My gaming PC randomly blackscreens and powers off during games such as CS2, Split Fiction, and Valheim. The system becomes completely unresponsive, and I have to switch the PSU off and back on before it will start again. The fans and lighting remain on during the failure, but there are no obvious crash logs or error messages.
My hardware is a Ryzen 5 7600X, GeForce RTX 3080, Gigabyte B650 Eagle AX motherboard, 32GB of DDR5 memory, and a Corsair 850W Gold-rated PSU. CPU temperatures are typically 55–85°C while gaming, and the GPU stays around 60–70°C. I have tried factory settings, Eco Mode, and negative CPU offsets, and the RAM is capable of running at 6000 MT/s.
Individual CPU and GPU stress tests, along with an OCCT power test, ran for at least 30 minutes without errors or shutdowns. The CPU reached about 95°C and 120–130W under full load, while the GPU reached about 70°C and drew roughly 339W in FurMark2. I normally use three monitors, although only two are active while gaming, with Discord on the second display. Regular peripherals and a desk lamp are also connected.
I suspect the PSU or a protection circuit is tripping because the computer needs a full power cycle before it can restart. What could be causing this, and what are the best ways to test it?
2 Answers
The PSU, motherboard, CPU, or a short could be triggering a protective shutdown. Since the system will not restart until the PSU is switched off and back on, a PSU protection circuit is a strong possibility. Inspect the motherboard, GPU power connectors, and both ends of every modular cable for discoloration, melted plastic, loose pins, or other damage. If everything looks normal, testing with another known-good PSU is the next practical step.
An RTX 3080 can produce very brief power spikes during real games that steady stress tests may not reproduce. The average load can look safe while a transient spike trips the PSU's over-current or over-voltage protection. That would explain why FurMark and OCCT pass but games still cause an immediate shutdown.
Check Event Viewer under Windows Logs > System for Kernel-Power Event ID 41 at the crash time. It does not identify the failed component by itself, but an Event 41 with no preceding driver or hardware error supports the idea of a hard power loss. The best confirmation is to run the same games with a known-good PSU and properly connected GPU power cables.
I found several Kernel-Power 41 entries, with no other errors immediately before them. I also tried an RTX 1060 and the shutdowns did not happen, although it uses far less power. I have a spare 700W PSU—would that be suitable for testing?
A decent 700W unit can be useful as a short diagnostic test if it has the correct PCIe connectors and is in good condition, but it may be marginal for a 3080 depending on the model and its transient handling. Do not mix modular cables between PSUs. If possible, use a known-good quality 750–850W unit for the final test, and avoid running the 3080 through questionable adapters or a single daisy-chained cable.
I checked the board, connectors, and cables and could not find any damage or discoloration. I am going to swap in another PSU and see whether the shutdowns stop.