Question Ryzen 3600 CPU blue screening on windows login

[Red Hawk]

Hi all. So I was messing around with a PC I had built for my brother a couple months ago, and had been operating without any issue in that time. I was trying to swap out the stock cooler for another cooler I had lying around. Turned out the other cooler wasn't compatible with the AM4 socket, didn't fit. So I put the stock cooler back on. I had completely disassembled the PC to do this, so I assembled it back together the same as it was before.

Upon reassembling and powering it on, the PC successfully POSTed, though it did give me a new hardware configuration detected message. I booted into Windows and...within 10 seconds, I got the BSOD. Rebooted, same thing. Went into the BIOS and made sure everything was set to the defaults. The PC was consistently BSODing within seconds upon booting into Windows, sometimes before logging in and sometimes after. It even BSODed when booting into safe mode. And the error message was different each time. Sometimes a memory controller error, sometimes an ntfs.sys error, sometimes system service exception. But the crash behavior was consistent.

So for troubleshooting I trying disassembling and reassembling the whole PC again to make sure I hadn't messed up reinstalling anything. Same behavior. Then I took each RAM stick out individually and put the RAM sticks in different slots, to eliminate the chance that a stick had spontaneously gone bad. Same BSOD. Next I swapped out the Ryzen 3600 CPU for a Ryzen 2600 CPU from another PC I had access to. And the PC successfully booted into Windows, appears to be stable and I'm typing this message from that PC.

So...the problem is isolated to the 3600. My immediate thought is that the pins on the 3600 were somehow damaged, but upon inspecting the CPU I don't see any damage. My other thought is that the thermal paste I use somehow got around one of the pins and is causing it not to make full contact with the socket, but I don't see any sign of that either. So...any thoughts on what could be causing this?

The PC in question is not the one in my signature. The relevant specs, aside from the 3600, are as follows:

MSI B450 Gaming Pro Carbon
2x8 GB Evo Potenza 3200 MHz DDR4 RAM
Gigabyte Radeon RX 580 8 GB
Crucial P1 1 TB NVMe SSD
Powerspec 650w 80+ Bronze PSU
Windows 10 64 Bit Home

All of the parts are no more than a couple months old, they were all bought brand new.

 

[Red Hawk]

Update: Ok, I took the 2600 out of my brother's PC and put the 3600 back into it, tried booting it up. BSODed the same way it had before. So, to eliminate the possibility of there being some corrupted file or setting on my brother's PC that just was not playing nice with the 3600, I put the 3600 in the alternate PC that the 2600 came from.

...and I'm typing this message from that PC now, with the 3600 installed. So. The CPU isn't dead, because another PC seems perfectly stable with that CPU installed.

I guess the thing to do now is reinstall Windows on my brother's PC.

 

[randomhero]

If that does not work, try booting from another plug and then with another PSU.
Electrical problems can hide as something else.
Good luck!

 

[coercitiv]

See if changing input voltage for the 3600 on the problem system has any effect (introduce something like +0.02-0.05 offset). Try this with RAM @ 2133 Mhz.

If the 2600 is still inside the problematic system, have you tried running heavy workloads on it? (Prime 95 for example, as it may also catch borderline instability)

 

[DAPUNISHER]

Don't just ensure everything in the bios is defaults. Kill the power, pull the battery, let it sit a bit, maybe repeat the the spell of making a few times, and then give it another go.

 

[Red Hawk]

Ok, finally reinstalled Windows on my brother's computer. He had been using the 2600 on his computer for the last couple months without issue. Ran the reinstallation with the 3600 installed, not the 2600. The installation had one hitch...midway through, I got an "installation failed" message, that if I want to install Windows I have to restart the installation. I hit ok...and the PC continued with installing Windows. I It finished the installation, and was able to boot into the new Windows installation without crashing. I was able to run Windows Update, and rebooted after the updates.

Upon updating though, I logged into Windows and...boom. Blue screen. Back to the drawing board...going to try some of the other suggestions here.

Edit:

So I booted back into Windows successfully. It was behaving different than before reinstalling Windows. Before, it was BSODing within 10 seconds of logging into Windows with the 3600 installed, every time. It wasn't repeatedly BSODing on me this time. I finished installing all the motherboard and graphics drivers, then I tried running Prime95. Upon starting the test, boom. Instant blue screen. So the system could boot reliably, but it was still unstable. So...

See if changing input voltage for the 3600 on the problem system has any effect (introduce something like +0.02-0.05 offset). Try this with RAM @ 2133 Mhz.

