PC Drivers & Software

[Deadchamber]

Got a 7900XT a few weeks ago and ever since, I've been getting driver timeouts that often times (not every time, but often enough) result in my PC freezing. Waiting around for it to become responsive again is useless as it doesn't come back unless I manually restart the PC.  

Checking Windows Reliability showcases Hardware error & LiveKernalEvent 141 which I am now speculating is the error received due to a loss of power, which could be due to forcing my PC to shut down due to the freeze and not exactly the cause of the problem. 

I've done many troubleshooting steps like..

•uninstalling Drivers using DDU 

•Reinstalling Drivers 

•Reinstall Windows 

•Reinsert GPU 

•Replace PSU & RAM Sticks 

•Disable DOCP/Resize BAR 

•CMOS reset 

•BIOS Default settings 

•GPU Benchmarks & Stress Tests 

•track CPU & GPU temps (both below 70°c at all times)

•Brought my PC to my local Micro center and they Stress tested my PC for a couple days and found nothing. They couldn't get it to crash or freeze. (It has froze since getting it back from the micro center once.) 

My final solutions are to completely wipe my SSD, Reinstall Windows after said wipe, ensure my CPU and CPU fan are installed correctly.

and I've already filled out a form and waiting on a RMA number to send my GPU for RMA. 

I'm lost on what else to do so any tips or advice is appreciated. 

[radu1006]

The GPU is fine, the drivers are the problem. In all 2024 drivers they implemented a very aggressive power saving mechanism that cause the GPU to run at a very low voltage and it causes instability.

Solution 1:

Remove current driver and install 23.12.1.

Solution 2 (removes most drivers timeouts, but not all):

Go to Registry in this path:

Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Video\{4719DBA3-7925-11EF-93D5-806E6F6E6963}\0000

Disable ULPS by setting these 2 keys to 0.

Update 23 Sept:

Solution 3 ( testing in progress)

Activate Overlay from Adrenalin software on desktop and leave it open. This causes a very small load that increases the GPU voltage to around 600mV.

Right now I am using a very small configuration, so that it does not bother me if it is on all the time.

Performance > Metrics

Make sure that Game detection is disabled

[Truckovsky]

Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Control\Class\{4d36e968-e325-11ce-bfc1-08002be10318}\0000]

"EnableUlps"=dword:00000000
"EnableAspmL0s"=dword:00000000
"EnableAspmL1"=dword:00000000
"DisableSAMUPowerGating"=dword:00000001

************
Try not only ULPS but ASMP too.
Works for my Asrock 7900xtx

[visario]

And what video game do you run, that you got driver time out?

[radu1006]

From what I understand his issues are random, but maybe he can confirm.

I have the same "random" timeouts even when I have a browser open or just on the desktop left idle.

[theufan06]

I'm going through this really for months, I have a Sapphire 7900 XTX Nitro+.

I currently have a 9 year old AVR and thinking that it might be the cause. Lastly, I will suspect my FSP Hydro PTM pro 850W Platinum PSU which makes 0 sense because I don't experience that crash/driver timeout during gaming. I only experience this when idle or doing normal tasks.

I also tried to RMA my card twice and they found nothing.

Below so far are the troubleshooting methods I have tried and nothing worked:

• Re-seated GPU
• Currently using the GPU Support bracket provided
• Currently using 3 individual power cables from the PSU
• Up to 3x fresh installs of Windows (2x Windows 11, 1x Windows 10)
• Tried many, many different driver versions, using Display Driver Uninstaller each time – tried drivers from 23.11.1(November 2023 Drivers) to the most current Driver is 24.8.1(August 2024 Drivers)
• Disabled MPO
• Ran at stock GPU (advertised settings)
• Tried 3 different motherboard BIOS version for my motherboard
• Tried Minimal, Full and Driver only installation for Adrenalin
• Reset all settings in Mobo bios to default settings
• Swapped BIOS switches on the GPU itself
• Disabled Smart Access Memory
• Disabled Freesync
• Changed display cables for both HDMI and Displayport
• Ensured power mode set to balanced in Windows
• Tried Removing MSI Afterburner
• Checked event viewer after a hard crash, there is nothing there that would explain the crash
• Disabled fastboot in BIOS
• Disabled faststart in power options
• Cleared CMOS
• Disabled ULPS
• Disabled HAGS (Hardware Accelerated GPU Scheduler)
• Increased RAM Voltage
• Disabled PSS (amd cool n quiet) support in BIOS
• Disabled XMP
• Forced PCI-E Gen 3
• Forced PCI-E Gen 4

[veskojl]

I'm having a very similar issue. Running ASRock AMD Radeon™ RX 7900 XT 20GB on Windows 10 Pro with 750 W Seasonic PSU gave me a lot of headache with all (almost) 24.X.Y versions of the Adrenaline software. I've got constant driver timeouts, black screens and I had to manually restart my PC every day.

