Question PC bluescreens and sometimes crashes without BSOD

[VitusB]

Hello.

For the past few months i've experienced a number of crashes and BSOD happening daily.
Most of the BSOD have the driver ntoskrnl.exe as the cause, which i've heard could be because of a hardware problem.
I've tried almost everything people have suggested, but nothing has helped.
Some things I've tried:

• Clean install
• MemTest86 (and tried another set of RAM)
• Installing another graphics card
• Stress tested CPU
• sfc /scannow
• Reinstalled graphics card drivers

If anyone have any ideas of how to solve it please let me know.

Some minidumps (I've attached a minidump from a driver verifier I just ran): https://drive.google.com/drive/folders/1gyb6GCE4dMGL_-ytXLeVucALeFgPLO5d?usp=sharing
Speccy: http://speccy.piriform.com/results/zlRGXMB1KWFtidMWtp1dO4Z

 

[VitusB]

When I went to change the DRAM voltage I saw that it was set to Auto, and that the current voltage was 1.216v where as the default would be 1.200v, should I still raise it by 0.010v - 0.020v, so that it would be 1.226v - 1.236v?

 

[Darkbreeze]

If your voltage is set to 1.216v then XMP is not enabled. The XMP profile voltage value for that memory kit should be 1.35v (Sometimes shows as 1.36v depending on the motherboard)

I would recommend that you ENABLE XMP, then save settings and exit the BIOS. Then, go back into the BIOS and make sure that it is actually still set to the XMP profile values. Then exit the BIOS and boot into Windows. Then download CPU-Z and look on the memory and SPD tabs to verify that it IS running at the XMP profile speed (Will look like half usually, so, 1600 for 3200mhz, 1500 for 3000mhz, since DDR4 is "double data rate" type memory), voltage and timings.

If it is, then I would suggest that it would be a good idea to run the full four passes on the memory at the XMP configuration settings.

Running it at the stock default configuration is only intended to tell you if there are PHYSICAL problems with the memory, ie, it's bad. Running it at the XMP configuration should tell you if there are problems with the compatibility of the memory (As in, the motherboard doesn't "like" that specific memory kit) or with the timings from the profile as well as any secondary or tertiary timings the motherboard has configured to run the kit at.

 

[VitusB]

I enabled XMP again, and the voltages then said 1.356v - the same as you said.

Then I installed CPU-Z and everything looked normal I think, you can check for yourself here:
https://drive.google.com/file/d/17D-du8A77aVDlCEaTpjZu2XbzQh2gTyP/view?usp=sharing

SPD.PNG

drive.google.com

https://drive.google.com/file/d/1RCSKh97OwPoBU1W4j9VQL-UiP02h_EiE/view?usp=sharing

Before I head to bed, I will run memtest again, but this time with XMP enabled.

 

[Darkbreeze]

Your primary CS latency timing (CL) is not right. It is showing as CL16 on your memory tab screenshot but the PROFILE CL latency for that kit with XMP enabled SHOULD be CL15. That is also what it shows in your other screenshot below the "XMP 2998" profile. This is probably because it is a Ryzen platform and Ryzen does not like to run memory kits with odd numbered latency timings. Personally, I believe, even if it passes Memtest86 (Which is NOT a guaranteed 100% validation that there are no memory problems, more like a "90%" validation) that your problems likely stem from this kit not having evenly numbered timings that are favorable to Ryzen. Combining the fact that the memory is a frequency that Ryzen doesn't like (3000mhz) and has odd (Intel) timings, I think it's very possible that this is part of the problem HOWEVER it is seemingly a shot down theory considering that your other 2400mhz kit had the same problems.

Before you do anything else, and after you either complete or fail the Memtest run, find the DRAM voltage setting (Memory voltage) in your BIOS and try raising the voltage to 1.36v. Save the settings and then exit the BIOS. Then try it all again to see if it has improved at all. If not, try raising it in .005v increments for no more than three raises total to see if there is any improvement. If not, then return it to 1.35v and we'll look at other possibilities.

 

[VitusB]

Your primary CS latency timing (CL) is not right. It is showing as CL16 on your memory tab screenshot but the PROFILE CL latency for that kit with XMP enabled SHOULD be CL15. That is also what it shows in your other screenshot below the "XMP 2998" profile. This is probably because it is a Ryzen platform and Ryzen does not like to run memory kits with odd numbered latency timings. Personally, I believe, even if it passes Memtest86 (Which is NOT a guaranteed 100% validation that there are no memory problems, more like a "90%" validation) that your problems likely stem from this kit not having evenly numbered timings that are favorable to Ryzen. Combining the fact that the memory is a frequency that Ryzen doesn't like (3000mhz) and has odd (Intel) timings, I think it's very possible that this is part of the problem HOWEVER it is seemingly a shot down theory considering that your other 2400mhz kit had the same problems.

