Intermittent issues, possible memory failure?

[mrluckypants96]

Hello Everyone! I've been trying to track down intermittent issues on my machine since a major rebuild and wanted to get other's opinions on the matter. Errors have ranged from a strange soft-lock where Windows mostly stops working over the course of a few minutes, to random reboots without any perceivable cause, to various bluesceen codes. The issues have been going on for the past couple months and I've been doing my best to keep notes on them as I go. I believe, at this point, that it's either the memory or the CPU's memory controller that are giving me the issue... or the motherboard is messing with the memory and breaking things, I'm not sure, but that's why I'm here. Interestingly, the original problem I was facing, the soft freezes, was resolved but now there are other issues.

System Specs. Basically everything is new except the GPU and Motherboard that were brought over from the old build:
CPU: Ryzen 7 5800X3D
Memory: 64GB (4x16GB) Corsair Vengeance RT 3600MHz DDR4 (or 16GB (2x8GB) Corsair Vengeance RGB when testing)
Cooling: Be Quiet Dark Rock Slim, upgraded to Be Quiet Pure Loop 280 when I realized that was nowhere near enough cooling.
Motherboard: MSI B450M Mortar (non-Max) on latest BIOS for 5800X3D support.
GPU: XFX 5700XT Thicc III (Also run with R7 240 and GT710 when testing things)
PSU: Originally a 1000W Seasonic Titanium Prime, but the fan was loud and annoyed me so I replaced it with a 700W Fanless version.
Storage: 2x4TB XPG Spectrix S40G SSD's
OS: Windows 11, reinstalled several times over the course of testing.



Here's the copy of my notes I've been taking, in mostly chronological order, hopefully they're readable enough for y'all to get the gist of what I've been trying:

Three times so far, a few days apart, I’ve noticed my computer soft-locking. The first sign of failure is YouTube video playback pausing and failing to restart. Discord will freeze, the clock and taskbar freeze, control alt delete does nothing, nor does control shift escape. Some apps can still be clicked on, some can be scrolled through and the windows moved around, but once clicked off of will be locked and unusable again. The start menu can be brought up but the search bar will not function and the menu will go unresponsive when clicked off of. A hard reset is the only solution. Update: Probably up to a dozen so far. Can be a few days or less than an hour between freezes. Also can happen when playing games, the game will freeze at the same time as video playback.

No errors appear on the Event Viewer until after the reboot, and then it's the usual "Windows unexpectedly rebooted" errors.

SFC /scannow found no corrupted files.

Turning on driver verifier makes the system run very slowly and use lots of processor time (expected). Attempting to run a game causes the system to hard lock and reboot. There is no BSOD or error in the event viewer. Disabled driver verifier.

Task Manager shows no abnormal resource usage when the freeze occurs. MSI Afterburner stops monitoring when the freeze happens, but the graphs can still be scrolled through and show nothing strange apart from slightly lower GPU usage for a the few seconds between YouTube playback freezing and Afterburner monitoring freezing (expected).

Memory tested, 5 advanced passes of windows memory test. No errors found. Froze again the next day.

Motherboard on latest BIOS, only BIOS available that supports the new processor.

Updated Graphics drivers. Frozen again less than an hour later.

Reinstalled Windows from scratch. A few days later, it’s frozen again.

Noticed before the reinstall but sometimes when leaving the machine on with the monitors off and returning later (eg. sleeping and coming back in the morning), turning on the monitors will lead to no GPU output and requires a restart. Possibly the same freezing as before but with the GPU not outputting? Noticed it happening after swapping from VBIOS 1 to 2, and still happens after changing from 2 to 1. Update: Hasn't happened since Windows reinstall.

Keyboard sometimes fails to connect. On one occasion I played a game for several minutes without noticing the keyboard wasn’t working. Opening iCUE reconnected it. Upon a reboot (after freeze) it once again wasn’t working for several seconds, followed by windows installing a device labeled “PCI Bus”. Update: Seemingly fixed after reinstalling iCUE

Replaced power supply, no change in behavior. Not really a troubleshooting step but it was loud and annoying me, so I got a new one.

Checked memory speed. A-XMP was never enabled since rebuild, RAM speed was limited to 2400, won’t POST at A-XMP 3600 and resets to 2400.

Tested with known good memory. Froze again. Put original memory back in.

