Overclock.net banner

1 - 8 of 8 Posts

·
Registered
Joined
·
7 Posts
Discussion Starter #1 (Edited)
So I've been working on OCing my old i7-6700K over the last few weeks, and I've got a stable OC at a mere 4.5GHz. Hoping to get more soon, but after all this work, I realized that I don't actually know what I should actually care about when it comes to CPU temperatures. I've heard "Try not to go beyond 80-85C", but which reading needs to be kept to that level? Max individual core temps? The "CPU" temp? "CPU Package" temp? Nothing I've read has made that clear.

Here's a screenshot I just took of HWINfo64 after running Prime95 Small FFTs for 15 minutes:


Which of these readings are most important to me, as an overclocker? And should I care about occasional maximum temp spikes, or overall average temps?

Thanks!
 

·
Overclocker in training
Joined
·
12,951 Posts
Hi,
Max core and cpu package and what ever tj max is and stay under that by 15c

Vcore and current/ min/ max readings which you're showing no difference so you must be using manual override probably not a good thing for 24/ 7 but fine for short term benchmarking.
 

·
Registered
Joined
·
7 Posts
Discussion Starter #3
Oh? I've never heard that manual voltage override was bad for 24/7 OC. Why is that? What should I use instead?
 

·
Overclocker in training
Joined
·
12,951 Posts
Oh? I've never heard that manual voltage override was bad for 24/7 OC. Why is that? What should I use instead?
Hi,
Well not on max clocks
But if you get a lower everyday clock nothing wrong with manual voltage
I usually shoot for whatever clock maxes at 70c which is far below tj max plus might work well gaming/....

Max clocks are mainly good for benchmarks so little different that any game would require.

Plus enable c-states and balanced power plan in os or adjust a performance power plan so cpu drops to 5% on idle or non activity. so your vid or vcore drops and reads lower than max reading.

Load line calibration LLC also what is being used ?
Try and use llc-4 or 5 so it allows some vdroop in vcore/ vid readings.
 

·
Registered
Joined
·
2,722 Posts
There's nothing wrong with it really, and I presume you opened that monitoring program after you started running the stress software which is why the minimum and maximum are the same? Due to Vdrop/Vdroop, you'll likely have load voltages a bit lower than idle anyway. From what I remember, you want this, as otherwise you'll be getting transient spikes above your set value during sudden CPU load increases. The danger of voltage is linked to temperature, so there is no harm from having a slightly higher voltage at idle than you have under load. That being said, different CPUs and different platforms behave differently (for example, my current my Ryzen does seem to have idle voltage at lower than it does under load, and I haven't changed it's behavior any over stock/default), so this isn't an absolute.

Your voltages seem fine (not 100% sure on Skylake's unofficially agreed upon limit so someone correct if wrong, but I think 1.3V isn't excessive for it; I was running about that on my Sandy Bridge for a mere 4 GHz) and temperatures are high but okay. I personally wouldn't push it much further myself, but if you like to live dangerously, you probably have a little bit of headroom accounting for the fact that most games shouldn't load the temperature as much as that stress program is. If that's stable though, and you don't see temperatures higher than that, I would (personally) say that's good.
 

·
Registered
Joined
·
7 Posts
Discussion Starter #6
> I presume you opened that monitoring program after you started running the stress software which is why the minimum and maximum are the same?

Yeah, to get a realistic load average, I reset HWINFO's log after I started the Small FFTs torture test.

Here's what HWINFO looks like at idle (after running for a while as I tried to get my CPU to calm down):



> I think 1.3V isn't excessive for it

I've been reading up a lot on this, and people have gone up to 1.424V (as displayed in Windows) without issue on 6700K. Though I personally seem to have lost the silicon lottery, as I cannot get long-term stable above 4.5Ghz, even at 1.42V (at which my cores just barely squeek by under 90C at load, which is too hot). That's why I'm going to delid today, and hope that the temp drop will aid with stability at higher clocks.

> enable c-states and balanced power plan

I'm using a balanced power plan, and I'm fairly sure C-states is enabled, though I haven't checked my BIOS in a bit. I think that's on by default, though.


I believe LLC is level 4, or possibly "Auto". I've used both in recent testing, and I'm not sure which one it's at right now.
 

·
Overclocker in training
Joined
·
12,951 Posts
Hi,
Policy is if you can't afford to replace it don't max oc it 24/ 7 lol
Looks more reasonable atm.
Think tj max is 100c on this chip.
 

·
Registered
Joined
·
7 Posts
Discussion Starter #8
I only OC as far as I can still remain stable after an 8-hour Prime95 Blend test and an 8-hour RealBench test. I've been "stable" with decent temps at up to 4.7GHz, probably fine for gaming, but torture tests always crash after a while. So I don't keep those as 24/7 OCs.
 
1 - 8 of 8 Posts
Top