• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

GPU dead out of nowhere? Weird symptoms.

Joined
Feb 29, 2016
Messages
659 (0.20/day)
Location
Chile
System Name Shinano
Processor AMD Ryzen 9 5900X
Motherboard ROG Strix B550-F GAMING WIFI II
Cooling Thermalright Peerless Assassin 120SE
Memory 32GB DDR4 3200MHz
Video Card(s) XFX RX 6700 10GB
Storage 970 EVO Plus 1TB | A1000 480GB
Display(s) Lenovo G27q-20 (1440p, 165Hz)
Case NZXT H510
Audio Device(s) Sony WH-1000XM4 | Edifier R1000T4
Power Supply SuperFlower Leadex Gold III 850W
Mouse Logitech G305
Keyboard Nuphy Air 75 | IK75 v3 (QMK) | HyperX Alloy Origins
Title... again. My two months old XFX 6700 just died in the weirdest way possible.

My PC was in sleep mode, woke it up and the screen lighted up as usual but 10 seconds later it filled with what I can only describe at static and it shut off. Previous to this (like before making the PC go to sleep) I was playing Cyberpunk 2077.

I got to safe mode and used DDU, which landed me a display. Restart, install drivers, same issue.

Oh well, I reinstalled Windows after the fact to make sure. Nope, same issue as soon as the automatic driver installation happened.

Here's a video:

Whole screen becomes static and then it dies.

Now for what makes this bizarre: if I put a Linux (tried Fedora GNOME and Manjaro KDE) and turn on my PC with it, it... works. I can use my PC and normal and even fire up light 3D stuff, watch YouTube videos and everything else.

But switch back to a new install of Windows and boom, dead.

I haven't tried actual games on the GPU under Linux, but I'm utterly confused. How does that even make sense?

So far I've tried:
- reset bios to defaults: nope
- forcing the slot to pcie gen 3: nope
- reseat everything: nope
- praying: nope
- new windows install: nope
- trying under linux: yes?

EDIT: It was the DP cable. It was the forsaken cable lol.
 
Last edited:
Weird that praying didnt work ^_^.
Did you try to reseat the GPU in the slot?
Maybe a vBIOS reflash could work?
 
Try disable driver enforcement.
How so? also honestly can't really get windows to boot now, getting to safe mode is nigh on impossible.

Weird that praying didnt work ^_^.
Did you try to reseat the GPU in the slot?
Maybe a vBIOS reflash could work?
Tried reseating everything really. Also Furmark works under Linux ugh. So we have 2D *and* 3D working under Linux but... can't do anything on Windows? The moment even the automatic driver install kicks in, it's dead.
 
welp if it displays ok with Linux, W10 driver perhaps? Sorry no clue how get around this, but the GPU etc is still Ok. This sounds like the old time Catalyst driver issues many had back in time.
Is this on W10 or W11? But to get going, try another graphics card, then start working on troubleshooting from that onwards.
 
Last edited:
How so? also honestly can't really get windows to boot now, getting to safe mode is nigh on impossible.


Tried reseating everything really. Also Furmark works under Linux ugh. So we have 2D *and* 3D working under Linux but... can't do anything on Windows? The moment even the automatic driver install kicks in, it's dead.
Could you try install some older drive from safemode?
 
welp if it displays ok with Linux, W10 driver perhaps? Sorry no clue how get around this, but the GPU etc is still Ok. This sounds like the old time Catalyst driver issues many had back in time.
Is this on W10 or W11?
Windows 11. Seems to also happen when the automatic driver install kicks in, which installs an eons-old driver.
This is weird, because the static pattern seems to be more of a DP handshake issue, but then it works fine under Linux.

I could try an HDMI cable or another DP cable, but this is pretty weird (and don't have any on hand)

Could you try install some older drive from safemode?
Windows Update installs a driver from mid 2022, so that's definitely old, and it fails. But yeah, I wanted to do that but... yeah, can't even get into safe mode now.

But yep, we have furmark working under Linux. Go figure. Can't even boot into a NEW windows install without blackscreening.
20230310_223531.jpg
 
Windows 11. Seems to also happen when the automatic driver install kicks in, which installs an eons-old driver.
This is weird, because the static pattern seems to be more of a DP handshake issue, but then it works fine under Linux.

I could try an HDMI cable or another DP cable, but this is pretty weird (and don't have any on hand)


Windows Update installs a driver from mid 2022, so that's definitely old, and it fails. But yeah, I wanted to do that but... yeah, can't even get into safe mode now.

But yep, we have furmark working under Linux. Go figure. Can't even boot into a NEW windows install without blackscreening.
View attachment 287313
I think you should try installing a driver from safemode , download the driver and remove the internet connection, maybe that will help.
 