Testing with another GPU (R7 240). Multiple bluescreens. GPU driver errors?
- netio.sys (missed stop code)
- kernel security check failure
- tcpip.sys system service exception (in safe mode w/networking)
- Pool corruption in file area
- kernel security check failure (in safe mode)
- classpnp.sys kmode exception not handled
Eventually got it to download and run DDU, removing GPU drivers. Still crashing within 15 seconds of windows boot with similar errors. Reset Windows from within the recovery menu.
- IRQL not less than or equal, black screen of death mid-reset, no file listed as the cause.
Tested *another* GPU, get a few more immediate bluescreens related to system files. From this point, the machine is running a GT710.

I figure windows is just borked at this point, reinstall windows.
- Cache Manager, crash mid install
Try again
- kernel security check failure
Try again, unplugged all other drives except the USB key with the iso on it and the drive I’m installing it to.
- “Windows cannot install required files. Make sure all files required for installation are available, and restart the installation. Error code: 0x8007025D
Try again
- Cache Manager crash again
Try again, swapped to a known good SSD
- “Windows installation encountered an unexpected error, verify the installation sources are accessible, and restart the installation. Error code: 0xC0000005”
Try again
- Cache Manager crash again
Bad .iso? Redownload .iso and retry.
- IRQL not less than or equal
Try again
- fvevol.sys, kmode exception not handled
Swap back to known good memory again. Made it through the install.

Now running with old memory, GT 710 GPU, old SSD. Ran for a day without issues. Started slowly putting new components back in.

Reinstalled new SSD and reinstalled Windows. Ran for a weekend without issues.

Reinstalled new memory. Won’t POST if more than one stick is installed at rated speed.
Testing revealed that at any preset over 3000MHz memory speed, slot 3 won’t function with the new memory kit.
Strangely worked fine at 3200 with the old memory kit in slot 1 and 3.
Running it now at 2933MHz preset in motherboard BIOS because it won't post at A-XMP speeds.
Crashed after about a day: PFN List Corrupt

No crashes for a few more days, reinstalled new GPU.
Couple of days later I woke up to a driver timeout error crash report from the GPU software and later in the day the computer restarted for no notable reason.
Memory Management error maybe 20 mins later, then again 2 minutes after that.

 

[Deleted member 185158]

Checked memory speed. A-XMP was never enabled since rebuild, RAM speed was limited to 2400, won’t POST at A-XMP 3600 and resets to 2400

This stands out the most for me.

Good notes by the way, wish everyone did that before posting!

This should be defaulting at 2133mhz once you've cleared the cmos.

I'm thinking an issue with bios.

Updating or recovering the bios with 1 stick of known good working memory in the proper slot is advised at this time.

However, I would like to note, that I read all of that and no mention of clearing the cmos. You should do this every time you change memory configurations.

 

[mrluckypants96]

This stands out the most for me.

Good notes by the way, wish everyone did that before posting!

This should be defaulting at 2133mhz once you've cleared the cmos.

I'm thinking an issue with bios.

Updating or recovering the bios with 1 stick of known good working memory in the proper slot is advised at this time.

However, I would like to note, that I read all of that and no mention of clearing the cmos. You should do this every time you change memory configurations.

I did find that a little weird, but assumed it was a part of the new memory. My old memory kit defaults to 2133 when installed, but the new one defaults to 2400.

I reflashed the BIOS with the latest version from MSI's website, cleared the CMOS, and reinstalled the new memory kit. Still defaults to 2400 and still won't post at 3600. I also had a number of random application crashes yesterday with the new kit before replacing it with the old kit (which runs fine at full XMP speed), after which it ran fine. I've got the new kit back in for now.

 

[ir_cow]

First thing that looks like it could be troublesome is 4x16GB. Dual Rank is already a bit hard on the IMC for Ryzen CPUs. Double it to 4 DIMMs and it could be unstable or just not boot. Try to lower it to 3200 and see if that fixes it.


Other thing you could try is run a memory test, if it fails then at least you know its probably memory related. That way you aren't wasting time chasing ghosts.

Edit: I see you say 3600 doesnt work. I that case I would go back basic trouble shooting.

 

[Chomiq]

Input primary timings only, leave everything on auto. Don't use A-XMP.

DRAM Voltage to 1.35V
Vcore SOC to 1.1 V
VDDG's to 1.05 V
VDDP to 0.95 V

Try 3600 first, if that won't work try 3200.

Also:

 

[Shou Miko]

Input primary timings only, leave everything on auto. Don't use A-XMP.

DRAM Voltage to 1.35V
Vcore SOC to 1.1 V
VDDG's to 1.05 V
VDDP to 0.95 V

Try 3600 first, if that won't work try 3200.

