Question OCing an old i7-6700k: need some advice to get past 4.5GHz

[coredumperror]

I've got an old system running on mostly 5-year-old hardware, which I'm finally getting around to OCing. I'm trying to eek out a little more life from the hardware, delaying the need to upgrade. But I'm having some issues with stability and heat that, based on what I've read about this processor, don't seem typical. So I'm hoping you folks can help me eek out a few more MHz, while keeping my CPU cool and stable.

Here's my specs:
i7-6700k @ 4.00GHz (base)
Asus Z170-A mobo
2x8GB Crucial Ballistix Sport 2666 MHz CL16 DRAM
Noctua NH-U12A cooler
Phanteks P400A-Digital case
GIGABYTE RTX 2070 Gaming OC 8G

The case and cooler are new, purchased this week because my previous case and cooler were badly set up and too loud. The RTX card was purchased after I fried my 980Ti in late 2018 .

My first attempts at OCing were met with surprising success. I was able to easily get my system to 4.4Ghz by simply bumping the multiplier and doing nothing else. I ran Prime95 torture test, and it worked great for hours, and my core temps never cracked 75C. So I went to 4.5GHz, and had similar results. Temps started climbing into the high 70s, but it seemed fine. So I tried 4.7GHz, which went well enough to try 4.8Ghz. That one finally gave me pause, as my Core 0 temp started to hit 95C, which I knew was way too much. So I went back to 4.7GHz.

In an attempt to figure out how best to drop my temps, so I could try again at 4.8GHz, I read up on how to properly OC i7-6700Ks, and I realized that I'd screwed up: I'd accidentally left my mobo's automatic VCore setting enabled. So the reason those temps got to 95C sometimes was likely due to the mobo pushing VCore way too high in its attempt to make my rig remain stable.

So I started doing it "properly", based on this guide from TweakTown. Particularly the flowchart.

Unfortunately, I got severe stability issues even at the very first OC that the chart recommends (4.5Ghz at 1.3v Vcore). Prime95 would throw a "HARDWARE ERROR" very soon after starting it, and IntelBurnTest would freeze.

After a bunch of experimentation over the weekend, and trying new stress tests, I was disappointed to ultimately settle on a 4.5Ghz OC @ 1.34v Vcore. That would pass 20 minutes of Prime95 (AVX disabled), and 15 minutes of RealBench stress testing (which I read was a better way to do AVX stressing), without cracking 85C on the hottest core (Core 0 is often 3-4 C hotter than the other 3 cores). Higher multipliers would make the system unstable, and higher voltages would make the CPU too hot (>85C at load after a few mins).


Did I just badly lose the silicon lottery? Is it because my CPU is 5 years old? Or is there something I can do to improve this?

Maybe try to apply thermal paste better? I have some Kryonaut paste from Thermal Grizzly, which I got just in case I'd need it. But my Noctua cooler came with NT-H1 thermal paste, so I used that.

Any help would be greatly appreciated!

 

[helpstar]

Which temperatures are you getting without ocing under load?

Is the motherboard´s BIOS up to date on version 3802 ?
reset the BIOS afterwards and try ocing again, following the guide

 

[coredumperror]

Sorry for the long delay. Been really busy.

I managed to get my BIOS set back to stock settings (and yes, it's on version 3802). I also realized that I needed to disable ASUS MultiCore Enhancement to get a true "stock" temp test, since that setting is on by default and applies surprisingly aggressive overclocks and voltage boosts all by itself.

Anyway, with the CPU properly clocking itself to 4.0GHz max, with the stock vCore of 1.296v, I got idle temps in the mid-20s and load temps (Prime95 Small FFTs for 10 mins) in the mid-60s. HWINFO recorded a single instance of 69C, but while I watched, max core temp rarely deviated from 66C.

I'm currently running Prime95 Small FFTs with the BIOS settings and Prime95 version suggested by the Tom's Hardware overclocking guide, and a multiplier of 44, with vCore manually set to 1.300. It's going well after 12 mins, with temps now averaging in the low 70s, except that Core 1 is averaging only 65C.

EDIT 1: The Small FFTs stress test went great, so I'm gonna start a 1hr RealBench stress run and then go to bed.

EDIT 2: Stayed up to finish the RealBench test at 4.4Ghz (44x100.0) and 1.300 vCore. It passed, with temps averaging in the high 60s and low 70s, depending on core. Though it did spike up as high as 79C a few times, and 80C at least once.

