I replaced my failed power supply a little over three weeks ago, and since then the PC has intermittently entered a strange partial lockup. New applications usually refuse to open, existing programs behave inconsistently, Task Manager and hardware-monitoring graphs stop updating, File Explorer and Windows Search may not respond, and many interface elements no longer react to mouse hovering. Some applications continue working partially: browser pages may still load slowly, Notepad++ can still save files, and games may keep running for a while even though input and saving stop working. After roughly two or three minutes, Windows forcibly restarts. Event Viewer and Reliability Monitor report a watchdog timeout, without providing a more specific cause.
The system contains a Ryzen 9 7950X3D, Gigabyte B650 Eagle motherboard, GeForce RTX 4080 Super, 64GB of DDR5 memory, a WD Black SN850X 4TB SSD, and a Corsair RM1000x PSU running Windows 10 Home. I replaced all PSU cables with the cables supplied with the replacement unit, reseated and swapped the RAM, and ran SFC, CHKDSK, and a seven-hour MemTest64 test without errors.
The problem has occurred while downloading a file, playing 30XX, playing Vintage Story, and backing up a USB drive while watching a video. It did not happen during several hours of playing another game immediately after the repair, but the failures now seem to occur more frequently. What is the best way to identify whether the replacement PSU, memory configuration, motherboard, storage, graphics card, or another component is responsible?
1 Answer
The first thing I would investigate is memory stability, including the memory speed and any automatic overclocking profile such as EXPO. DDR5 can pass a basic memory test and still fail under a particular workload, so temporarily loading BIOS defaults and running the RAM at conservative JEDEC settings would be a useful diagnostic step. If the problem disappears, test the sticks individually and then work back up to the rated settings.
I had assumed the seven-hour MemTest64 run ruled memory out, and I have not intentionally overclocked anything. I will check whether the motherboard enabled a memory profile automatically and try the default settings.