PC Graphics

[neodium]

Wanted to share a solution I found that resolved the black screen and driver timeouts for world of Warcraft. I’ve posted on the wow forums in a thread and a popular reddit thread as well as submitted feedback via the bug report tool to amd. But I’ve resolved my issues for now.

TLDR: I believe trace logs for the game were filling up to a max size and causing a driver crash. Disabling game logs in the adrenaline drivers and increasing the log size have resolved this issue for me on both dx 11 and dx 12. My post below, hope it helps someone else.

*******

I’ve been having this issue for awhile now after coming back to wow with a 7900xt installed.
I had tried both dx11 and dx12 as recommended but was experiencing the same black screen, hang, driver recovery process after an hour or so of gametime, maybe 2, and then every so often after that.

I noticed that in the windows event logs that the driver crash occured, and in each instance there was a event log 1 kernel warning within 20 or so seconds afterwards.

“The backing-file for the real-time session “RSXTraceSession” has reached its maximum size. As a result, new events will not be logged to this session until space becomes available. This error is often caused by starting a trace session in real-time mode without having any real-time consumers.”

I couldnt find anything really on this apart from some other comments in german on the web referencing issues with an AMD radeon GPU.

I attributed this to some sort of Radeon Software tracking log.

I applied 2 fixes.

1. I changed the log size under the kernal logs in event viewer from 1024kb and added an extra 1 to the front which auto rounded to some number close enough.**
2. This article explains a similar situation and the steps to increase size are here too “https://answers.microsoft.com/en-us/windows/forum/all/warning-kernel-eventtracing-1-logging/95f17ecb...

Probably main fix: I also opened the adrenaline driver software and disabled the following:

1. *Settings cogwheel:
   *Preferences tab:
   *Locate the “Game adjustment tracking and notifications” setting and change to disabled (default was enabled)

I have since now had a session over 6 hours long and not encountered any issues:
I am still currently on DX11 from previous testing but will swap back to DX12.

Following day on do 12 I’ve now had it running for 5 hours without a timeout.

My ‘theory’ is that whatever is being logged is getting a lot of info from wow in particular and thus causing a driver crash as the file hits max capacity. The delays in logging are likely due to the time to recovery and the system recovering and logging the error afterwards. 

Once again, Hope this works for others

[mr_feist]

Could you please explain the steps required to apply this fix again? I don't quite understand what I'm supposed to do. 

[LazerDefender]

It's in AMD Adrenalin settings cogwheel, under preferences. I've apparently had mine disabled and I'm still constantly crashing. I've tried low and high settings, reinstalling drivers, everything, it doesn't matter. I just swapped from NVIDIA, and I'm going to try to go back as soon as I can afford it.

[Darthsa]

I unfortunately cannot find these settings to disable. I've been having these crashes on most games with the same exact errors so I was hoping this was a solution

[TonyTheTerrible]

I'm also getting crashes like crazy. I'm turning off all In-Game Overlay related settings and I'll try DX11.

[neodium]

It’s potentially the event log size then. I’ll make something up when I’m home to demonstrate how to show the event logs. Displaying them in a way where we can see the events that occur around a driver crash and see if there are similar issues with others here. 

[kyberpunk]

Hey, I am also following this thread as I just built a new computer with 7800xt and experience the same problem with World of Warcraft crashing every few hours. I hope this might be the reason, but I am new to PC - So it could also be other reasons.

I am not quite sure which log file the in the Event Viewer I need to increase, so I hope you can provide some guidance here?

My event viewer is structured like this:
- Custom Views

- Windows Logs

- Applications and Services Logs

- Subscriptions

I tried to increase all the logs under Windows Logs (Application, Security, ...)
And found a log under Application and Services Logs / Microsoft / Windows / "Kernel-EventTracing" and increased this.

But from your post I can't confirm if this is the right logs to increase?

Your post and time is very much appreciated already!

[neodium]

Unsure if you get a notification, but i've replied directly to the main post with a guide.

