Random BSOD ntoskrnl.exe

drapos · 06 Dec 2015

Hi.

I revamped an old X58 build some months ago. It was an inherited computer with several problems. I have been changing most of it (CPU, MOBO, RAM, HDD) but random BSODs keep happening.

I'm running a fresh install of Windows 10 Home x64, installed on a new SSD. This is the 3rd time I have to restore Windows this week cause it keeps corrupting itself given a time. I tried default drivers, old drivers, new drivers, but nothing works, and eventually I get BSODs related to IRQ/Memory/DRIVER_IRQL_NOT_LESS_OR_EQUAL/DPC_WATCHDOG_VIOLATION/Threads bad handled, etc. I have it slighty oced, and heavily tested with Linx/Prime/Memtest/Cinebench/etc for stability.

BSViewer mostly blames ntoskrnl.exe, but I can't dive into the minidump. I will attach last one.

I'm really desperate. I have an even older setup with mostly same onboard devices on a X48 platform running flawless.

Any help would be very much appreciated.

axe0 · 09 Dec 2015

Hi drapos,

What tests exactly have you ran?
Could you give details about the results of the tests?

drapos · 09 Dec 2015

Hi, thanks for attending my problem.

Ram profile is 3x4 2133 CL 10-11-10-28 1.5V, but I'm using it at 1600 CL 8 1,5V because I would need to push CPU OC harder to get memories there. Raising memory multiplier too high on X58 would result in no boot. My OC is just going from 133 bus speed to 160, but I got BSOD at 133 too, so I left 160 as default.

I have left memtest86+ several hours, 10 passes of Linx using all available memory and a couple of hours of Prime just to catch what Linx could miss. BSOD never comes when stressing the computer, but when casual surfing the web instead. Most of the crashes come right after booting the computer or when it is idling. Every restart is a rolldice.

Temps are pretty low on the CPU and DDR3 side, and OKish on the GPU.

Many thanks for your time.

axe0 · 09 Dec 2015

First thing I would recommend is to revert any overclock back to stock on your system, on an unstable system overclocking could make it worse and difficult to troubleshoot.

drapos · 09 Dec 2015

Sure, I never worked on proper voltages for default clocks -Mobo default voltages are just too high surpassing 1,200VCore for a 32nm CPU- but I will do something quick. Will try current voltages with default clocks and torture it for some hours.

Is there any clue at dump? I have some more since first post if needed.

Thank you.

axe0 · 09 Dec 2015

It is a 0x1A(MEMORY_MANAGEMENT) which is caused by the memory, memory frequency, memory voltages, bad drivers, slot of the motherboard or power issue.
As usual, it doesn't flag any driver.

drapos · 09 Dec 2015

I will attach all the 4 dumps I have from the last days since my last fresh install.

I only have stability issues with Windows, never had any problem with DOS stuff such as Hirens utilities even when Windows was looping into BSODs, that's why I believe RAM shouldn't be a problem. Even when obviously I use those way less, it's also true I use those utilities to stress and torture.

I really appreciate the time you are using to help me. I'm running out of ideas whith this build.

drapos · 10 Dec 2015

This is from tonight.

axe0 · 10 Dec 2015

Diagnostic Test

RAM TEST

Run MemTest86+ to analyse your RAM. MemTest86+ - Test RAM - Windows 10 Forums

Note

MemTest86+ needs to be run for at least 8 passes for conclusive results. Set it running before you go to bed and leave it overnight. We're looking for zero errors here. Even a single error will indicate RAM failure.

drapos · 10 Dec 2015

Is the number of passes as important as the number of hours? Memtest have sang me faulty modules/mobo pins in other setups in a couple of hours, while I have tested this build as long as 11h with no errors.

12GB on a trichannel system will take 19h at least. But then, testing modules one by one could miss IMC unstabilities because of the abnormal lack of stress going from tri channel to single channel.