New-ish build is blue-screening randomly - SSD, port or BIOS issue?

[jw2k_fr]

Hello friends

I rebuilt my office PC over the winter, and have gone from having a rock solid PC which I didn't even think about to something that's become a flaky, neurotic, needy mess.

I went through several iterations of mini ITX motherboard - I changed from my old Asrock ITX to a Gigabyte B550i Aorus Pro ITX motherboard (because of GB's reputation both for build quality and support). I couldn't get hold of a 500 series Ryzen so I went with a Ryzen 7 3800XT. Rest of the build is as follows:

Gigabyte Aorus B550i Pro AX mini ITX
Ryzen 7 3800XT
Corsair Vengeance LPX DDR4 3600 64GB (2 x 32GB)
nVidia GT 1030 - now Sapphire RX480 (from my gaming machine)
Samsung 970 EVO+ 1TB NVMe Samsung BOOT DRIVE - cloned from the 860 EVO
860 EVO 1TB SATA
840EVO 500GB SATA
Fractal Design Ion+ 660P PSU
Noctua NH-D15

As a point of note, I did not do a clean install. The old Win10 (Intel drivers) booted without issue with the AMD hardware after it did a bunch of reorganization of drivers (which blew me away), so that's a possible contributing factor. Ideally I would have done this, but after I got rid of the initial hardware issues documented here (read 'most helpful critical review'), it's been rock solid for four months. I've also done the equivalent of several Service Pack installs as I've upgraded to 2H20 and now 1H21 updates of Win10 which would have written over a good part of the Win10 OS with clean files to help eliminate any possible file corruptions.

About a month ago I started getting random hard reboots - no warning, no blue screen, just *poof* and it would reboot. I struggled for a few weeks with that wondering whether it was a USB current draw issue (5 external, bus powered USB HDs connected), but fixed it by doing a boot repair (which actually worked for once!!). However, last night I got my first bluescreen reboot "Your system ran into a problem" and at lunchtime I had a second. Both happened while I was watching video (last night was local video file, today was 4K YouTube). I don't believe it's a GFX card issue as the 480 doesn't get hot enough to spin the cooling fans half the time.

Given that I appeared to have a file corruption issue last time, and that during both bouts of reboots I've seen the 970 NVMe drive disappear from the BIOS, I'm fairly sure it's either an SSD issue or something to do with the NVMe bus or port. This wouldn't occur if it was a simple file corruption or driver issue. I'm able to boot from the old 860 EVO SATA SSD and run for as long as I want without a reboot. This would seem to rule out the motherboard as an issue, although not the NVMe drive or that part of the bus.

Samsung Magician software reports that the 970 EVO is a little on the hot side at around 55-60C, but that's the only red flag I can find in the setup.


Would very much appreciate any input y'all can provide on similar issues you've seen and fixes you've tried which did or didn't resolve the issue.
Thanks in advance for any help!

 

[chameleoneel]

Hello friends

I rebuilt my office PC over the winter, and have gone from having a rock solid PC which I didn't even think about to something that's become a flaky, neurotic, needy mess.

I went through several iterations of mini ITX motherboard - I changed from my old Asrock ITX to a Gigabyte B550i Aorus Pro ITX motherboard (because of GB's reputation both for build quality and support). I couldn't get hold of a 500 series Ryzen so I went with a Ryzen 7 3800XT. Rest of the build is as follows:

Gigabyte Aorus B550i Pro AX mini ITX
Ryzen 7 3800XT
Corsair Vengeance LPX DDR4 3600 64GB (2 x 32GB)
nVidia GT 1030 - now Sapphire RX480 (from my gaming machine)
Samsung 970 EVO+ 1TB NVMe Samsung BOOT DRIVE - cloned from the 860 EVO
860 EVO 1TB SATA
840EVO 500GB SATA
Fractal Design Ion+ 660P PSU
Noctua NH-D15

As a point of note, I did not do a clean install. The old Win10 (Intel drivers) booted without issue with the AMD hardware after it did a bunch of reorganization of drivers (which blew me away), so that's a possible contributing factor. Ideally I would have done this, but after I got rid of the initial hardware issues documented here (read 'most helpful critical review'), it's been rock solid for four months. I've also done the equivalent of several Service Pack installs as I've upgraded to 2H20 and now 1H21 updates of Win10 which would have written over a good part of the Win10 OS with clean files to help eliminate any possible file corruptions.

About a month ago I started getting random hard reboots - no warning, no blue screen, just *poof* and it would reboot. I struggled for a few weeks with that wondering whether it was a USB current draw issue (5 external, bus powered USB HDs connected), but fixed it by doing a boot repair (which actually worked for once!!). However, last night I got my first bluescreen reboot "Your system ran into a problem" and at lunchtime I had a second. Both happened while I was watching video (last night was local video file, today was 4K YouTube). I don't believe it's a GFX card issue as the 480 doesn't get hot enough to spin the cooling fans half the time.

Given that I appeared to have a file corruption issue last time, and that during both bouts of reboots I've seen the 970 NVMe drive disappear from the BIOS, I'm fairly sure it's either an SSD issue or something to do with the NVMe bus or port. This wouldn't occur if it was a simple file corruption or driver issue. I'm able to boot from the old 860 EVO SATA SSD and run for as long as I want without a reboot. This would seem to rule out the motherboard as an issue, although not the NVMe drive or that part of the bus.