Also:
View attachment 257324

One i know not 3 weeks from now purchased a AMD Ryzen 9 5950X system with a set XMP of 3600MHz it was validated from the builder but not 2 weeks after he started to experiencing crashes and so on ram could only run at 2933MHz.

He shipped the gamer computer back the company tested everything and found out the memcontroller was bad could do run higher than 2933MHz sadly.

I told him it was either the memcontroller in the cpu or the ram.

 

[AlwaysHope]

OP, the ram or the board itself is suspect, run MT86 at least 2 whole runs bare minimum, if all well - them move on to Microsoft memory diagnostic tool. Passes that, then try another 3rd party tool of your choice for triple checking memory issues.
I've had issues with OC ram on MSI B450 Gaming Pro Carbon AC before & that high end B450 board only supported up to 3466MHz ram with OC. So forget about 3600MHz speed, its not going to happen on that platform especially with the capacity of ram you want to run.
Reduce your expectations unless you want to upgrade the board cause' with a Zen 3 chip, that IMC will be more robust than when I did it on AM4 with a Zen+ chip so 3466MHz should be your max with bios tweaking for stability.
A 5800X3D really deserves an X570 board imo.

 

[mrluckypants96]

First thing that looks like it could be troublesome is 4x16GB. Dual Rank is already a bit hard on the IMC for Ryzen CPUs. Double it to 4 DIMMs and it could be unstable or just not boot. Try to lower it to 3200 and see if that fixes it.


Other thing you could try is run a memory test, if it fails then at least you know its probably memory related. That way you aren't wasting time chasing ghosts.

Edit: I see you say 3600 doesnt work. I that case I would go back basic trouble shooting.

In the notes I said I ran a Windows memory test for five passes on the "advanced" setting and it gave no errors. Can't really blame you for missing it, there's a lot there. Anyways, I set the system at A-XMP timings but with the frequency turned down to 3200 and ran a Prime95 Large FFTs test to stress test the RAM. It was stable after half an hour, which I know isn't really enough to fully test stability, but it's a start. I can understand that going from two small to four larger DIMMs and trying to crank up the speed could overstress the memory controller, it's just disappointing.

Input primary timings only, leave everything on auto. Don't use A-XMP.

DRAM Voltage to 1.35V
Vcore SOC to 1.1 V
VDDG's to 1.05 V
VDDP to 0.95 V

Try 3600 first, if that won't work try 3200.

After running ir_cow's test, I then went back and disabled A-XMP and set it to 3200 with slightly tighter timings (16-20-20-40 instead of 18-22-22-42) and everything else set to auto. My BIOS is locked down and won't let me change any of the voltages besides the SOC Vcore. I really wish I could mess with voltages because man this processor is spicy. Any sort of stress test and it just pings straight to 90C. Running a stress test for the last 30min or so, it hasn't bluescreened yet.

One i know not 3 weeks from now purchased a AMD Ryzen 9 5950X system with a set XMP of 3600MHz it was validated from the builder but not 2 weeks after he started to experiencing crashes and so on ram could only run at 2933MHz.

He shipped the gamer computer back the company tested everything and found out the memcontroller was bad could do run higher than 2933MHz sadly.

I told him it was either the memcontroller in the cpu or the ram.

I don't really have anything to add to this, I just like the profile picture. :3

OP, the ram or the board itself is suspect, run MT86 at least 2 whole runs bare minimum, if all well - them move on to Microsoft memory diagnostic tool. Passes that, then try another 3rd party tool of your choice for triple checking memory issues.
I've had issues with OC ram on MSI B450 Gaming Pro Carbon AC before & that high end B450 board only supported up to 3466MHz ram with OC. So forget about 3600MHz speed, its not going to happen on that platform especially with the capacity of ram you want to run.
Reduce your expectations unless you want to upgrade the board cause' with a Zen 3 chip, that IMC will be more robust than when I did it on AM4 with a Zen+ chip so 3466MHz should be your max with bios tweaking for stability.
A 5800X3D really deserves an X570 board imo.

Yeah, in hindsight pushing a three year old motherboard to its limits with new gear probably wasn't the best idea. I hope 3200 will be able to run stably. As mentioned above, I ran a Windows memory test early in the debug process and it found no errors after 5 passes on the "advanced" setting. I may try running 3466 since that's the highest the board officially supports. Looking into X570 boards, apparently there is all of one option for that chipset in mATX form factor.

