be OCing again, for a new rig.

 

[Woomack]

100% load showing by OS doesn't mean that CPU is constantly loaded in 100%. Test changes and also load changes so temps are sometimes higher or lower. I can only tell you it's normal. There are tests which will keep CPU at real 100% load but then you won't be able to even move mouse. Tests like that are pretty pointless for gaming/home PC.

~1.4V is reasonable max for these cpus. They're pretty fresh so it's hard to say what will be max safe voltage. I wouldn't believe in voltages declared by Intel as they're not testing that on overclocked hardware.

 

[EarthDog]

Doesn't the M7 have voltage read points on it? Use those to confirm your voltages. CPUz (software in general) sometimes does not like to read the correct voltage.

 

[Mordachai]

Good point about it not being truly 100% loaded - but do you experience 20c see-saw patterns under XTU's cpu stress test? Some variation makes sense, the magnitude of 40, then 60, then 40, ... just seems outrageous to me (which probably just shows my ignorance!)

I'll break out my voltage meter and see what it says. I just figured that folks here would have run into this on the new 170 mobos if anyone had. Or that I was just missing some basic piece of info (oh, XTU is reporting core voltage, whereas CPU-Z is VID)?

PS: Love Banksey! And those who work with him - great folk, great art - visual poetry!

Thanks for the feedback- much appreciated.

 

[Woomack]

On 6600K I have +/- 15*C in XTU but I was testing mainly 4.2GHz so at higher clock/voltage difference can be bigger. It's like voltage/temp readings have some delay and CPU is never constantly loaded so for short period of time it drops to 0% ( even though you still see 100% ). You see that temps/voltages are updating every 1 second+. In this time CPU could finish one thread and start another.