[neodium · Accepted solution]

I haven't had issues since i applied my fixes and posted this solution.
Nor have I seen the problematic log since I changed it.

Step 1 of the kernel change:

Open the event Viewer, you can do that from hitting the windows key and typing "Event Viewer"

Step 2: Open the correct view, Check to see if we are seeing the same naughty log.
Select the Administrative Events view which should be under custom views.
This is a pretty good combination of reports from many different places, giving us a good timeline.

Step 3: Example of logs:
3a: Pair of logs:
Note the time the last time you had a driver crash when playing World of Warcraft and look for that rough time.
A good indication will be looking for logs that say "display".
Here is an example of the pair of logs i found that led me to make this change.

 

3b: Display log (4101) indicating driver crash and recovery aka "black screen"Display Warning

3c: Mystery "Kernel-EventTracing" (1) log referencing "RSXTraceSession"

The Kernel log continuously popped up after a driver crash while playing this game and as a result i saw this log as part of a pair each time I saw a driver crash log. So i made the assumption they are connected.

Solution starts here:

Step 4: Increasing the size of the event tracing log.
Navigate in the Event Viewer, using the left hand panel:
Applications and Services Logs
--->Microsoft

   --->Windows

      --->Kernel-EventTracing

Step 5: Open the properties of the Admin log under Kernel-EventTracing by right clicking it and selecting "properties" from the menu

 

Step 6: Modify the size
Default size will be 1024KB

I changed it to 11008KB which is just the nearest to 1000 that is a multiple of 64.
If you change it to 1000KB and select apply, it will warn you and adjust it to the closest multiple of 64kb.

-Change the Maximum Log size ( KB to 11008
Select "OK" at the bottom of the window to apply this change.

The rest of the settings on this window are default, but it should be set to overwrite older events as needed.
This fix was gleaned as it is the solution to address other log issues that have file size issues, is it perfect? no. But it worked for me.

This completes the fix: Bonus are my adrenaline settings which i referenced as the other half of my fix, as i changed both at once.

Radeon Settings

Hope this helps and resolves some issues, if it does or doesn't, please let the thread know.

[mr_feist]

Unfortunately this didn't work for me at all. Crashed three times back to back after doing this. 

[neodium]

when you viewed your event logs and located a display driver crash, was there anything of interest around the same time occurring?

[kyberpunk]

Thank you so much of the guide! I have done the above but unfortunately I still experience crashes.

I have taken some screenshots of the Event Viewer of the logs.

This is the overview, the crash happened around 09.51, and before this it was 10 minutes since last log.

This is the overview:

This is the warning log:


And this is the Kernel-EventTracing error log:

Display Warning: Display driver amduw23g stopped responding and has successfully recovered.

Kernel-EventTracing Error: Error setting traits on Provider {8444a4fb-d8d3-4f38-84f8-89960a1ef12f}. Error: 0xC0000001

All of the DeviceSetupManager looks like this:

Metadata staging failed, result=0x80070490 for container '{20B9CDE5-7039-E011-A935-0002A5D5C51B}'

Metadata staging failed, result=0x80070490 for container '{3C9DD332-6689-11EF-8C5D-806E6F6E6963}'

... Let me know if these are of interest.

Also let me know if you need more information? I am also trying to investigate on the error codes.

[neodium]

Well there definitely seems to be something going on there. 

A lot of my searches on the other errors you have in the log that occur near simultaneously with the driver crash have been frequently associated with USB devices, Bluetooth devices or problems with drivers.

When was the last time you updated your chipset drivers from your motherboards website? Try this as well.

Are you using any older usb equipment and can you reduce this down to the bare minimum or use other peripherals for awhile to see if the issue persists?

[kyberpunk]

Yes a lot of things going on.

I tried to also follow the suggested things in this post with the intention to solve the other errors: https://forums.tomshardware.com/threads/graphics-issues-after-moving-pc.3845648/

It has been crashing everyday.

