I think, hopefully, I might have found the source of all my troubles.
I played Sauerbraten today to test whether the freezes were caused by my CPU or GPU overheating. Sauerbraten ran fine, so I decided to do a more scientific test by using Phoronix Test Suite. When looking at the tests available, I thought “Hmm, I’ll try this CPU stress one”.
It went for an hour without incident. So I decided to do “memory-bandwidth”. No problem. So I decided to do “iozone” (I’d noticed that the crashes only occur when there is data being written to disk). Anyway, I ran the first test and the computer crashed. I rebooted and ran it again, and it crashed again. This time I got a dmesg log - the actual crashing process was “kswap”. I rebooted and ran the test yet again, and it crashed yet again.
So I rebooted and turned off the swap. The iozone test didn’t crash the computer nor make it unstable. I’m now encoding videos, with the only notable incident so far being a single program crash (Prism); probably unrelated.
So, it looks like the problem is corruption of the swap. I don’t know if it’s a software problem or a hardware problem (bad sectors?), but so far turning off the swap has stopped the crashes. We’ll see what happens next morning though! 
Entries (RSS)