I think you should try installing a driver from safemode , download the driver and remove the internet connection, maybe that will help.
That is the second thing I tried. After restarting from DDU and installing the drivers the screen would flash, show static and turn off, sadly.

Still weirded out. On Linux I've tried countless GPU stress tests both light and harsh. Everything works.
 
Well the only things i can think you can still try is to install win10 instead of 11 or try to reflash the vBIOS (or if you have dual bios just switch).
 
Once you boot into Windows again somehow (even if you have to install it again), open command prompt as administrator and type:

bcdedit /timeout 30
bcdedit /set {globalsettings} advancedoptions true

This will display the startup options such as safe mode and last known good configuration (which ignores recently installed drivers), and low-resolution mode (forces 640x480 60 Hz on boot), every time you boot the PC, and give you 30 seconds before it boots. This will make troubleshooting the problem from the Windows side much easier.

The 22.5.1 driver is regarded in the AMD community as exceptionally stable, keep a copy of it handy in addition to the newest one: https://www.amd.com/en/support/kb/release-notes/rn-rad-win-22-5-1

CRU utility to clean the monitor store is also going to be useful: https://www.monitortests.com/forum/Thread-Custom-Resolution-Utility-CRU

Download and run the AMD Cleanup Utility if you manage to somehow recover your current install: https://www.amd.com/en/support/kb/faq/gpu-601

Make a registry file to disable automatic driver updates from Windows Update, save this as .reg, merge and then reboot:

Code:
Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SOFTWARE\Policies\Microsoft\Windows\WindowsUpdate]
"ExcludeWUDriversInQualityUpdate"=dword:00000001

Try installing 22.5.1 and see if the monitor goes blank, good luck!
 
Last edited:
Try csm off if not. If rebar is auto, csm will blank out video.
 
Uh huh... I tried to reinstall Windows and now I'm in a sticky situation.

20230310_230642.jpg



I... don't think this is meant to happen lol. Now I wonder if the *drive* died instead (the linux system runs off a live usb?). Guess I'll try off my spare SATA SSD.
 
Uh huh... I tried to reinstall Windows and now I'm in a sticky situation.

View attachment 287316


I... don't think this is meant to happen lol. Now I wonder if the *drive* died instead (the linux system runs off a live usb?). Guess I'll try off my spare SATA SSD.

It's possible, but I think you ran into a rare bug that AMD has acknowledged recently:

 
It's possible, but I think you ran into a rare bug that AMD has acknowledged recently:

I've tried reinstalling Windows a few times. Like, once when I realized the issue and once when I couldn't get into safe boot now. Nope, nothing.

Installing on the SATA SSD now.

It's possible, but I think you ran into a rare bug that AMD has acknowledged recently:

Nope. Same issue. I guess I corrupted the old NTFS partition by shutting down the PC mid-setup as it went blank, now I waited a bit.

Here's a better video:

This is in a completely fresh install. Slow it down to 0.25x at 00:05

Also, reconnecting the cable shows an image for a split second then it does the same as the last video.
I would chuck it off to a bad DP cable but... it works under linux?

Try installing 22.5.1 and see if the monitor goes blank, good luck!
Yep. Static then blank.
 
Last edited:
I've tried reinstalling Windows a few times. Like, once when I realized the issue and once when I couldn't get into safe boot now. Nope, nothing.

Installing on the SATA SSD now.


Nope. Same issue. I guess I corrupted the old NTFS partition by shutting down the PC mid-setup as it went blank, now I waited a bit.

Here's a better video:

This is in a completely fresh install. Slow it down to 0.25x at 00:05

Also, reconnecting the cable shows an image for a split second then it does the same as the last video.
I would chuck it off to a bad DP cable but... it works under linux?


Yep. Static then blank.

This reminds me of a problem I had a long time ago with the R9 Fury X... say what? Does HDMI work? If HDMI works, try setting up Windows using that even if it's a nuisance, install the driver normally and plug it on DP again. Even if it does glitch up and go blank like that, the driver will write a file named atikmdag_dce.log in C:\Program Files\AMD that will be very useful in saying what went wrong. This file contains monitor EDID, the status of the last display link, and relevant frequency data. Good luck, man. Looks like you have a long night of troubleshooting ahead of you...
 
I'll have to buy an HDMI cable first. Somehow got none of those. Will do. The current install is serviceable (can enter safe mode and can get to the login screen before it goes blank) so at least I don't have to keep reinstalling Windows.