Just now I get these logs:

1. Warning (Display)
   • Display driver amduw23g stopped responding and has successfully recovered.
2. Warning (Kernel-Eventracing)
   • The maximum file size for session "PerfDiag Logger" has been reached. As a result, events might be lost (not logged) to file "C:\Windows\system32\WDI\LogFiles\ShutdownPerfDiagLogger.etl". The maximum files size is currently set to 20971520 bytes.
3. Error (Kernel-Eventracing)
   • Session "PerfDiag Logger" stopped due to the following error: 0xC0000188

[kyberpunk]

I have also looked into this and tried to increase the log size with the Registry Editor - But it seems like I'm not authorize.

Following this thread: https://www.tenforums.com/general-support/137114-session-perfdiag-logger-failed-start-error-0xc00000...

[AverageWowPlayer]

Did the fix as per your instructions and played on DX12 no issues so far for ~8 hours. So fingers crossed but that long on DX12 normally would result in at least 6 crashes for me.

Prior to doing your fix I had tried many other fixes and only one that seemed to work for me was using DX11 and having anti-aliasing disabled which gave me at least 2 weeks of no crashes.

Does the size of the log matter in relation to time played per session? As in if I were to leave my game running and playing over the course of the day, would I need to increase the file size? How long does it log info onto the log before creating a new one?

[neodium]

Glad that it’s confirmed to have helped someone. I’m not sure exactly, but it was definitely a ~10 fold increase in size.  I’d have a crash after an hour or two.
So id assume that if the problem is consistent, 10+ hours before it popped back up.

I was hoping also disabling the adrenaline setting that tracks game data may also help reduce the amount of info going into this log.

i wish I could locate the log as I’d love to see what it’s filling up with. 

Definitely feels like an oversight in the driver design if an embedded log file maxing out was not accounted for causing an issue.

[Veavx]

Hi,
Ive tried your guide, sadly does not work for me. its really hard and enoying having this issue since buying the new pc (7900xtx + 7800x3d), Im looking constantly for solutions and reading a ton of forums posts. 

On WoW forums ppl sharing their problems since november '23 . Almost 1 year later and no solution. Feelsbadman

[AverageWowPlayer]

Although Neodium's fix has worked for me, another thing that worked was disabling anti-aliasing and v-sync while running the game in DX11. I would try give all 3 of those fixes a go. I tried just DX11 and it reduced the frequency of crashes. Then also with v-sync off which lessened them again to every few days. Turning off anti-aliasing as well completely stopped the crashes for weeks until I tried Neodium's fix.

I still had/have every other setting maxed out running at 100-165 fps at 1440p running on:

i7-13700k

7900xtx

32Gb RAM

[AverageWowPlayer]

Edited out double post

[neodium]

Yep, I actually had a windows update today and ended up having major issues with display as a result. 

in the end I ended up safe moding with network connections.

installing and running ddu to remove my drivers entirely.

Rebooting and updating to the recent amd drivers that I was holding off on.

with the log file size change in place I raided tonight over 4 hours and had no issues.

For those interested I did also locate and find a way to stop the RSXTraceSession. When you run “perfmon” from start and launch the performance monitor you can check under ‘data collection sets’ and then ‘event trace sessions’ to actually see the RSXtraceSession running with a play symbol next to it. But you will only see it if wow is running, if it isn’t, launch wow and refresh the perfmon.  You can actually just right click it and select “stop”. From looking at the logs and diagnostics, this actually seems to be a log generated with a open source proper diagnostic software tool that is used by amd in this case, they have likely borked something when using it. 

[AverageWowPlayer]

Ok so it seems for me the fix didn't work, just crashed for the first time in weeks in a warsong gulch bg. Only things I changed yesterday was turning anti-aliasing back on and also setting resample quality to FidelityFX Super Resolution 1.0 and increasing the sharpening slider to 0.5. I set image-based techniques to CMAA 2 and multisample techniques to colour/depth 4x. I might try a combination of those 2 over the next few days to see if I can get some sort of AA working.