I decided to downgrade the driver to version 23.12.1 and I haven't experienced any problems so far (a few days).
If this doesn't solve my issue, I plan to buy a new PSU and give another chance to the latest AMD drivers. Until then, I'll keep my fingers crossed.

I don't know what the AMD software guys are doing, but without proper logging, error handling & QA procedures troubleshooting problems like this is a pure nightmare.
If they (AMD) don't have enough resources (money, people) then to stop wasting other peoples time and at least open source the driver, so the community can help themselves. 

[JozefTaktyka]

There is also one other solution which actually helped with my friend's 6600XT.

- Open the Adrenalin app and go to Performance -> Tuning,
- scroll to the GPU section and on the right side click on Custom under Manual tuning. A new section will appear below,
- enable GPU Tuning and set 'Max frequency (%)' to 90-95% (start with 95% and see if it worked).

Remember to save the changes in the top right corner of the app! Then check if it helped.

[veskojl]

Thanks for the suggestion, @JozefTaktyka !

I've got another timeout today, so I'll try @radu1006 's solution with the disabling of "EnableUlps" first and if it doesn't work I'll try with reducing the GPU Max frequency. 

This is getting ridiculous.

[radu1006]

Reducing GPU max frequency does not work, unless your problem is related to some PSU issue. But with 750W, you should not have any problems. The GPU does not reach the max frequency anyway most of the time. You can reduce it just for a lower power consumption.

Disable ULPS and if that does not fix it, enable the Overlay all the time on desktop.

How did the 23.12.1 drivers perform? Did you have any issues with them?

[veskojl]

"The GPU does not reach the max frequency anyway most of the time. You can reduce it just for a lower power consumption." 

Has anybody actually tried to overclock the GPU instead of reducing the performance?

"Disable ULPS and if that does not fix it, enable the Overlay all the time on desktop."

Thanks. I'll try this too.

"How did the 23.12.1 drivers perform? Did you have any issues with them?"

Initially my system felt much more stable. I was able to run it without any problems for several days. However...,

After one resume, the PC was idling for a while and I've got a driver error, I waited for about a minute and the system was able to recover itself from the problem. This is the first time this happens. With all 24.x.x. drivers I get black screen and I have to restart the PC.
I've checked the Event Viewer and noticed a Warning in the log: 

"Display driver amduw23g stopped responding and has successfully recovered."

So in that sense the 23.12.1 driver looks much more "reliable". 

Maybe it is worth mentioning, that I don't use the GPU for gaming. Most of the time it is loaded (VRAM: 100%, GPU: ~10%) by the hardware acceleration from Firefox (tons of open windows & tabs), and I often put the PC to sleep with the FF open. And usually this driver timeout happens after I wake up my computer. It looks like the GPU cannot keep up (voltage, VRAM, clock) with the fast resumes and switching between apps, Windows/load context or I don't know....

Having said that, both fixes:

1. Disable ULPS 

2. Activate Overlay from Adrenalin software on desktop and leave it open. This causes a very small load that increases the GPU voltage to around 600mV.

look quite logical from my perspective, but the solution is not here yet.

[JozefTaktyka]

Hey, have you tried new drivers 24.9.1?

So far so good for my 7700.

[veskojl]

Hi Jozef, 

No, I haven't tried it yet, but I'm planning to.

Just waiting for the next crash of the 23.12.1 driver and once I'm fed up with it, I will install the latest one.

[radu1006]

I have already tested 24.9.1 and it has the same low voltage problem and driver timeout on my 7900xt.

However  the best solution so far is what I have described as Solution 3 in my previous post:

1. Install latest driver (24.9.1 in this case)

2. In Adrenalin go to Performance > Metrics and enable Metrics Overlay

3. Set Transparency to 100% and set Adrenalin to start at windows login.

Works like a charm and because of the 100% transparency, nothing is displayed.

[veskojl]

I've tried installing 24.9.1 and applied all your recommendations. This didn't help.

Yesterday, after I woke up my PC, it run without issues for about an hour and then it hanged & froze, so I had to restart it.

I've checked the Event Viewer and there was present the typical log entry:

Display driver amduw23g stopped responding and has successfully recovered.