Before you do anything else, and after you either complete or fail the Memtest run, find the DRAM voltage setting (Memory voltage) in your BIOS and try raising the voltage to 1.36v. Save the settings and then exit the BIOS. Then try it all again to see if it has improved at all. If not, try raising it in .005v increments for no more than three raises total to see if there is any improvement. If not, then return it to 1.35v and we'll look at other possibilities.

Ok, I passed the memtest, but I will try to increase the voltages as you said

 

[VitusB]

Btw, what do you mean by this?

Then try it all again to see if it has improved at all

I changed my DRAM voltage to 1.360v and went back into BIOS to double check, and saw that it was at 1.368v.
Then I opened CPU-Z and noticed that it still said 1.350v under the XMP-2998 tab under the SPD tab.

Edit: my pc crashed some time after setting the voltage to 1.360v, so I tried to raise voltage again but realized I couldn't set it to 1.365 but only raise it in increments of 0.01

 

[Darkbreeze]

The value under the XMP-2998 listing won't ever change. That is the hard coded XMP value. It is only reporting what is hard coded there. CPU-Z was really only for verifying speed. Better would be to use HWinfo, choose the sensors only option when you open it.

If it still crashed at 1.36v, then I'd suspect that it may not be the memory voltage, but you can certainly try bumping it up another .01v and see if that helps. If not, then I'd just reset the DRAM voltage back to 1.35v because the problem then is likely elsewhere.

 

[VitusB]

My pc crashed at 1.370v, so i've reset it to 1.350v.
Here's some dump files since some time has passed since I last sent them: https://drive.google.com/file/d/1GGx3NTDe3YF5SQWw7Yaqaoh1MI_Oj7Hb/view?usp=sharing

 

[Darkbreeze]

Ok, then I'd suggest you contact AMD or your motherboard manufacturer. I don't suggest setting a manual voltage on this, since you don't have a manual OC, and if you are getting Prime errors with the stock configuration it would seem as though either something is wrong with the CPU or the motherboard. Without getting my own hands on it, that would be my early guess.

 

[gardenman]

I ran the dump files through the debugger and got the following information: https://jsfiddle.net/n7fxq8u6/show

Spoiler

File information:	111820-48125-01.dmp (Nov 18 2020 - 08:37:48)
Bugcheck:	ATTEMPTED_EXECUTE_OF_NOEXECUTE_MEMORY (FC)
Driver warnings:	*** WARNING: Unable to verify timestamp for amdkmdag.sys
Probably caused by:	memory_corruption (Process: System)
Uptime:	0 Day(s), 23 Hour(s), 38 Min(s), and 55 Sec(s)

File information:	111720-43968-01.dmp (Nov 17 2020 - 08:58:15)
Bugcheck:	UNEXPECTED_KERNEL_MODE_TRAP (7F)
Driver warnings:	*** WARNING: Unable to verify timestamp for vgk.sys
Probably caused by:	memory_corruption (Process: System)
Uptime:	0 Day(s), 0 Hour(s), 10 Min(s), and 32 Sec(s)

File information:	111620-44187-01.dmp (Nov 16 2020 - 15:36:23)
Bugcheck:	DRIVER_IRQL_NOT_LESS_OR_EQUAL (D1)
Driver warnings:	*** WARNING: Unable to verify timestamp for amdkmdag.sys
Probably caused by:	memory_corruption (Process: System)
Uptime:	0 Day(s), 0 Hour(s), 42 Min(s), and 18 Sec(s)

File information:	111620-42609-01.dmp (Nov 16 2020 - 14:53:27)
Bugcheck:	IRQL_NOT_LESS_OR_EQUAL (A)
Probably caused by:	memory_corruption (Process: Discord.exe)
Uptime:	0 Day(s), 3 Hour(s), 50 Min(s), and 09 Sec(s)

File information:	111520-48828-01.dmp (Nov 15 2020 - 06:46:01)
Bugcheck:	DRIVER_IRQL_NOT_LESS_OR_EQUAL (D1)
Driver warnings:	*** WARNING: Unable to verify timestamp for amdkmdag.sys
Probably caused by:	memory_corruption (Process: System)
Uptime:	0 Day(s), 20 Hour(s), 05 Min(s), and 33 Sec(s)

According to the dumps, it looks like 3 of them were caused by your GPU. 1 of them was caused by Vanguard.

 

[Darkbreeze]

Have you done a clean install of the GX drivers using the DDU? If this is GX related then it's a whole different ballgame. I thought we were talking about rounding errors in Prime95, which would have zero to do with the graphics card.

 

[VitusB]

and if you are getting Prime errors with the stock configuration

I didn't get Prime errors with stock configuration, the Prime errors were because I was using another set of RAM that didn't fit my motherboard correctly.
Since I reinstalled my original RAM again I've been able to pass both Prime95 and memtest

Have you done a clean install of the GX drivers using the DDU?

