My PC suddenly powers off when launching or playing certain games, with no blue screen or error message. After rebooting, Windows only reports Kernel-Power 41 and Event 6008 for the unexpected shutdown. There is no bugcheck code, WHEA error, NVIDIA driver error, display-driver error, or minidump.
The system passes a 10-minute OCCT Power test, reaching about 380 W of GPU board power and roughly 514 W combined CPU and GPU power. Temperatures peaked at 77.8°C on the GPU core, 92.2°C on the hotspot, 86°C on GPU memory, and 97–97.5°C on the CPU package, with brief CPU throttling. The motherboard MOS/VRM sensor reached 100°C and reported VRM thermal alerts.
The system contains an Intel i5-14600K, MSI B760M Bomber WiFi motherboard, MSI RTX 3080 Gaming Z Trio, 32 GB of DDR5-5600, and an almost five-year-old ASUS ROG Thor 850W Platinum power supply. XMP is disabled, and there are no manual overclocks or CPU undervolts.
Reducing the GPU power limit to 70% did not prevent the shutdown. I also lowered the CPU PL1 and PL2 limits to 95 W, down from 135 W and 150 W, but the PC still shut down at the same point while loading a game. BIOS System Power Fault Protection is enabled. OCCT showed no PCIe errors, and its reported 12 V readings were generally normal, although one GPU connector briefly reached 11.790 V.
What could make the system shut down specifically during game launches or gameplay when it survives a sustained OCCT combined load?
4 Answers
It is still worth testing the memory even though a RAM problem often causes a blue screen. Run several passes of MemTest86 or another bootable memory test, and also check the system with one DIMM at a time. A marginal module or memory-controller issue can appear only under a particular game workload.
The power supply should still be investigated. It is nearly five years old, and an RTX 3080 can produce very short transient spikes that a synthetic test may not reproduce. Make sure each GPU power connector is fed by a separate PSU cable rather than using one daisy-chained cable. If possible, test with a known-good, modern high-quality PSU with sufficient transient headroom.
A graphics-driver issue is also worth ruling out. Completely remove the current NVIDIA driver with a clean-install utility, then test an older stable driver version. Game launches can exercise different driver and shader paths than OCCT, so passing the stress test does not fully clear the driver or GPU software stack.
The motherboard VRM is a strong suspect. OCCT may produce a steady load, while launching a game can create rapid CPU power transients and shader-compilation spikes. The VRM sensor already reached around 100°C and reported thermal warnings, so the board could be triggering protection even though the average OCCT power draw looks acceptable. Try setting the CPU limits substantially lower, such as 95 W or below, and see whether the behavior changes.
I tried setting both PL1 and PL2 to 95 W in BIOS. The shutdown still happened at the same point, so lowering the CPU limit did not appear to solve it.