Question Solutions and component diagnosis to fix PC stability issues during gaming ?

[wortmalerei]

TLDR: My PC keeps crashin during games. Don't know if it is a GPU, CPU or RAM issue. I tried everything within my limited knowledge but to no avail

My PC:
CPU: AMD Ryzen 9 7900 (not overclocked)
CPU-Cooler: Bequiet! Dark Rock Pro 4
GPU: ASUS TUF 4090 OC (not overclocked)
MB: ASUS TUF GAMING X670E-PLUS WIFI
RAM: G.Skill Flare X5 32GB DDR5 Kit
PSU: 1200 Watt be quiet! Straight Power 11 Platinum Modular 80+
OS: Windows 11 Pro

Issue:
PC runs stable during tasks like browsing, word processing, and development. However, it crashes during gaming after 2-20 minutes. I've noticed that lowering the graphics settings allows for longer gameplay. The crashes mostly drop me back to Windows, occasionally resulting in different BSOD (e.g., String: IRQL_NOT_LESS_OR_EQUAL, Code: 0x0000000a, Driver: ntoskrnl.exe). Additionally, I've observed recurring errors in the Event Viewer during each restart or normal operation, although they don't manifest during regular use. Examples include:

• Error: Source: Firefox - The system cannot find the specified file.
• Error: Source: DistributedCOM - The server "{ }" failed to register within the specified time period with DCOM.
• Error: Source: ServiceControlManager - The "Windows Security Service" service terminated unexpectedly. This has occurred 1 time already. The following corrective action will be taken in 60000 milliseconds: Restart the service.

Here's what I've tried so far:

• Installed the latest drivers (including BIOS and GPU).
• Completely reinstalled Windows 11
• Set RAM to default settings from EXPO 1.
• Ran hardware tests:
  • RAM: Memtest86 - 6 passes -> no errors.
  • CPU: CineBench -> stable after multiple tests -> no errors.
  • SSD: Samsung Magic Health Check -> no errors.
  • GPU: FurMark -> stable ->I stopped it myself after approximately 40 minutes.
  • GPU: EzBenschmark -> Benchmark crashes at higher settings every 2 or 3 attempt
  • GPU: HeavenBenschmark4.0 crashes at higher settings every 2 or 3 attempt-> Exception code 0xc0000005.
  • Ran various recommended Windows scans (e.g., sfc /scannow) -> no errors.
  • Checked the temperature during stress test → All is fine

I've reached the limit of my knowledge... and it seems, my gaming career ☹️

What else can I do to fix these crashes?
How can I determine which component (CPU, GPU, RAM, power supply, Windows, etc.) is causing the crash?

Thanks for your answers

 

[wortmalerei]

Update that might help narrow down the problem:
I just tested the game Metro Exodus Enhance Edition (again) and this time limited the frame rate, reducing it from 144 to 60 using the NVIDIA Control Panel in Windows. The game ran without any issues for 1 hour (then I exited without experiencing any crash !). However, as soon as I set the frame rate back to 144, the game crashes within a maximum of 3-4 minutes. In 90% of cases, the game crashes directly to Windows, meaning the screen freezes, and I use the Task Manager to return to Windows. In 10% of cases, I receive a BSOD (Blue Screen of Death).As mentioned above, the system is otherwise stable (without heavy GPU usage).

Based on this, I suspect there might be a problem with the power supply or the GPU, am I correct?
Can someone provide me with a tip on how I can determine whether there is an issue with the GPU or the PSU (or another component)?

 

[wortmalerei]

I now found a more consistently reproducible way to crash the system, apart from GPU-intensive games. I purchased the 3DMark Advanced Edition. CPU and SSD tests are running smoothly. However, GPU tests crash intermittently, similar to games. When I run a test approximately 4-7 times, it crashes once, meaning the program closes and returns to Windows. After that, I can use Windows normally.

A more severe and "reliable" crash occurs when I perform the GPU test with DLSS ("NVIDIA DLSS feature test") using the setting: DLSS 3 Super Resolution - Quality. The test sequence initially runs without DLSS and then reloads to test with DLSS. The first test without DLSS runs stably (never experienced a crash), but during the second part with DLSS, the computer crashes consistently at a fixed point after about 20 seconds.

Sometimes the computer crashes back to Windows, and shortly thereafter, I receive a BSOD*, or the computer crashes directly to a black screen, where the computer is still on and remains on even after waiting for a while without any changes. The only option then is to press the reset button to completely restart the system.


*The BSOD error messages vary, such as DRIVER_IRQL_NOT_LESS_OR_EQUAL, CRITICAL_PROCESS_DIED, SYSTEM_SERVICE_EXCEPTION, but the cause is always "ntoskrnl.exe+42c280."

Can someone tell me what is special about DLSS? Is this an indication of a faulty PSU/GPU or other?"

 

[thestryker]

While I'd lean towards it being a faulty GPU at this point it wouldn't hurt to use HWMonitor and MSI Afterburner while you're doing anything that will cause a crash. These will allow you to see what the GPU and CPU are doing when the crash occurs. HWMonitor does have logging capability so that could capture all of the system when it crashes.

I have a feeling you're going to find out the only thing impacted is the GPU, but it's still good to verify. Generally speaking if the problem was CPU/storage/DRAM/motherboard/PSU you would see problems doing other things even if it wasn't as often.

 

[Misgar]

I experienced similar problems with Windows rebooting during long video rendering sessions running overnight earlier this year.

The problem was partly due to instability in Topaz Video AI which has improved considerably after numerous weekly updates.

A quick fix at the time was to set the maximum GPU power to 95% in MSI Afterburner. The GPU did not appear to be getting too hot when it was running at 100% and I'm not overclocking the GPU or the CPU.

CPU: 7950X
Noctua NH-D15
GPU: Gigabyte RTX 3060

I'm using the NVidia Studio drivers, not the Gaming drivers.

IRQL_NOT_LESS_OR_EQUAL faults are often due to driver problems but can be very difficult to fix.