Samsung Magician software reports that the 970 EVO is a little on the hot side at around 55-60C, but that's the only red flag I can find in the setup.


Would very much appreciate any input y'all can provide on similar issues you've seen and fixes you've tried which did or didn't resolve the issue.
Thanks in advance for any help!

Do you have the SOC voltage in the bios, at 1.2v?

If not, try that. Some Ryzen board and CPU combos won't run fast RAM stable, without a bump to the SOC voltage.

*I'm also a fan of fresh Windows installs

 

[lopoetve]

Do you have the SOC voltage in the bios, at 1.2v?

If not, try that. Some Ryzen board and CPU combos won't run fast RAM stable, without a bump to the SOC voltage.

*I'm also a fan of fresh Windows installs

Do 1.1v first. 1.2 is a bit high for some boards. 1.1v, and set the DRAM voltage to 1.38v or so. Bet it boots right up stable as could be.

 

[jw2k_fr]

Do you have the SOC voltage in the bios, at 1.2v?

If not, try that. Some Ryzen board and CPU combos won't run fast RAM stable, without a bump to the SOC voltage.

This would be CPU VCORE SOC correct?

*I'm also a fan of fresh Windows installs

Yeah, me too. Just don't need the aggravation of trying to do a clean install at the moment when most of the contents of my office and my apartment are buried in storage, including physical copies of software with needed license keys

Instead... I thought I would try and do an update to Win10 21H2 which should be available through the Windows Insider preview program, however I couldn't get it to show up. Instead, I decided to over-write with a beta version of Windows 11... because that's bound not to cause any problems

More seriously though, from what I see the back-end is supposed to be largely the same as Windows 10, so hopefully it won't be as bad an all-new beta, and will get rid of any file corruption issues I have going on. Also, the release version of 11 should be out in a couple of weeks, so hopefully this is the short-lived lesser of two evils.

Thanks for the suggestions!

 

[chameleoneel]

This would be CPU VCORE SOC correct?


Yeah, me too. Just don't need the aggravation of trying to do a clean install at the moment when most of the contents of my office and my apartment are buried in storage, including physical copies of software with needed license keys

Instead... I thought I would try and do an update to Win10 21H2 which should be available through the Windows Insider preview program, however I couldn't get it to show up. Instead, I decided to over-write with a beta version of Windows 11... because that's bound not to cause any problems

More seriously though, from what I see the back-end is supposed to be largely the same as Windows 10, so hopefully it won't be as bad an all-new beta, and will get rid of any file corruption issues I have going on. Also, the release version of 11 should be out in a couple of weeks, so hopefully this is the short-lived lesser of two evils.

Thanks for the suggestions!

It could be named something different, depending upon the brand. or even where you are looking, in the Bios.

In my asus board, it something like VDDR SOC Voltage. You can see it on the same page as the CPU voltage adjustments.

But you can also adjust it in at least two other places, via the AMD overclocking menu. Which is in the "advanced" section of the Asus bios. and one of those other two places, its labeled completely differently as something way more straight forward, like SOC voltage or something like that.

 

[chameleoneel]

Do 1.1v first. 1.2 is a bit high for some boards. 1.1v, and set the DRAM voltage to 1.38v or so. Bet it boots right up stable as could be.

I think it depends upon the board, cpu, and memory combo.

On the combo I have running right now----I get all sorts of delayed clicks and processes where I have to often wait a few seconds, with default soc voltage.

with 1.13 voltage set, I get delayed clicks on just about everything. but no longer waiting several seconds.

with 1.2 for the SOC, the PC feels like it should.

 

[likeman]

with 3600 ram usually need to bump up the SOC voltage as long as it isn't over 1.2 normally fine (some motherboards automatically set it to 1.15v when ram XMP profile is set on the ram witch should also set ram speed, CL and ram voltage automatically)

 

[jw2k_fr]

Do 1.1v first. 1.2 is a bit high for some boards. 1.1v, and set the DRAM voltage to 1.38v or so. Bet it boots right up stable as could be.

It seems to be idling at around 1.09v, so I will try 1.1V to begin with and then 1.2V

DRAM voltage is set to 1.3V but is being reported at 1.38V in the status window to the right.... going to see how it goes for a bit

Thanks for all the suggestions!

 

[lopoetve]

It seems to be idling at around 1.09v, so I will try 1.1V to begin with and then 1.2V

DRAM voltage is set to 1.3V but is being reported at 1.38V in the status window to the right.... going to see how it goes for a bit

Thanks for all the suggestions!

Ram should be 1.35 or 1.4. Corsair at 64 or more I end up at 1.4 it seems

 

[jw2k_fr]

After playing around this weekend (which involved fitting a heatsink to the Samsung EVO 97) PLUS SSD which is installed in the M2 slot underneath the mobo (with zero airflow), I was able to drop my drive temps from 53C to 43C at idle. This, combined with the memory voltage tweaks and an upgrade to a late Beta of Windows 11 (to replace as many OS files as possible, in case of corruption) appears to have cured the reboots.

I'm fairly sure the issue was the drive as opposed to the memory. As it turns out, I was sever underclocking the RAM, so the voltages I was using were almost undoubtedly fine. Not a bad thing though to be back at full 3600Mhz speed using the XMP profile.

Thanks for all the very kind assistance!