The new thing is the following 2 lines in the log:

1. Information

Process C:\Windows\System32\DriverStore\FileRepository\u0407465.inf_amd64_25af0ae2eb807985\B407987\atieclxx.exe (process ID:9368) reset policy scheme from {9935e61f-1661-40c5-ae2f-8495027d5d5d} to {9935e61f-1661-40c5-ae2f-8495027d5d5d}

2. Error

The AMD External Events Utility service terminated unexpectedly. It has done this 1 time(s).

In that sense, the 24.9.1 driver doesn't fix the PC freezing issue for me.

[JunoSword]

Hello all,

seems like i am not the only one with those problems...

Occurence:

• while idle
• while ingame
• while browsing
• while using discord or spotify
• while watching videos

There is no pattern identifiable when the driver crashes and i am not able to reproduce the issue. Browsing, idling or using different apps like discord etc. are not represented in every crash occuring, meaning that it happends without using one or another app or activity.

Windows Event Viewer Output:

Protokollname: System
Quelle: Display
Datum: 24.09.2024 23:13:09
Ereignis-ID: 4101
Aufgabenkategorie:Keine
Ebene: Warnung
Schlüsselwörter:Klassisch
Benutzer: Nicht zutreffend
Computer: DESKTOP-TSJT83G
Beschreibung:
Der Anzeigetreiber "amduw23g" reagiert nicht mehr und wurde wiederhergestellt.
Ereignis-XML:
<Event xmlns="http://schemas.microsoft.com/win/2004/08/events/event">
<System>
<Provider Name="Display" />
<EventID Qualifiers="0">4101</EventID>
<Version>0</Version>
<Level>3</Level>
<Task>0</Task>
<Opcode>0</Opcode>
<Keywords>0x80000000000000</Keywords>
<TimeCreated SystemTime="2024-09-24T21:13:09.4878050Z" />
<EventRecordID>20277</EventRecordID>
<Correlation />
<Execution ProcessID="0" ThreadID="0" />
<Channel>System</Channel>
<Computer>DESKTOP-TSJT83G</Computer>
<Security />
</System>
<EventData>
<Data>amduw23g</Data>
<Data>
</Data>
</EventData>
</Event>

I am sorry for it being in german, but most of you should have the same Event Output with the information that the "Display driver amduw23g stopped responding and has successfully recovered".
Which is not always the case, that sadly forces me to reset my pc manually.

Troubleshooting:

I've already tested Driver Versions 23.9.2, 24.7.1, 24.10.37.10 (recommended Driver for Warhammer 40k Space Marine 2), 24.2.1 (IMO the most stable with 2 occuring crashes in a 2 month time frame) and currently 24.1.1.
All deinstalled via DDU and reinstalled in safe mode.

Since User @theufan06 already tried a lot of methods, i can confirm that most of them have no effect against preventing a driver crash.

I can say with confident that this is only a software related problem, since i have not found any hardware related issues or OC problems. Checked with HWiNFO64, stress tested via AMD Adrenalin and FurMark.

No driver crashes occured during multipe stress tests with a max span of 10min.

There are no problems with my PSU or a issue with power. If 1000W with 20% Surplus do not suffice, i don't know what will. Also tried with a dedicated power outlet.

Driver Versions match and Windows interferance (Automatic Driver Updates) are disabled.

I was not able to fully reproduce the driver crash but experienced after 2 months while playing Space Marine 2 another driver crash. It was always around the same final boss fight state and were only able to fix it by deinstalling and reinstalling the same driver (Version 24.2.1).
The recommended driver version 24.10.37.10 on AMD's Website did not fix the occuring crash in the boss fight. Neither did Version 24.8.1.

My whole life i went with AMD and sadly my whole life i had to deal with some driver issues.
I tend to go to NVIDIA on my next GPU or mayber even sooner, if i can't fix this issue.
It breaks my heart since the RX7900XT has such a great performance and value, but always investing vast amounts of time to troubleshoot a terrible software is no solution for me anymore.