This reminds me of a problem I had a long time ago with the R9 Fury X... say what? Does HDMI work? If HDMI works, try setting up Windows using that even if it's a nuisance, install the driver normally and plug it on DP again. Even if it does glitch up and go blank like that, the driver will write a file named atikmdag_dce.log in C:\Program Files\AMD that will be very useful in saying what went wrong. This file contains monitor EDID, the status of the last display link, and relevant frequency data. Good luck, man. Looks like you have a long night of troubleshooting ahead of you...
It was the cable. Damnit how did it work on Linux? lmao

20230311_114355.jpg
 
Nice! I'm glad you got it sorted out. That's a pretty good question, too! Enjoy, my friend. :lovetpu:
 
What lead me to this whole "might be the cable" was that linux (well, the amdgpu driver and X11/Wayland modeset implementation) is actually WAY better at dealing with broken cables. If the DP handshake fails, it will chug along with the first EDID list the monitor sent at boot time and ignore the malformed data.

Windows seems to take the wrong DP handshake data and request the mode to the monitor (I think it was requesting 1440p@165Hz 10-bit, which isn't possible with this cable), which caused a monitor black-out (technically "unsupported mode" but I guess this monitor just bails out).

Old cable works now, if I run it in a different place than it was before, btw. My best guess is that since in the cable management the old cable was hugging the monitor power cable, the DP cable has bad shielding and it was receiving electrical noise (maybe the power was dirtier than usual?), which corrupted the data the DP cable was meant to send for the handshake.

Pretty cursed, honestly. At least it doesn't seem to be any kind of GPU issue, but goes to show how cursed troubleshooting can be, and I don't really know what sparkled this to happen out of nowhere ('twas working for the past year and a half on that run)
 
Last edited:
What lead me to this whole "might be the cable" was that linux (well, the amdgpu driver and X11/Wayland modeset implementation) is actually WAY better at dealing with broken cables. If the DP handshake fails, it will chug along with the first EDID list the monitor sent at boot time and ignore the malformed data.

Windows seems to take the wrong DP handshake data and request the mode to the monitor (I think it was requesting 1440p@165Hz 10-bit, which isn't possible with this cable), which caused a monitor black-out (technically "unsupported mode" but I guess this monitor just bails out).

Old cable works now, if I run it in a different place than it was before, btw. My best guess is that since in the cable management the old cable was hugging the monitor power cable, the DP cable has bad shielding and it was receiving electrical noise (maybe the power was dirtier than usual?), which corrupted the data the DP cable was meant to send for the handshake.

Pretty cursed, honestly. At least it doesn't seem to be any kind of GPU issue, but goes to show how cursed troubleshooting can be, and I don't really know what sparkled this to happen out of nowhere ('twas working for the past year and a half on that run)
The electrical noise idea is interesting - the only time I've ever gotten static was with my old monitor and cable, where if I had any static electricity built up on me and I touched the metal computer case, or sometimes anything metal at all in the room, even across the room, the screen would show moving static for 2-3 seconds, then black out and recover (never had any corruption of any sort from it).
 
The electrical noise idea is interesting - the only time I've ever gotten static was with my old monitor and cable, where if I had any static electricity built up on me and I touched the metal computer case, or sometimes anything metal at all in the room, even across the room, the screen would show moving static for 2-3 seconds, then black out and recover (never had any corruption of any sort from it).
Another option someone gave me is that one or two cables were bad and running the old cable differently "fixed" it. Only the god of computers knows at this point.
 
That's a pretty good question, too! Enjoy, my friend.
Did some research and it seems like it just... happens? https://linustechtips.com/topic/725578-displayport-suddenly-not-working/

Same result (tho I had signal before it switched to 165Hz), same trigger (computer coming back from sleep), a lot of comments saying there was a similar issue. Would also explain why the old cable works now, outside what could've happened with higher electrical noise.

Man, computers are weird.
 
Last edited:
Do tell, I use HDMI in general due to using TVs as a display but this one would definitely have me worried.

That file I brought up earlier, atikmdag_dce.log would be quite useful to know if that's precisely the problems you ran into, as it has the link speed negotiation history logged in there.

Weird...
 
Do tell, I use HDMI in general due to using TVs as a display but this one would definitely have me worried.

That file I brought up earlier, atikmdag_dce.log would be quite useful to know if that's precisely the problems you ran into, as it has the link speed negotiation history logged in there.

Weird...
Sadly it didn't get created on the old installation, and I managed to get it working before we got there.

Considering how many people vouches for the weird fix, it seems rather plausible. I did power-cycle the monitor unintentionally (disconnected the wrong cable, it fell down to the floor and then I had to fish for it on the back of my desk :p, then had to rearrange them around. Monitor spent all in all around 3 minutes disconnected from power) when reconnecting the cables, so it could've been.
 
Nice! I'm glad you got it sorted out. That's a pretty good question, too! Enjoy, my friend. :lovetpu:
So it happened again. This time I went straight to power off the monitor, plug off the cable, wait 3 minutes and plug it in again. Voilá, instantly works again.

I hate hardware sometimes, I love it another times. This time it skews towards hate :p
 
Last edited:
Back
Top