OS issues during memtest

[toyo]

Hello,
Tried to run this and it really doesn't play nice with Windows 10. While it ran, it froze Windows a couple of times, Desktop Windows Manager crashed 3 times (was watching some YT), SettingSyncHost.exe crashed once too, screen would go black and freeze for a while. Two times it recovered, test continue to run with 0 errors. 3rd time DWM crashed however, I was logged off and asked for PIN. In Event Viewer, a lot of Windows services timed out, such as:
"The Windows Push Notifications User Service_57cd8 service terminated unexpectedly. It has done this 1 time(s). The following corrective action will be taken in 10000 milliseconds: Restart the service."
In my humble opinion we should have some OS stability during testing so we can see final results, let it run over the night etc.
If this is not possible and the app requires some specific OS setup to run, like no apps running during the test or user activity, which is OK, but should be clearly specified.

 

[W1zzard]

Reduce the amount of memory tested. If you choose "maximum", it will aggressively force applications into the page file, which will lead to system slowdowns and possibly some apps crashing because they're not optimized for such memory-pressure situations.

Never heard of DWM crashing though. MemTest64 is primarily written for Windows 10

 

[toyo]

Reduce the amount of memory tested. If you choose "maximum", it will aggressively force applications into the page file, which will lead to system slowdowns and possibly some apps crashing because they're not optimized for such memory-pressure situations.

Never heard of DWM crashing though. MemTest64 is primarily written for Windows 10

Yeah I figured it's because running out of RAM. Maybe DWM crashed because of me watching Youtube at the time on Vivaldi? Videos weren't really normally running either, there were interruptions.

In case it's helpful, here is what the Event Viewer is saying:

Faulting application name: dwm.exe, version: 10.0.17763.1, time stamp: 0xe52aabf3
Faulting module name: dwmcore.dll, version: 10.0.17763.1, time stamp: 0xa3929d7c
Exception code: 0xe0464645
Fault offset: 0x00000000000af069
Faulting process id: 0x454
Faulting application start time: 0x01d4835254fdf117
Faulting application path: C:\Windows\system32\dwm.exe
Faulting module path: C:\Windows\system32\dwmcore.dll
Report Id: 4ffe2da2-0b1a-4ea9-90f5-aa72cec9dfde

The Desktop Window Manager process has exited. (Process exit code: 0x000000ff, Restart count: 1, Primary display device ID: NVIDIA GeForce GTX 1070 Ti)

Did this 3 times, and then Windows logged me off. Never seen this behavior.

 

[W1zzard]

I could imagine there being a connection between DWM crashing and video playback.

That the video is getting choppy isn't surprising, because your system will move a lot of processes from RAM into pagefile (which is A LOT slower), so the system will start to stutter, which affects the running video, too.

MemTest64 locks all the memory it's testing, so effectively it's taking it away from the OS. This is equivalent to trying to run Windows 10 with 1 GB of physical memory

 

[Gorstak]

Uhm, this kinda sounds funny...like: I tried to fix the carburator with engine running sort of funny...

There is a builtin windows memtest, just click on start and type memtest...it will test it after reboot, but you should probably do it when you go to sleep as it takes a few hours to complete.

 

[toyo]

Uhm, this kinda sounds funny...like: I tried to fix the carburator with engine running sort of funny...

There is a builtin windows memtest, just click on start and type memtest...it will test it after reboot, but you should probably do it when you go to sleep as it takes a few hours to complete.

That test is completely useless for my use case, just like memtestx86. These tests are for defective RAM. I am more interested in actually testing if OC timings/voltages such as SA/IO/VDIMM are stable. memtestx86 and Windows memtest would pass even while the settings are EXTREMELY unstable, in fact, I've passed Windows memtest in cases where Windows would BSOD on any game launch.

As for the "fix the carburator while the engine is running", you should know that using the PC while running prime95 with AVX is entirely possible. It's how i do my 24-48 hrs of testing, it's not ideal, there are slowdowns, but for example work in Indesign and Photoshop is OK, and browsing the web/watching videos is also almost normal. Same can be said for Intel Burn test and LinX, or for running GSAT inside Windows 10/Ubuntu. So yeah, in some instances it's entirely possible to keep using the PC while stress testing.

 

[Gorstak]

you sound tired

 

[toyo]

you sound tired

It can get tiring, yeah, but it's all good now. It's not that I wanted to, but I had to do this much testing, or else my PC would be unstable.
It's all because of me buying RAM that was not on the motherboard QVL, which for me it meant that the motherboard would implement XMP wrongly. My RAM is also very weird. For example I cannot run it with the 1.35V listed in XMP. I run 3000MHz with 1.2V VDIMM, the same as in the JEDEC profile. I cannot use any voltage above 1.35V, which is really peculiar, as most RAM on Z370 would overclock with 1.4V easily. So for example, even for 3400MHz stable, I'd have to run 1.3V.

Lesson learned here, don't cheap out on RAM, especially don't buy RAM not on QVL. Next time I'll just buy some GSKILL stuff that's universally acclaimed and be done with it, but that will probably be when DDR5+new platforms will be here.

 

[Gorstak]

I have noticed a lot of brand name machines use double slotted ram chips, not sure why. I don't mean chips on both sides of the ram stick, I mean two lines of chips on both sides. I never saw any ryzen owner complain on incompatibility with those chips.

 

[John Naylor]

Why is memtest86+ worthess ?

http://www.memtest.org/

 

[toyo]

Why is memtest86+ worthess ?

http://www.memtest.org/

It's not, if you have potentially defective RAM. But it's worthless for unstable RAM because of voltages such as VDIMM, VCCIO, VCCSA, and badly implemented XMPs.

 

[Gorstak]

If you're overclocking, a general rule is to loosen all timings to max, then oc as much as you can. Once you're done with stability testing for that part, you start tightening RAM timings...basically it's a very long and tiring process without any real effect in every day work.