Did this. Set a +0.05 offset and ran the memory at 2133 MHz. Ran Prime95 for 15 minutes with no issue. Set the memory back to 3200 MHz but kept the offset. Running Prime95 now with no issue. The voltage offset seems to ensure the computer's stability. Is there any downside to keeping this offset?

 

[coercitiv]

Did this. Set a +0.05 offset and ran the memory at 2133 MHz. Ran Prime95 for 15 minutes with no issue. Set the memory back to 3200 MHz but kept the offset. Running Prime95 now with no issue. The voltage offset seems to ensure the computer's stability. Is there any downside to keeping this offset?

No immediate downside considering the CPU is running at stock, but it's not a good sign, as it means some type of degradation might have happened in the power delivery chain. (PSU or motherbaord, assuming BIOS settings are all at stock except the voltage offset, including stock LLC settings).

You can try leaving the offset and see if problems come back in days/weeks/months even with the offset. If that happens, don't increase the voltage any further, but instead bring back settings to stock which will likely exacerbate instability (system might not even manage to POST) and test the system with another power supply. If it's stable, then it's the original power supply, if it's still unstable then it's probably the mainboard acting up. (you alreadu had the CPU in another system IRC, and it was deemed stable)

 

[Chicken76]

Besides resetting the BIOS, I would take out all the memory modules and re-insert them again, paying attention to do it properly. If you took out the motherboard to fit that cooler, the DIMMs might have been knocked/moved and not sit properly in the slots. It's enough for one of them to not make proper contact to experience that kind of instability. Validate the RAM is OK with memtest86+.

 

[Red Hawk]

No immediate downside considering the CPU is running at stock, but it's not a good sign, as it means some type of degradation might have happened in the power delivery chain. (PSU or motherbaord, assuming BIOS settings are all at stock except the voltage offset, including stock LLC settings).

You can try leaving the offset and see if problems come back in days/weeks/months even with the offset. If that happens, don't increase the voltage any further, but instead bring back settings to stock which will likely exacerbate instability (system might not even manage to POST) and test the system with another power supply. If it's stable, then it's the original power supply, if it's still unstable then it's probably the mainboard acting up. (you alreadu had the CPU in another system IRC, and it was deemed stable)

Alright, I'll see how it goes. If the problem comes back, I'll transplant a PSU from another (relatively new) PC I have access to. If that solves the problem, I'll RMA the original PSU. If it doesn't, seems like I'll have to RMA the motherboard.

Thanks for the help!

Besides resetting the BIOS, I would take out all the memory modules and re-insert them again, paying attention to do it properly. If you took out the motherboard to fit that cooler, the DIMMs might have been knocked/moved and not sit properly in the slots. It's enough for one of them to not make proper contact to experience that kind of instability. Validate the RAM is OK with memtest86+.

One of the first things I did when this started was reseat each of the DIMMs, trying them individually and in different slots. Didn't resolve the problem, thanks though! I'll double check with memtest to be sure.

 

[blckgrffn]

Man, I love it when I get punished for good intentions like this too, OP.

I am glad you seem to have found a happy place, as others have said make sure you pounce if there is future instability. If it's the PSU it could fail spectacularly and cost other components their life.

 

[VirtualLarry]

Hmm, as long as you didn't actually OC the CPU, I would consider RMA'ing it, and maybe follow a little stricter ESD precautions in the future.

 

[Red Hawk]

Ok, the PC went unused for a couple weeks for unrelated reasons, and when it was booted up after that it seemed to fail to POST with the last settings (memory running at 3200 MHz and CPU offset of +0.05) The PC was not completely unusable though, it would show a message along the lines of "Memory overclock failed, press F1 to go to BIOS". So I could still go into the BIOS and turn the memory down to 2133 MHz. With that setting, the PC is able to POST and boots into Windows fine. But obviously running at about 60% memory speed is not ideal. Guess I'll try transplanting a PSU next.

 

[DAPUNISHER]

Ok, the PC went unused for a couple weeks for unrelated reasons, and when it was booted up after that it seemed to fail to POST with the last settings (memory running at 3200 MHz and CPU offset of +0.05) The PC was not completely unusable though, it would show a message along the lines of "Memory overclock failed, press F1 to go to BIOS". So I could still go into the BIOS and turn the memory down to 2133 MHz. With that setting, the PC is able to POST and boots into Windows fine. But obviously running at about 60% memory speed is not ideal. Guess I'll try transplanting a PSU next.