The fact that it went to 80C at 1.300v is giving me pause. I am feeling more and more confident that my thermal paste application is imperfect. I'm going to reseat it tomorrow after work.

But for now, sleep.

 

[coredumperror]

So I reseated my CPU cooler, and discovered some stuff.

1. I'm pretty sure I used too little thermal paste, because there was basically none in the middle of the IHS after I took the cooler off. It was almost totally clear, and most of the text was clearly legible.
2. I believe this also indicates that I screwed the cooler down much too tight, pressing all the thermal compound out to the edges of the IHS.

So this time I applied a little more thermal paste (it's a pain to apply the Noctua stuff, because they don't give you a sharp nozzle. It's just a flat hole in the tube at the end), and screwed down the cooler much less forcefully.

Results so far are promising. I saw a 3-4C drop in average load temps during Prime95 Small FFTs (high 60s rather than low 70s, with max at 77 rather than 79), with RealBench also topping out at 77 with averages in the high 60s.

I'm currently testing a 4.5GHz OC, and I'm running RealBench now. After 15 minutes, it's averaging high 60s, with a single, short spike at 80 on Core 0, which is still by far the hottest core under load (interestingly, Core 1 is clearly the hottest when idle).

Much to my surprise, though, this 4.5GHz is stable with stock vCore of 1.300v. In my previous attempts at 4.5, before I followed all the advice on the guide I linked in my previous comment, I couldn't even boot with a 45 multiplier at 1.300v. I needed to go all the way to 1.340 to remain stable during stress tests. But now I'm (apparently) stable at 1.300v, which is awesome.

 

[coredumperror]

Well, I've now given up on 4.6GHz, as it is no where near stable at any voltages below 1.36, and it's too hot and mildly unstable at 1.36. So I'm working on the 8hr tests for 4.5GHz at 1.310v, since that seems likely to work out.

Guess I either lost the silicon lottery, or my CPU is just too old for the good OCs. Oh well, now I at least know that I tried my best and did everything as well as I could. And I've got a 12.5% OC with temps low enough for my fans to remain silent, which is sweet. Plus, I learned a lot for next time!

 

[bishopi5]

TBH running at 4.4 on stock voltage is kinda nice, so you didnt miss out on the silicon lottery, It might even be stable at less volts

 

[coredumperror]

running at 4.4 on stock voltage

Note that I wasn't actually on stock voltage (1.300v). I was on "Auto" voltage, which means my mobo was bumping the voltage up automatically in order to maintain stability at higher clocks.

Anyway, since you brought this thread up again, I thought I'd give a further accounting of my OCing adventures.

I delidded the CPU using the razor method, which went very well. I got a 10-12C temp drop, which spurred me to once again attempt 4.6 GHZ. I zeroed in on 1.400v, and it had seemed to work, surviving an overnight RealBench and and overnight Prime95 Blend with max core temps below 85C, and averages in the mid 70s. But when I started gaming, things went south.

After 3 BSODs during a day-long session in Prey, I had to once again give up on 4.6GHz, dropping back to 4.5GHz at 1.34v. It's working great after another day-long session, so I'm just going to stick with that until I upgrade.

The fact that I apparently need more than a .06v bump to maintain partial stability at just 100 more MHz, with a CPU that people regularly overclock by 200 MHz beyond the highest I could get to even with a delid, tells me I definitely lost the silicon lottery on this one. 🤷

 

[bishopi5]

Haha my bad did not realise how old the thread was, Yeah the 6700k is fickle, mine runs at 4.5 @ 1.268 which is amazing, but i too cannot run 4.6 without close to 1.4v, That old silicon lotterly, Good to hear you had some success with delidding

 

[Burigeller]

Haha my bad did not realise how old the thread was, Yeah the 6700k is fickle, mine runs at 4.5 @ 1.268 which is amazing, but i too cannot run 4.6 without close to 1.4v, That old silicon lotterly, Good to hear you had some success with delidding

I am going with literally the same, i was trying to get the 5.0 mark but it was nowhere near stable. After all the i7 6700k is a terrible CPU for OC. I remember having my i7 4790K @5ghz without an issue with acceptable thermals lmao. Man i was so young and had no idea about PC back then i have no clue how that didnt go sideways lmao. But yea i would not recommend pushing the i7 6700k too far, all in all i can say its not worth the OC since it becomes unstable after 4.8ghz and the going for 4.6 or 4.7 is not really worth the stress just to have it run on 90° under full load for 0.3ghz ,