Multiple problems Ryzen build (black screen, BSOD, Prime95 failing, freezes)

[191x7]

My PC seems to have got a poltergeist. Any Ghostbusters here?

 

My configuration is under my profile, but I'll also add it here https://docs.google.com/spreadsheets/d/1zFdH7zP_pc6SdkQQ-hEYbVBAdGG9ihRlqzYBg9vvAWc/edit?usp=sharing so you can find it faster.

Some parts are old, but most are under 3-4 years, the platform some 6 months, and some parts (like the SSD-s) are under a month.

The OS is from the clean install I did when upgrading to Ryzen last December, cloned it from a 120GB SSD to the new one before Easter.

 

Problems started happening a few weeks ago (maybe a month). At first really rare, but the issues started increasing and now they happen often.

 

1st issues noticed (main issues!) - the PC would enter a hibernation/sleep "like" state with no image on the display and not reacting to input.

After hitting the reset switch it would boot, and when it happens holding the power switch also does shut down the machine.

On next boot, the mouse would stop working and has to be re-plugged to work.

I thought it to be AsMedia USB drivers (to be blunt, not a fan of the company, had issues with those on any board using then), but with different driver versions, nothing changed.

One time (I think the first time it happened), BlueScreenView showed a ntoskrnl BSOD after the next reboot, but that was only that one time.

This issue happens unrelated to the PC utilization; happens when booting into Windows, when resuming from standby (or whatever it's called when you take Win10 sleep and remove hibernation), when IDLE or while surfing.

 

2nd issue noticed (after one of the "resets") - Chkdsk and SFC fixed something. Happened only once. DISM clean, SFC fine. I believe it to be related to the forced shutdown, not a sign of the issue.

 

3rd issue - the Samsung F3 disk had 1 sector marked for reallocation (for a long time), a few weeks back it showed 270 bad sectors and 273 marked sectors.

A few days later (since I did not have the time to remove the drive) the health returned to 100% and S.M.A.R.T. shows no bad sectors nor marked sectors.

Since the first marked sector, the drive was emptied and used only for unimportant stuff. Everything from/on the drive works.

The drive is some 10 years old. I might remove it to see if the main issue is somehow related to this.

 

4th issue - the last time I ran Prime95 Blend the "Core 11" stopped after an error (got 0.5 when it should have got 0.4). The CPU was never overclocked (I didn't have time to do it and hardly any need to do it).

Temperatures (with the fans set to "Quiet) did not exceed 62°C so it wasn't overheating. Ran Passmark Memtest86 after that and the 4 complete passes passed without issues (with the 3200MHz XMP).

This was last weekend, I did not have time to re-run Prime95 again, but I did not change anything that could affect the 4th issue, except updating the bios from f31 to f40 (latest).

[What can I say, updating the BIOS might solve the issues but the issues cause the PC to crash so there's a huge risk in updating. Not a nice feeling accepting the risk.]

 

5th issue (strongly related to the 1st issues) - the last thing I did (last night) was cleaning the old drivers (uninstall + DDU) and installing the latest chipset and GPU drivers from AMD.

After the procedure, the PC not only crashes (the 1st issue) but also completely FREEZES. The "freeze" like you've set a wallpaper on the screen and disconnected the peripherals. Image stuck, nothing reacts.

This time, WhoCrashed reads the DUMP files and shows an AMD related BSOD, atikmpag.sys. So the freezes aren't, they're BSOD-s.

 

I let Windows update to 1903 last night. Will check if it did something that could mitigate the issue(s).

 

What I might/will do next (not necessarily in that order):

- Check if the update to 1903 helped. No high hopes though.

- Clean the GPU and replace the paste (the card is 3,5 years old, max temps don't go over 75°C but the paste has never been changed). Don't think it's actually needed.

- Remove the Samsung F3 drive and maybe the WD drives (those were added two months ago, used drives with a clean S.M.A.R.T. status).

- Remove all the drives and clean install Windows to a 120GB SSD to see if it would continue happening. (might do this first, would help a lot)

- Check if the issues (the 1st one) happen under a live Linux.

- Try using the PC without it being connected to the C-Lion Blazer 800VA 480W UPS.

- Trying some really old drivers that I know did not cause any issues (for example the ones I used 5 months ago).

 

No, it's not the PSU. The one I have is a year and a half old and the system doesn't even draw 400W under MAX load (Prime95+Furmark).

 

I just hope it's not the new ADATA 8200pro SSD (but if it is, I can RMA it) or the graphics card.

I had planned to upgrade the graphics card with a new one closer to the end of the year. Maybe the Na'Vi that will perform close to an RTX 2070. But I don't have the funds at the moment.

If it's one of the hard drives, not a biggie, but if it's multiple of them causing the issues together that might pose a problem.

And no, I don't like reinstalling Windows just when I got the previous installation set up the way it should be.

 

Can you help? Any idea on what to do next? Maybe encountered a similar issue?

[Skiiwee29]

Does these issues continue when everything is on factory default and not overclocked?

[191x7]

7 minutes ago, Skiiwee29 said:

Does these issues continue when everything is on factory default and not overclocked?

Everything is on factory default and not overclocked except XMP. But happens even when everything is on default.

[GoldenLag]

3 minutes ago, 191x7 said:

Everything is on factory default and not overclocked.

might be reffering to XMP. you might not have gotten unlucky with a poor memmorycontroller. 