PC Specs (fresh PC Build from July):

• MSI Gaming Plus WIFI AMD X670E So.AM5 DDR5 ATX Retail
• AMD Ryzen 7 7800X3D 8x 4.20GHz So.AM5 WOF
• 20GB Sapphire Radeon RX 7900 XT Vapor-X OC Aktiv PCIe 4.0 x16 2xHDMI 2.1 / 2xDisplayPort 2.1 (Retail)
• 1000 Watt be quiet! Straight Power 12 Modular 80+ Platinum
• 32GB (2x 16GB) Patriot VIPER VENOM DDR5-6400 DIMM CL32-40-40-84 Dual Kit
• 2TB Samsung 990 PRO M.2 PCIe 4.0 3D-NAND TLC (MZ-V9P2T0BW)
• 1TB Samsung 980 Pro M.2 PCIe 4. 3D-NAND TLC (MZ-V8P1T0BW)

Current Driver (GPU1 AMD Radeon RX 7900 XT Primary/Diskret):

• 24.1.1
• Driverversion 23.40.02-240111a-400139C-AMD-Software-Adrenalin-Edition
• AMD Windows-Driverversion 31.0.24002.92

Current Driver (APU2 AMD Radeon(TM) Graphics Integrated)

• Driverversion 23.40.02-240111a-400139C-AMD-Software-Adrenalin-Edition
• AMD Windows-Driverversion 31.0.24002.92

[radu1006]

Ok, I have to clarify some aspects of my tested solutions.

There are different types of crashes/black screens, freezes:

1. Driver crashes when there is no/little load, idle or browsing.

2. Driver crashes during a high load situation: gaming, benchmarking, 3d rendering etc.

The 3 solutions provided by me only work for crashes of type 1 on 7900XT (possible all 7000 series).

For crashes of type 2, there has to be a separate investigation and root cause analysis. There can be a lot of reasons for them: overheating, insufficient power, GPU defects, bad driver implementation, other components failure etc.

Also if you are using a 5000 series AMD CPU, I recommend to increase the Minimum processor state in windows power plan to at least 20% (based on the silicon of the CPU, you may get crashes caused by the CPU in low power state)

[veskojl]

At first glance, my case falls in point 1). However, low GPU clock speed doesn't mean the GPU is not working hard. In my case, the VRAM of the GPU is at 100% for long periods of time. 

BTW,  my CPU is AMD Ryzen 9 3900X and this Windows setting is already at 100%.

For me, it looks like that the CPU/ GPU clock & VRAM/RAM get out of sync for some reason. 
It could be the GPU voltage, it could be the GPU clk, it could be the RAM, have no idea, but without low-level debugging this case looks unsolvable.

[alfata]

At this point im RMA-ing my second 7900XT for this reason.
In my talks with the retailer they said they've got insane number of issues with these cards.
They issue apparently is that the card disconnects. So if you are not straining it the pipeline is not as full and you can minimize crashes but as soon as you start straining it the disconnect cannot help but hit an unrecoverable number of frames so it just crashes.

[radu1006]

I have checked your posts, but you seem to have a lot of different issues that are not related to each other and also related to the PSU.

What do you mean that the GPU disconnects? From the socket or turns off?

This card can easily consume over 340W in some cases and depending on your config.

In your case I would first make sure that the PSU can handle it and also if it has 3 connectors, use 3 separate cables.

If you want to test it with lower power consumption, just limit it to -10 in power tuning in Adrenalin and see if you still have issues.

[InTheFlatField]

I'm also experiencing this with a 7900XT. There are several issues being discussed in this thread. The one I'm running into is that the drivers timeout when idling on the desktop. Once a game is running it is dead stable. I believe that some of the cases of the card crashing during gameplay could be power related. I ended up having to upgrade my PSU to make my 3070 stable.

However like others in this thread I am seeing issues at idle instead. I don't think they are power delivery related, and I'm not even sure that they are hardware level at all. I think there is some kind of bug at the driver level related to power saving techniques at idle.

Like others I started with ULPS which didn't fix it. I also tried disabling MPO and increasing the driver timeout and those didn't work either. The new to me suggestions in this thread are the ASPM settings, and keeping the overlay enabled all of the time (the overlay setting is super clever btw).

I'll give those ideas some soak time and report back, however I'm really responding in the hope that some developer sees this as another data point. I've also been reporting the crashes with the bug reporting tool. Again I think there is something funky going on with the idle power stuff at the driver level.

[veskojl]

@InTheFlatField , welcome to the club and good luck with troubleshooting this!
And please remember to share your findings here!

@radu1006 seems to be the most active & persistent guy on trying to resolve this issue, but unfortunately his tips didn't help in my case. I'm also trying to give my perspective on the problem, but no luck so far.
I've just woken up my PC and got another freeze and log entry in the Event Viewer:

Display driver amduw23g stopped responding and has successfully recovered.

Gratefully this time the driver recovered itself, so I didn't have to restart my system.

IMHO, this is somehow related to the VRAM (voltage/clock) of the GPU. I can't think of another logical reason why sleeping my PC with ~100% VRAM consumed and then waking it up would cause a crash of the entire system.