I bought a second-hand PC that was built in January and has mostly been running well, but it randomly crashes during longer gaming sessions, especially in Overwatch. It can happen after 15 minutes, after a couple of hours, or not at all. The game freezes, the display goes black, and I hear a short buzzing sound before the system becomes unresponsive. Usually I have to hold the power button to restart it, and there is no blue screen or useful crash report afterward.
I have clean-installed the graphics drivers, updated the BIOS, disabled overclocking, and checked temperatures; everything appears to be within a normal range. Event Viewer only shows that Windows shut down unexpectedly. I did have two brief driver-timeout crashes while using the desktop, but those have not happened again.
The system has a Ryzen 5 5500, Gigabyte B550 Gaming X V2 motherboard, Gigabyte RX 7600 8GB, 16GB of G.Skill 3600MHz memory, an MSI MAG 240R cooler, a DeepCool 700W power supply, and both a 500GB WD drive and a 1TB Samsung 980. I paid €700 and would prefer to diagnose and repair it rather than return it. Could this be a power supply, graphics card, memory, or voltage-spike issue?
3 Answers
A hard lock with a black screen and no useful Windows log often points to something below the operating-system level, with the power supply, GPU, or cabling being the main suspects. Start with the simple checks: reseat the motherboard and GPU power connectors, check the wall cable and outlet, and remove any sleeved extension cables or adapters temporarily. If possible, testing with a known-good PSU would be one of the most useful ways to narrow it down.
The occasional driver timeout makes the graphics card worth investigating too. Make sure it is firmly seated and that its power connector is fully inserted. A short stress test passing does not completely rule out the GPU or PSU, since games can create different transient loads than synthetic tests. Try another demanding game or a longer OCCT test, and see whether lowering the GPU power limit or clock slightly changes the behavior; if that helps, it strengthens the case for a GPU or power-delivery problem.
Do not completely dismiss the memory. Run a proper memory test such as MemTest86, and also try temporarily disabling the memory's XMP/DOCP profile so the RAM runs at its default speed. If the crashes stop, the 3600MHz setting may be unstable even though the modules themselves are not defective. You can also test one memory stick at a time in the recommended motherboard slot.
I ran the CPU and GPU at full load in OCCT for about 10 minutes without a crash. I was wondering whether voltage spikes could still happen only in games, since I do not know enough about hardware testing to check that directly.