[191x7]

1 minute ago, GoldenLag said:

might be reffering to XMP. you might not have gotten unlucky with a poor memmorycontroller. 

But the issues weren't there, I have the platform since early December.

Unless if the bios updates... but then it would hardly pass Memtest (although, I'm no Linus to let it run for 3-5-10 days).

[scottyseng]

Could also try running memtest x86. I would also say to update the BIOS (I'm doubtful it would crash when no OS is being loaded because you're updating through BIOS).

 

Otherwise, you might've lost the silicon lottery like my friend did with his 1700 (Was not stable at stock clocks with stock voltage. I had to overvolt it to get it stable)

 

Nevermind, saw that you ran memtest already.

[191x7]

1 minute ago, scottyseng said:

Could also try running memtest x86. I would also say to update the BIOS (I'm doubtful it would crash when no OS is being loaded because you're updating through BIOS).

 

Otherwise, you might've lost the silicon lottery like my friend did with his 1700 (Was not stable at stock clocks with stock voltage. I had to overvolt it to get it stable)

Memtest86 has been run. And the bios is the latest. Didn't I mention that? My bad.

[porina]

On Prime95, if you still have the install, look in the results log and see what FFT size it was running when it failed? If you run it again on small FFT, what actual core clock and voltage does your system use in that state? Oh, what version of Prime95 are you using?

 

Apart from that, as wild guesses, you could try reseating both CPU and ram.

[Skiiwee29]

With so many seemingly random issues all starting to rear itself randomly.. im starting to think you may have a bad board. I had a similar problem with an x370 Taichi and after spending a few months of troubleshooting it, I swapped boards to the Crosshair VI Hero and all my troubles went away. As you mentioned in #1, you have a pretty old drive in there. I would definitely remove any drive not used for boot and test it again for a few days/week. Basically take it down to bare bones only needed to boot and browse. 

[191x7]

8 minutes ago, porina said:

On Prime95, if you still have the install, look in the results log and see what FFT size it was running when it failed? If you run it again on small FFT, what actual core clock and voltage does your system use in that state? Oh, what version of Prime95 are you using?

 

Apart from that, as wild guesses, you could try reseating both CPU and ram.

I will check it when I get home. I'm using the latest Prime95 if I'm not mistaken.

Do you think it might be voltage related? I totally did not think of that. I must have gotten rusty.

 

8 minutes ago, Skiiwee29 said:

With so many seemingly random issues all starting to rear itself randomly.. im starting to think you may have a bad board. I had a similar problem with an x370 Taichi and after spending a few months of troubleshooting it, I swapped boards to the Crosshair VI Hero and all my troubles went away. As you mentioned in #1, you have a pretty old drive in there. I would definitely remove any drive not used for boot and test it again for a few days/week. Basically take it down to bare bones only needed to boot and browse. 

Might be the case, but I have to rule everything out first. When I'm sure it's the board, I can RMA it, but at the moment I don't have a spare AM4 board.

[porina]

1 hour ago, 191x7 said:

I will check it when I get home. I'm using the latest Prime95 if I'm not mistaken.

Do you think it might be voltage related? I totally did not think of that. I must have gotten rusty.

Ok, just wondering in case it was an ancient version. I don't know what it might be, but we can look at parts and try to narrow it down. By knowing the FFT size it failed on, that could narrow things down more. Smaller FFTs would be more indicative of the core being unstable. Stable small FFT but unstable large FFT would implicate more the IMC and ram. There's a transition area where caches could come into play too.

[cretsiah]

@191x7 cant help with all of those issues,

 

but the hibernate/sleep problem could be related to the power settings/management.

 

for instance I have never turn off display..... however further down the list of options I had to also disable PCI management to avoid the monitor turning off.

 

this stupid PCI express management setting would cause the screen to turn off (due to no signal from graphics card to monitor), which required rebooting to repower the graphic card slot.

[191x7]

13 hours ago, porina said:

Ok, just wondering in case it was an ancient version. I don't know what it might be, but we can look at parts and try to narrow it down. By knowing the FFT size it failed on, that could narrow things down more. Smaller FFTs would be more indicative of the core being unstable. Stable small FFT but unstable large FFT would implicate more the IMC and ram. There's a transition area where caches could come into play too.

I've attached the result of the Prime95 test. The version is 29.4 build 8.
 

Prime95 crash.txt

[porina]

Just now, 191x7 said:

I've attached the result of the Prime95 test. The version is 29.4 build 8.

8k FFT is tiny and easily fits within L2 cache, maybe even L1. That would point towards the core being unstable, not IMC or ram.

 

Might be worth trying a small voltage offset increase?

[191x7]

Turns out the latest bios was faulty without updating something first.

Gigabyte released the bios F40 with a short description, then they added a tool to upgrade something WITHOUT any description and then today updated the description of the F40 bios stating that you need to use the tool before upgrading to F40 for f40 to work fine.

And also MSI Afterburner. 

 

Since I uninstalled Afterburner and flashed F40 the way it was supposed to be I did not experience any issues. It has only been a couple of hours so I can't say for sure.

[scottyseng]

Hopefully that fixes it. My other friend's ryzen build I had to update the BIOS to get it stable because they added support for the latest windows OS build

[191x7]

On 5/23/2019 at 4:44 AM, scottyseng said:

Hopefully that fixes it. My other friend's ryzen build I had to update the BIOS to get it stable because they added support for the latest windows OS build

 

The main issue (black screen) has returned, even after a clean Windows install.