[SOLVED] Graphics card restarting in loop after monitor turns off

[gnn]

Hi,

I have an issue where my graphics card restarts in loop immediately after my monitor turns off. If I move the mouse, or hit a key, my monitor comes back on with no issue. It is very annoying ! Can anyone help me fix this issue....

My computer is mainly used as desktop and workstation, and some light gaming.
The following are my build specs:

• MSI MPG X570 Gaming Pro Carbon Wifi
• Ryzen 9 3900x, OC'ed (43.5 x6/ 41.5x6)
• Patriot Viper Steel 32GB (4x8GB) PC4-35200 (4400 MHz - CL19) Running at 3600 CL14
• Sapphire RX 570 Nitro+ 4 GB (in PCIEx16_1 in Gen 3x16 mode) Not OC'ed
• Rosewill Photon 850W 80 Plus Gold PSU
• Samsung NVME 980 Pro in M.2_1 as boot drive
• Samsung NVME 970 Evo in M.2_2 as data drive
• Ziyituof ZYT-SA3006 ASMedia Sata 3 Expansion card (in PCEx16_2 in Gen 2x2 mode)
• Computer connected to the following display devices (in clone mode)
  • ASUS VE278 - 27 in monitor in DP mode
  • Samsung TV in HDMI mode
  • Panasonic TV in HDMI mode

Here are a few things I have tried:

• Checked the power profile
  • Running on AMD Ryze High Performance settings
  • PCI Express Link State Power Management is OFF
• Running Radeon Software Only
  • The graphics card does the cycle boot, fans ramp up to idle speed.
• Running MSI Afterburner
  • If MSI Afterburner is running, the graphics card does the same cycle boots, but the fans cycle up to 100% upon each restart (really annoying)

 

[InvalidError]

If the monitor output comes back as soon as you move the mouse, then it sounds more like the video outputs have simply gone in power-saving state after the computer has been idle for however many minutes your power plan profile is turning off monitors at. In that case, fans ramping up/down periodically might indicate that something is periodically putting heavy load on the GPU.

If the GPU is really 'restarting' then I'd expect the Windows log to include a mention that drivers have been reloaded due to a GPU or driver issue. In that case, try the GPU in a different system to rule out unrelated hardware and software issues.

You could also try unplugging your TVs to see if the undesirable monitor standby behavior might be triggered by them.

 

[gnn]

Thanks for the quick response.

I believe my GPU is really restarting, as the Sapphire logo goes dark and fans stops, then comes back on blue and the fans ramp up again.

I cannot seem to locate any windows log mentioning the graphics driver. Where would those be located in the Event Viewer. I would have expected Hardware Events, but it indicates 0 events.

Nevertheless, I tried to unplug both TV's and the problem seems to have went away. When I plugged the TV's back in, the problem did not come back. This seems to indicate that the HDMI talk or CEC might be what is causing the issue. I will continue to do some testing and repost my findings tomorrow evening.

If that fails to lead to a conclusion, I do have an old Core 2 - Q9650 system laying around I can put the card in to see if the issue comes up.

 

[InvalidError]

Driver/kernel issues are usually in Windows Logs\System

 

[gnn]

I get this log that pops up, but it seems unrelated.

Log Name: System
Source: Service Control Manager
Date: 2021-04-08 1:07:03 AM
Event ID: 7040
Task Category: None
Level: Information
Keywords: Classic
User: SYSTEM
Computer: darica
Description:
The start type of the Background Intelligent Transfer Service service was changed from demand start to auto start.
Event Xml:
<Event xmlns="http://schemas.microsoft.com/win/2004/08/events/event">
<System>
...
</EventData>
</Event>

Also, I can confirm that the HDMI connections to the TV's have no effect on the issue. The problem persists when I only have the DP on the monitor. I also tried HDMI on the monitor, same problem.

I will try the card in a different PC tomorrow to isolate driver issues. I could also try to load the Windows default drive to see what happens.

 

[InvalidError]

The system log likely has hundreds of entries, BITS is just a background thing. You need to look for GPU-specific ones or at least ones that involve driver files with names that look GPU/graphics-related.

 

[dorsai]

Check if you have Wake On USB turned on in the BIOS.

Also make sure you have Windows set to Shut Down and not Standby when you power down.

 

[gnn]

OK, so I finally figured it out. It was the driver update that caused the issue.

The card never actually restarted, this was hinted by the fact that there are no log entries that could be found in the Event Viewer. It seems that the card only looped off and on for the LED logo display on the side of the card, and the fan controller. This is why I could move the mouse and it would resume immediately. I found this out by moving the mouse while on idle and immediately after the logo turned off. The monitor displayed for a while and I had control over the computer, but the fans and display were still off for a few seconds before they turned back on again. This hinted be towards a driver issue, as this was a really specific internal failure of the graphics card that could not be associated to a video signal or a simple power failure. Since it only occurred after a monitor off event in windows, it was also unlikely to be a hardware issue. So I stated to investigate driver version to see if I could find anything.

Here is the impact of each driver trialed from most recent to oldest:
Base windows Catalyst driver - Solved issue, but no control over video parameters. No not really a solution. But it did point towards the possibility that the entire issue was driver related.
21.3.2 Non WHQL (29 Mar 2021) - Multiple crashes and PC reboot. Had to use safe mode to uninstall drivers.
21.3.1 WHQL (24 Mar 2021) - This was the driver I was using when I posted about this issue. Seems that the fan controller was restarting in loop when I was in idle mode.
20.11.2 WHQL (13 Nov 2020) - This driver worked well and solved the issue of the fan controller restarting. But I did get a DSOD. This got me thinking that since x-mas or so, I had unexplained random reboots a few days apart, which stopped happening late march. After a look at the logs, the time they stopped and the timeline corresponded to the 21.3.2 driver release. So I continued working my way back.
20.9.1 WHQL (9 Sep 2020) - This driver seems to be working well so far on all fronts. I will repost should an issue come up.

Next steps:
20.1.2 WHQL (23 Apr 2020) - Tried it briefly, seems to work. I will give it a longer run should I get more issues using 20.9.1.
Wait for another WHQL driver release and test to see if issue was resolved and ensure any crashes are absent.

Thanks for the help, while it did not give me an answer directly, it got me thinking harder about the issue and I was in a far better position to diagnose it by ordering my train of thought.