 

[eidairaman1]

In the notes I said I ran a Windows memory test for five passes on the "advanced" setting and it gave no errors. Can't really blame you for missing it, there's a lot there. Anyways, I set the system at A-XMP timings but with the frequency turned down to 3200 and ran a Prime95 Large FFTs test to stress test the RAM. It was stable after half an hour, which I know isn't really enough to fully test stability, but it's a start. I can understand that going from two small to four larger DIMMs and trying to crank up the speed could overstress the memory controller, it's just disappointing.

After running ir_cow's test, I then went back and disabled A-XMP and set it to 3200 with slightly tighter timings (16-20-20-40 instead of 18-22-22-42) and everything else set to auto. My BIOS is locked down and won't let me change any of the voltages besides the SOC Vcore. I really wish I could mess with voltages because man this processor is spicy. Any sort of stress test and it just pings straight to 90C. Running a stress test for the last 30min or so, it hasn't bluescreened yet.

I don't really have anything to add to this, I just like the profile picture. :3

Yeah, in hindsight pushing a three year old motherboard to its limits with new gear probably wasn't the best idea. I hope 3200 will be able to run stably. As mentioned above, I ran a Windows memory test early in the debug process and it found no errors after 5 passes on the "advanced" setting. I may try running 3466 since that's the highest the board officially supports. Looking into X570 boards, apparently there is all of one option for that chipset in mATX form factor.

Look at B550 motherboards as well

 

[Candor]

I've had a similar experience to this with my old X99 system. Various bluesceen codes, and Windows install failures.

It got to the point I could only successfully install Windows with only the install drive plugged in everything else disconnected.

After losing my mind for a couple of days, I finally was able to determine that the on-motherboard memory controller had become faulty, and 2 of the memory slots had issues.

 

[Shou Miko]

I don't really have anything to add to this, I just like the profile picture. :3

Thanks you, my thinking is just it could be a ram issue or memcontroller but since the memcontroller is in the CPU then you can only swap out the CPU to test in the board.

 

[ir_cow]

View attachment 257324

@Chomiq already pointed this out but basically that motherboard is not going to support high frequencies. 2x 32GB is only 2400 MT/s. 4x 32GB is probably not much better. I would buy a new MB that supports 3200. DDR4-3600 or even Gear 1 may be impossible with 4 dual-rank DIMMs.

 

[lowrider_05]

Maybe try it with 2 Sticks of Ram instead of all 4. For example, I can do 3800/C14 with 2 Sticks but only 3600/C22 with all 4 of them!

 

[mrluckypants96]

Alright, so I've got it running at 3466 with 18-22-22-42 timings across all four DIMMs and it appears to be stable. Thank you everyone for your help!

It seems the issue was that the board doesn't support 3600MHz speed at all, and when trying to use any of the presets that the motherboard offers, it tries to set the timings too tight and it won't post or won't be stable. Not sure why it still was having issues when at lower speed and loose timings, but it hasn't happened at all this week.

 

[Pictus]

Alright, so I've got it running at 3466 with 18-22-22-42 timings across all four DIMMs and it appears to be stable. Thank you everyone for your help!

It seems the issue was that the board doesn't support 3600MHz speed at all, and when trying to use any of the presets that the motherboard offers, it tries to set the timings too tight and it won't post or won't be stable. Not sure why it still was having issues when at lower speed and loose timings, but it hasn't happened at all this week.

For 4 RAM DIMMs you want a motherboard with 6 layer and better cooper wiring.
The only 6 layer PCB mATX motherboard for Ryzen I know is the Gigabyte B550M Aorus Pro AX

To check for VRM and PCB

VRM on the new AM4 motherboards

www.overclock.net

 

[eidairaman1]

Alright, so I've got it running at 3466 with 18-22-22-42 timings across all four DIMMs and it appears to be stable. Thank you everyone for your help!

It seems the issue was that the board doesn't support 3600MHz speed at all, and when trying to use any of the presets that the motherboard offers, it tries to set the timings too tight and it won't post or won't be stable. Not sure why it still was having issues when at lower speed and loose timings, but it hasn't happened at all this week.

It just means you would need to try and run at 3600 with looser timings than what xmp has

Yeah on this page it says up to 3466 OC

Overview B450M MORTAR

As a world leading gaming brand, MSI is the most trusted name in gaming and eSports. We stand by our principles of breakthroughs in design, and roll out the amazing gaming gear like motherboards, graphics cards, laptops and desktops.

www.msi.com

So if XmP3466 doesnt come up you would have to manually set it.