For the past couple of months, my computer has been crashing unpredictably, especially while streaming with OBS or playing games. The screens either go completely black or freeze, although I can usually still hear audio. Sometimes the display comes back after a moment, but the game, OBS, and most other applications have already crashed. Discord and a few background programs may remain running.
A technician has replaced nearly everything except the power supply, SSD, and some cables. The crashes happen less often now, but they still occur during streaming and occasionally outside of it. I also had a fresh Windows 11 installation done during the hardware replacement.
The system uses an RTX 5080, Intel Core Ultra 9 285K, Gigabyte Z890 AORUS PRO ICE motherboard, 32GB of DDR5-6000 memory, an 850W power supply, two monitors, and OBS. The relevant errors are Event ID 153 involving nvlddmkm, a LiveKernelEvent with code 141, and an OBS NVENC error saying the device passed to the API is invalid. What should I investigate next?
3 Answers
The technician replaced a lot of hardware without eliminating the problem, so the next step should be controlled testing rather than more part swapping. Record whether the crash happens with OBS closed, with NVENC changed to software encoding, and with only one display connected. Compare those results with a GPU stress test and a memory test. That can help separate an OBS/NVENC driver problem from a power or graphics hardware failure.
The power supply is still a major suspect, especially since it is one of the few original parts left and the crashes happen under GPU load. An 850W unit may also be marginal depending on its model, age, cabling, and transient handling. Testing with a known-good, appropriately rated PSU would be more useful than continuing to replace unrelated components. The graphics driver should also be completely removed and reinstalled, since the nvlddmkm and NVENC errors point toward the display driver or GPU losing communication.
A clean Windows installation was already performed, so reinstalling Windows again may not solve this by itself. I would check the exact PSU model, reseat the GPU and power connectors, avoid adapters or loose extensions, and test with one monitor and hardware acceleration disabled in other apps. If the crash still produces LiveKernelEvent 141 with a known-good PSU and a clean driver installation, the GPU or its power delivery could still be defective.
Windows was reinstalled when the replacement parts went in, but I may still try another clean setup after checking the power supply and drivers. The PSU is one of the only original components left.