Memory related BSODs (8700k, Maximus X Hero 1602)

[seb0r36]

I got an 8700k a couple of weeks back. I ran it sort-of stock for the first few days (MCE disabled, XMP enabled). It worked fine - mind you, I did not test it extensively, becaused I planned on delidding anyway.

Then, I got it delidded and started overclocking, following Der8auers OC guide for this exact combo of cpu and mb.
It first failed while trying 5.1 @ 1.312 V (AVX -3) in 1344k (non-AVX) with MEMORY_MANAGEMENT, so I tried 5.0 @ 1.28 V, trying to bring volts down rather than clock up.

That failed as well, witht the same BSOD, only later. So I went back to 5.0 @ 1.296 V and called it a day.

But then the problems really started: It would *BSOD rarely, but consistently with that setting, but never under load, always in idle or web browsing. That BSOD was always PFN_LIST_CORRUPT.

By then I figured RAM and/or IMC could be the problem, so I ran HCI Memtest 1000% and memtest86 without a single error.
But it kept crashing on me, so I went back to stock (except XMP). The BSOD persisted anyway.

Then I did a fresh windows installation, wanting to rule out driver corruption. Guess what, the BSODs kept on coming.

That was when I realized that in an asrock forum, some people were reporting the same problem — memory related BSODs in near-idle situations. In their case VCCIO was way to high and bringing it down to 1.10 V did the trick. I look in the BIOS and was shocked to find that VCCIO in Auto was at 1.33V (wayyy to high for a 3200 cl16 kit). So I brought it down to 1.10 V (and VCCSA as well from 1.25 to 1.15).

That worked for the last 2 weeks (with 4.8 @ 1.232 -0 AVX). *I dared to push higher yesterday, passing 2 hours of custom prime95 with 5 @ 1.28 V. Unfortunately, 5 @ 1.264 V failed with SYSTEM_SERVICE_EXEPTION, another BSOD not really related to insufficient vcore.

So now, for the grand finale: *I did a fresh Win 10 Install today on a new 970 Evo, having everything at default, even the ram. After installing all updates (but no third party applications) it crashed on me again, in idle (PFN_LIST_CORRUPT).

Strangely, now even Memtest finds errors (it found the exact same error in 2 out of 4 runs [test 7, same addres, same expected and actual values]).*

I‘m really lost now, some leads point to IMC and some to RAM, but it could also be something else entirely, e.g. my CPU cooler (see rig), which has a very high mounting pressure, probably interfeering as well.

Someone got any ideas or similar experiences which he/she likes to share?

Thanks for the feedback and sorry for the long post.*

Sebastian*

[Phillyflyer]

Please post what memory you are using...Brand...model and # of modules.

[seb0r36]

http://gskill.com/en/product/f4-3200c16d-16gvkb*


It’s 2 x 8 GB @ 3200 16 18 18 38. *It‘s also in the qvl for that mb

[chevell65]

That link didn't work for some reason. This one works. https://www.gskill.com/en/product/f4-3200c16d-16gvkb

The PFN is used by your hard drive to determine the location of each one of your files on the physical disk.

Are you installing Windows 10 version 1803?

[Phillyflyer]

First things first....Enable the XMP....set VCCIO @ 1.15v..set VCCSA @ 1.20v....for 5.0ghz..set Vcore @ 1.285v....for 5.1ghz..set Vcore @ 1.36v or 1.365v... set LLC @ 6....you can set the AVX to 0 if you wish...run Prime 95 version 26.6 as it has no AVX...and...HCI Memtest...try that..lemme know the outcome

[seb0r36]

@chevell65: Sorry for the link not working, I copied it in on my ipad, maybe that‘s why. And yes, it‘s 1803.

@Phillyflyer: I will test that shortly after I finished running my ram sticks separately. Since errors only popped up in test 7 (I tested both in test 7 already, exact same error as before). My first stick already came back clean (8 runs of test 7). Depending on the outcomes I will make the following assumptions:

- Other Ram stick passes as well in the same slot as stick1 did —> Ram is most likely ok, problem lies with either the mb (ram slot) or the imc

- Other ram stick does not pass in the same slot as stick1 did —> That ram stick is most likely faulty (if only even a tiny bit)

Mind you, these memtest86 runs are all at stock, non-xmp.

[mrvortep]

Phillyflyer wrote:
First things first....Enable the XMP....set VCCIO @ 1.15v..set VCCSA @ 1.20v....for 5.0ghz..set Vcore @ 1.285v....for 5.1ghz..set Vcore @ 1.36v or 1.365v... set LLC @ 6....you can set the AVX to 0 if you wish...run Prime 95 version 26.6 as it has no AVX...and...HCI Memtest...try that..lemme know the outcome

Hey Philly, sorry to jump in this thread - but on reading this thread I noticed that my Max Hero X has set my VCCIO to around 1.35V and VCCSA to 1.28V. This is using Corsair Vengeance Pro RGB 3600 (18-19-19-39). The DIMMS get quite warm, into the low 50s when gaming. Do you think it's worth manually adjusting their voltages down to VCCIO: 1.10V and VCCSA: 1.20V as suggested above? Thanks as always!

[Phillyflyer]

mrvortep wrote:
Hey Philly, sorry to jump in this thread - but on reading this thread I noticed that my Max Hero X has set my VCCIO to around 1.35V and VCCSA to 1.28V. This is using Corsair Vengeance Pro RGB 3600 (18-19-19-39). The DIMMS get quite warm, into the low 50s when gaming. Do you think it's worth manually adjusting their voltages down to VCCIO: 1.10V and VCCSA: 1.20V as suggested above? Thanks as always!

XMP enabled will set SA /IO voltages in the 1.30-1.35v range by default....no need for it to be that high....that being said...all depends on how good the chips memory controller is as well... you can start with IO @ 1.10v and SA @ 1.15v...see if that works...manually adjusting these voltages will take a little time...patience and effort to get the right configuration for stability....too much or too little will create instability....keep within .05v of each.

[mrvortep]

Phillyflyer wrote:
XMP enabled will set SA /IO voltages in the 1.30-1.35v range by default....no need for it to be that high....that being said...all depends on how good the chips memory controller is as well... you can start with IO @ 1.10v and SA @ 1.15v...see if that works...manually adjusting these voltages will take a little time...patience and effort to get the right configuration for stability....too much or too little will create instability....keep within .05v of each.

Cheers Philly. I've set the memory to IO@1.15v and SA@1.20v and just left it there for now. Seems stable running P95 and a little cooler on the DIMM (I didn't try any lower). I'm just running at 5ghz so not pushing the system really, CPU never hits 70C - I'm happy with the combination of speed/sound as a result. Thanks again!

[chevell65]

Sounds like a bad memory stick from the testing that you did. This explains the BSOD's.