Weird OC problem with 6700k and asus z170-a

[Ipso]

Hi,

I'm rather new to OCing but I think I have the basics down and succeeded at a modest OC, but then it started acting strange, which is why I'm asking if someone might know what's going on and what can be done.

First a sidenote, the first mobo I got acted up (dimm slot 1 stopped working after a week) so I returned it and have a new one now. I tried swapping dims in different slots in the old one and the dimms seemed to work, it was whenever slot 1 was populated it refused to boot with dim-warning-light lit. So I don't think that matters but thought I'd mention it in case.

The rig:

CPU: Intel I7 6700k
Cooling: Kraken x60 with fans swapped for Noctua NF-A14 PWM
Mobo: Asus Z170-A (Aisuite 3 and BIOS updated to latest version from their site, aisuite from late august, bios v 1003)
Mem: Corsair Vengeance LPX DDR4-3000 QC 32GB
PSU: Corsair AX760 760W
Graphics: GB AMD 280X 3GB

Let me know if you need any other info!

So what happened...

I aimed for 4.6 ghz with seemed what most can achieve at least, and it was pretty easy, I don't quite understand the adaptive/offset mode parameters, so I went for manual mode with 4.6GHz@1.35V and XMP for the RAM. With was fine at least for 20 min of Prime 95 torture, and a few hours of surf/gaming etc for about a week. (Those settings worked at the first mobo as well).

Then suddenly yesterday while I left it idling in win 10 for a couple of hours, when I returned, it hanged, and when booting the problems started: Almost any OC seems to not work mostly at boot, but if boot works, it mostly works afterwards. I now run at stock CPU and JEDEC 2100mhz for ram just to get it somewhat stable, and even so it got a weird bs after an hour or so yesterday.

So what I've tried
- After trying some minor changes and failing, I updated the BIOS successfully - no change, still oc problems
- Once, somehow, I got 4.6 working@1.39v (I've tried up to 1.42), it passed some prime95 and an hour surf/game, but then failed at boot next time and didn't work again!
- Tried loading default bios parameters and starting over, nothing new
- Temps are fine, 30c when idling (liquid temps seems to be 30 for some reason so the cpu wont go below that) and below 80-85c when running the "hard parts" of prime 95 torture, and 50-75c for "easier" parts, otherwise perhaps 50 during gaming (in games that use the cpu somewhat, less otherwise)
- Also trired to OC the RAM and leave the CPU but I only got that back once at XMP 3ghz (again for at least a few stable minutes and some testing), with a slightly increased ddr voltage (1.36 vs the xmp 1.35), but after that never again, raising it ti 1.38 doesnt help, tried loweing frequency to 2800 and 2600...no good.

I haven't tried say CPU 4.4ghz, but when it pulled of 4.6@1.35v for about a week...what's wrong now?

Running everything on "auto" in bios now, 4ghz with 4.2 turbo, and I may be mistaken, but it's now running at 1.4v and 1.45 at prime torture, and I think it was lower before...

I read in a few threads here about some problem with some Asus motherboards having some problem with lowering voltages after sleep, but I trired disabling c-states, and for some that was fixed in with bios, and that the z170-a hadn't gotten the fix yet, but that was 15 sept, and I think the current bios is newer (the fix may not be included of course, no mention of it on the description.)

But the weird thing is that is so inconsistent and worked a week, then acted up, but worked at oc settings a one or two times, then wont boot next time? Testing settings back and forth, suddenly it worked again using 4.4ghz@1.4v, then i tried 4.6@1.4, also ok (when I say worked I mean boot and about 10 min of prime 95 torture...usually it stops working before that if there's a problem..)! Then reboot to change to lower voltage in bios, and it doesn't even boot using the same parameters?! Back to 4.4, doesn't boot again!? I'm really at a loss here...memory at low speed all the time...

In fact, it seldom stops working during the 20 min prime 95-tests, mostly it doesn't even boot, and rarely bluescreen, but a few times now, a setting that passed a 10 min torture and some surfing wont even boot on next reboot...what's up with that, is the reboot test that thorough or is there a problem in the bios settings for booting somehow?

Now I finally managed to get the memory working at almost xmp, but lowering frequency from 3 to 2.933 and increasing voltage 0.01v to 1.36 (tried it before as well but didn't work then..)...but still stock auto cpu 4-4.2ghz@1.4-1.45v...

Any ideas?

[Menthol]

Is it stable at all default settings? After so many blue screens and OS errors your likely going to end up with a corrupted OS and need a fresh install
I would suspect it's memory related because your modules are more than likely designed for the X99 platform, I had a set of Gskill 300mhz that were for the X99 platform that also would not run at XMP 3000 but would be set to 2933 by AISuite.
A set of ram that is for Z170 is the best solution Gskill Ripjaw V series or Trident Z series have both proven to work well for me

[Ipso]

Menthol wrote:
Is it stable at all default settings? After so many blue screens and OS errors your likely going to end up with a corrupted OS and need a fresh install

I tried to remove all usb except keyboard and mouse and also graphics, but it doesn't seem to stable at all anymore, defaults wont boot, bs in windows before or at login screen (but bios think its ok and doesn't complain), however when i lower the clock, defaults in everything except all cores at 35 multiplier, windows seems ok...