I am going to ask a stupid question, because the last time I saw this issue on AM4 was due to the answer being no. You do have the ram in A2&B2 correct? The other ram issue on AM4 I see most frequently is the bios. Newer bios often resolve ram related woes on AM4. Make certain you have the latest AMD all-in-one drivers before updating the bios, too.

 

[Shmee]

Have you memtested yet? I would be interested in the results. Also DAPUNISHER makes a good suggestion on the BIOS update.

 

[Red Hawk]

Ok...I tried swapping the PSU with a relatively new PSU from another PC, and it had no effect on the issue. What I've done now is disabled XMP and manually set the RAM speed to the full 3200 MHz. That made the message about "memory overclock failed" go away and allowed the PC to boot into Windows, and validated as stable with Prime95. The +0.05 CPU offset is still necessary for stability, though.

So...with XMP enabled and set to the default 3200 MHz profile, the PC refuses to boot. With XMP disabled and memory manually set to 3200 MHz, the PC boots fine and is stable (with the offset). I think the darn PC is just being fickle at this point.

I am going to ask a stupid question, because the last time I saw this issue on AM4 was due to the answer being no. You do have the ram in A2&B2 correct? The other ram issue on AM4 I see most frequently is the bios. Newer bios often resolve ram related woes on AM4. Make certain you have the latest AMD all-in-one drivers before updating the bios, too.

Not sure if the DIMMs were installed in A1/B1 or A2/B2, but in any case, I did change them over and test them individually in different slots when this all started. I also made sure to flash the mobo with the latest BIOS and updated to the latest chipset drivers.

Have you memtested yet? I would be interested in the results. Also DAPUNISHER makes a good suggestion on the BIOS update.

I ran several passes with memtest, and it never registered any errors.

 

[DrMrLordX]

So...with XMP enabled and set to the default 3200 MHz profile, the PC refuses to boot. With XMP disabled and memory manually set to 3200 MHz, the PC boots fine and is stable (with the offset). I think the darn PC is just being fickle at this point.

XMP profiles can change settings like vddg, vddp, and SoC voltage without alerting the end user. You may have a touchy mem controller though, and/or some funky power delivery issues with the board (hence the requirement for a positive voltage offset).

 

[Red Hawk]

Alright, I think I've isolated the issue to the motherboard.

Since finding that XMP cause the PC to not even POST and thus disabling XMP and manually setting the memory to its rated 3200 MHz, I noticed one more issue. The Cinebench R20 score on my brother's PC was well below what it should have been. Opening up MSI Afterburner to monitor clock speeds revealed why -- it was downclocking to around 2.7 GHz under while running Cinebench. So I opened up Ryzen Master, enabled PBO, ran Cinebench again -- the clock speed went back up to boosting at about 3.8 GHz. Though running Prime95 revealed some wonky behavior still -- it would boost up to 4 GHz for Prime95, but after about seven minutes it dropped down to around 3.6 GHz.

So I wanted to decisively isolate if it was a problem with the CPU or the mobo. I took my brother's 3600 CPU and put it in my current build (in my signature), and took my 3800X and put it in his PC. Set both to the settings that were getting near immediate blue screens on my brother's PC with his CPU -- XMP enabled automatically to 3200 MHz, everything else stock. Ran Prime95 on both of them. Both appeared to be stable...however, opening up MSI Afterburner revealed that my 3800X in his PC was downclocking to 2.7 GHz. My brother's 3600 was running at a full 4 GHz in my PC. So I can't get his PC to crash with my 3800X installed, but it is demonstrably not running properly. I've already eliminated his PSU and memory as the issue. So, I think I can safely say the issue is with the MSI B450 Gaming Pro Carbon mobo in my brother's PC.

So I've recommended to him that he RMA his motherboard so MSI can either fix it or replace it. He's hesitant, as technically his motherboard and CPU can run stably at full speed -- but only with a +0.05 offset, XMP disabled and memory manually set to 3200 MHz, and PBO enabled. That's not ideal, and I've told him that keeping it at those settings could degrade the components. He's hesitant though, as a friend of his has his own homebuilt PC with an i7-8700K that he's been overvolting by .15v for over a year and it's still running fine. So, it's really up to him now if he wants to ship it in or just live with the quirky voltage settings. Does it sound to you guys like the mobo should be RMAed?

(Side note on his CPU cooler -- I recently upgraded to a Corsair H115i in my own build, and transferred the Wraith Prism that came with my 3800X to my brother's PC. So it's using that now, not the Wraith Stealth that the 3600 originally came with.)