I built a new PC about a week ago with an MSI PRO X870E-P WIFI motherboard, XFX Mercury RX 9070 XT OC Edition, Ryzen 7 7800X3D, Lian Li 850W Gold PSU, 32GB of G.Skill Flare X5 memory, and a Thermalright Peerless Vision AIO. Everything worked normally, including long gaming sessions at ultra settings, until the GPU began crashing while Apex Legends shaders were loading. Mortal Shell 2 is also affected.
The crashes report DXGI_ERROR_DEVICE_REMOVED and DXGI_ERROR_DEVICE_HUNG, while AMD Adrenalin reports driver timeouts. Windows Reliability Monitor shows repeated LiveKernelEvent 141, along with 117 and 193. Temperatures are normal, OCCT reports zero errors and zero VRAM errors, and there are no overclocks or undervolts enabled. I also disabled the 7800X3D's integrated graphics, set the PCIe GPU as primary, disabled hybrid graphics, updated the motherboard BIOS, and reseated and checked all power connections.
The graphics card uses three separate PCIe power cables rather than a daisy-chained cable. One unusual symptom is that the AIO's LCD briefly resets to its startup screen whenever the GPU or driver crashes. Could this be a faulty driver, graphics card, PSU or power-delivery problem, motherboard or PCIe instability, or a USB-related issue? What should I test next?
2 Answers
A clean driver reinstall and testing an older Adrenalin release would be a sensible first step. Use DDU from Safe Mode, install the driver without importing old tuning profiles, and leave the card completely stock during testing. Some recent driver branches appear to be causing more timeout problems on certain 9070 XT systems, while 26.5.2 has been stable for several users.
Turning off Adrenalin's issue-detection feature might prevent the notification, but it will not fix the underlying device hang. I tried that approach and the games still crashed, so treat it as a diagnostic setting rather than a solution. If the older clean driver does not help, test with other overlays and monitoring tools disabled, then try a different PCIe slot or another known-good PSU/GPU if available.
Good to know—the crash continuing after issue detection is disabled means the alert itself probably isn't the cause. I'll focus on isolating the driver and hardware next.
That gives me a clear starting point. I'll remove the current driver with DDU and test 26.5.2 before changing anything else.