So can it still be a windows problem when it behaves like that? Feels like a hw problem...can the cpu take damage but still "work" like now, or should it be fine if I sort any other problems out? Sure the AI suite gave me a 1.45v OC but that only ran for a few hours not weeks so I'm hoping it should be ok hardware-wise...

Also, for now to eliminate the memory as a problem, I run it a t low default speeds now, 2100 (xmp is at 3ghz)

[Ipso]

Now I'm going bananas...

I couldn't get windows to boot on stock or ANYTHING above....I was about to give up and wait for new bios or something, and then I thought I'll try one last asus ai 5-way OC and see if it can fix something...as usual, the first thing it does before actually starting is settings bios to stock and boot...but no, my win won't boot on that so fail...

I set bios to defaults, but underclock to 35 multiplier which had worked for windows a few times before. Also on a whim i try to disable speedstep but didnt't think that would change anything since I've trired LCC which should prevent low voltages at low frequencies. Back in windows asus ai resumes but I stop it immediately and think what the hell, I'll try an OC in the AI GUI ....

Gradually I increased the multiplier to my old 46 OC and then decreased voltage to -0.05v to arrive at about 1.4v at 100% "light" CPU load, 1.37v at 100% hard load, and 1.35v or so at the most extreme part of p95 torture...and it worked...almost!...10 mins of torture and max temp of 75 at the hardest part (after 14 min or so it seems to get "easier"/cooler again) and then it hanged...

This from not being able to boot windows at stock...wtf? Is somehow my speedstep broken but the rest ok?...Will try some lcc now to see if that get's me past the p95 15 min mark...if so I'm happy for now...and will start plugging usb devices back in...but somehow I doubt it will last...I'll probably get a bs tomorow or in few days and have the same problems...

(also running integrated graphivcs now, during the testing and after switching a fan my gcard got the vga-led during boot...hope it was just a glitch..)

[Ipso]

Ok, hoping this is my final "report" and things will keep working as is, (but I doubt it);

Changed bios to:
Overclock tuner: Manual
BCLK: 100
Multiplier for all cores: 46 ("group/locked mode")
CPU Voltage: Offsetmode@-0.07v
LLC: Level 3
-> Speedstep disabled

Everything else (except fan profiles) at default settings

This gives me (temps are max and voltages aproximate average):
Idle: 4Ghz@1.25v@30c
100% all cores "light" (first 5min of p95 torture): 4.6ghz@1.41v@50c
100% all cores "heavy" (5-10 mins of p95 torture): 4.6ghz@1.37v@70c
100% all cores "extreme" (10-15 min of p95 torture): 4.6ghz@1.36v@77c

So if that keeps up I'm satisfied...the "heavy/extreme" parts will never show up for longer periods of time in realistic use, and if I get a bs/year due to hidden instability that's fine with me..

Also, the rig is in a desk cabinet, with a bunch of holes, filters, a case without any walls and lots of fans in both desk and case...so all in all perhaps I could shave a coupled c of the temps of taling it out but I think this is ok.

So as far as I understand, my near insanity the past days were mostly due to a broken(?) speedstep or something like that...but not quite out of the woods, memory@xmp and usb-devices may still unfix this but I'll run this for a few days first to know where my baseline is.

Thanks for all the help for now! (I'll probably be back after a bs and same problems all over tomorrow or so...)

[x-rated]

try avx2 y-cruncher too and also two handbrake encoding from h.264 to h.265 at the same time 🙂 those two programs (at least handbrake) are stressing different parts of cpu and even stable-looking settings may crash

[Ipso]

x-rated wrote:
try avx2 y-cruncher too and also two handbrake encoding from h.264 to h.265 at the same time 🙂 those two programs (at least handbrake) are stressing different parts of cpu and even stable-looking settings may crash

Thanks for the tip, but I'm rather relaxed when it comes to absolute stability...as long as there's less than a bs every couple of months during normal use I'm ok with it, so I wont try to hunt special cases, but thanks for the info, if I change my mind I try those out!

[Ipso]

So a hopefully super-final report for anyone interested in what went wrong and how it turned out in the end:

I've now plugged in the graphics card and other USB devices, as well as set the memory to almost-xmp 2933/cas15@1.36v. The system is configured as a couple of posts up when it comes to OC (the offset variant), I tried adaptive mode and manual but this was the best compromise I could reach without speedstep when it comes to idle volts vs stable turbo OC..The only difference being that the OC ram gave the CPU more work faster in p95 torture tests and bumped temps almost 10c! But thinking only the first of those "temp/work-levels", or just a second here and there of the others in normal use, would still result in ok temps for everyday use...so now I wait and see if the system stays stable for 5 years and I don't have to touch it or if I'll be back in a week...

All in all, the only (strange) conclusion i can draw about what caused the problem is that somehow after a week my speedstep "broke" somehow (or "works" if voltages are above 1.4 when idle...) so disabling it was what made my OC work again (or even defaults boot for that matter).

Thanks for all ideas and help!