I've spent about four months troubleshooting a PC that randomly goes to a black screen and immediately restarts. Event Viewer only shows Kernel-Power 41, with no blue screen or useful error message. I've tried multiple Windows installations, older and newer GPU drivers with DDU, different power settings, BIOS updates, and thorough cleaning and repasting. Everything is currently running at stock settings.
The crashes are highly game-dependent. The Last of Us Part II runs perfectly on Ultra, and the GPU passes OCCT and FurMark without crashing while staying around 45–55°C. However, PUBG restarts the system within a few minutes even at 900p on the lowest settings. GTA V with RageMP and mods usually crashes after around 40 minutes, while Hell Let Loose causes an immediate restart when opening the server browser.
The Radeon software previously pushed the RX 6650 XT to about 2735 MHz instead of the listed 2635 MHz. Limiting it to 2350 MHz at 1050 mV improves stability in some situations, but PUBG and Hell Let Loose still cause restarts. I also replaced the original power supply with a new MSI 650W unit, with no change. A technician checked the GPU, CPU, and RAM and found no obvious physical damage.
Could this be related to motherboard power delivery or transient loads from certain DX11 games, or is another component or configuration more likely to be faulty?
System: Ryzen 5 5600G, Sapphire Radeon RX 6650 XT Pulse 8GB, MSI A520M-A PRO, 16GB Patriot Viper 4 DDR4-3200 CL16, MSI MAG A650BN 650W PSU, 500GB NVMe SSD, and 2TB hard drive.
1 Answer
Since this is a hard restart rather than a normal application crash, the most useful next step is to check whether Windows is producing any dump files. Look in C:WindowsMinidump, and if files exist, copy and compress them for analysis. If the folder is empty, configure Windows to create small memory dumps and see whether a future crash generates one.
Also, if XMP is enabled, try testing with it fully disabled first. If the system is stable at the default memory speed but not with XMP, the RAM settings or memory-controller stability may be involved. You can also test the kit at its rated settings with a small DRAM-voltage adjustment, such as around 1.38–1.40 V, but change one setting at a time and verify stability rather than assuming the motherboard VRMs are the cause.
XMP is currently disabled, but I’ll test the memory settings and check whether Windows has created any dump files.