If GX drivers is my GPU drivers the yeah, this was one of the first things I tried.

I'll try to contact AMD and MSI to see if they can help.

 

[Darkbreeze]

Well, if you have three crash dumps all related to the graphics card, then I'm thinking you kind of want to look that direction first maybe. Gardenman and Colif are far more experienced with sorting out the crash dumps than I am, so pay attention to what they say in that context. Any chance you have another graphics card you can plug in place of the current one, or borrow one, to test the theory?

 

[VitusB]

Well, if you have three crash dumps all related to the graphics card, then I'm thinking you kind of want to look that direction first maybe. Gardenman and Colif are far more experienced with sorting out the crash dumps than I am, so pay attention to what they say in that context. Any chance you have another graphics card you can plug in place of the current one, or borrow one, to test the theory?

I've already done that.

I think now I can safely say that it isn't my: RAM, GPU, PSU or storage.
However, it could still be either my Motherboard and maybe my CPU.
But then again, wouldn't my pc crash when running Prime95 if the CPU were faulty, since it's stressing it?

 

[Darkbreeze]

Prime95 only tests SPECIFIC instructions using a fraction of the CPUs capabilities. The CPU is responsible for processes related to pretty much everything, SATA, PCI, USB, everything, well, most everything. If there were a bent pin, it could ONLY have an effect in a specific set of circumstances, unrelated to the type of processes involved when running Prime. Prime is not really all that great of a stability test. It is, but it's not.

Try running Realbench. Choose the stress test option. Select HALF of the amount of RAM you have installed, so if you have 16GB installed, put 8GB in the option field. See how long it will run that. If you can make it to anywhere between 6 and 8 hours, it's probably not the CPU, but then again, a faulty motherboard could cause a failure too. The problem is, with computer hardware, that MOST of it is impossible to diagnose as faulty without replacing it unless you are a manufacturer or high end electronics repair facility with high end specialty test equipment and a lot of experience with electronics. And the schematics.

 

[Colif]

errors 1, 3 & 5 blame amd gpu drivers, though they are different to normal gpu errors, no mention of directx and they all happen after a page fault. Page faults are normal operations of any os with virtual memory. it essentially means the CPU couldn't find what it was looking for in ram and had to access the boot drive
error 2 blames Valorant

Can you remove valorant until we fix problem as If it wasn't a game I would classify it as malware

 

[VitusB]

Can you remove valorant until we fix problem as If it wasn't a game I would classify it as malware

Try running Realbench. Choose the stress test option. Select HALF of the amount of RAM you have installed

Ok, I'll do it when I get home.
Is it the Asus RealBench program?

 

[Darkbreeze]

ASUS Realbench

https://www.techspot.com/downloads/downloadnow/7050/?evp=5f8bf568c9ed15babb267e7895014b0a&file=1

 

[VitusB]

I let RealBench run for 8 hours with 8 gb RAM, on the stress test option, and my pc didn't crash once.
Can we now confirm that it isn't my CPU or?...

 

[Darkbreeze]

I'd say it's not the CPU, but I've been wrong before.

Try Furmark on the GPU stress test. And Unigine.

 

[VitusB]

I'll run furmark now, can you send a link for unigine?

 

[VitusB]

Ok so I let Furmark run for like 4 hours and it didn't crash once.
Then I had to leave my house because I had to do something, but I just let my pc stay on, and whilst I was gone it didn't crash either.
Then when I got home I opened Minecraft and after about 5 minutes it crashed without BSOD.

 

[Darkbreeze]

Well, I'm at a loss really.

If it has the latest BIOS, Windows is fully up to date as far as Windows updates INCLUDING the "optional updates", it passes 4 passes of Memtest86, it passes 8 hours of Realbench and it passes 4 hours of Furmark, there shouldn't be anything that makes it crash unless it's the ACTUAL game or application itself. Since you've somewhat recently done a clean install of Windows (I believe you said you did) then we can probably rule out OS corruption as well. No damn idea then.

 

[Colif]

Since 3 blamed the gpu driver, i would at least run ddu and instead of grabbing drivers from AMD, run windows update after the fact and let windows find drivers for the card.

did you ever run driver verifer? you probably have 5 pages in but just thought I ask. I generally don't suggest it as it can break windows... Microsoft should be trusted to not break their own operating systems really.

 

[VitusB]

Since 3 blamed the gpu driver, i would at least run ddu and instead of grabbing drivers from AMD, run windows update after the fact and let windows find drivers for the card.

I can try it. Currently, my last 3 crashes since I uninstalled Valorant, has been without BSOD.

did you ever run driver verifer? you probably have 5 pages in but just thought I ask. I generally don't suggest it as it can break windows... Microsoft should be trusted to not break their own operating systems really.

I did run it some time ago, but I can do it again. Do you know what settings I should enable?