It's very easy for minor memory issues to result in disk corruption rather than overall system instability. You just need something like a single bad bit that only shows up sometimes, eg while hot as nostr:nprofile1qqsr6tj32zrfn7v0pu4aheaytdnnc6rluepq73ndc2tdjzus34gat9qpz4mhxue69uhhyetvv9ujuerpd46hxtnfduhswulwwv pointed out.
Discussion
Hmm will run more tests to confirm
Write a memory test program that stores 4092 bytes and a CRC
memtest86 does this more thoroughly… ive always used that in the past
Memtest doesn’t tend to get your CPU hot, though. Different things can fail at different utilization levels…
There was that prime-something program i remember using a long time ago for that, not sure if there are more modern solutions
ycruncher?
I remember prime95 but maybe that was like 20 years ago
damn im old
whatever type of error you are having must be a more widespread issue than a few pages to be triggered so often
the ideal program would have a small pool it nonstop allocates to and deallocates from and a pool it very slowly checks and rotates allocations in/out
write random data and CRC it as I said
you then want to stop the process and get a debugger if there’s a mismatch and see the physical location along with identifying the RAM module
Prime95 will definitely get a CPU toasty, but you would need to ensure it's running with a mix of large and small FFT and for a while (e.g. hours/days). Small FFT maximizes CPU heat, but isn't the best for finding instabilities working with RAM. Large FFT helps there. In general, I find the overclocking community sometimes works on "vibes" rather than completely proven test methodologies, so take this with a grain of salt.
not that you're overclocking, but that group tends to accumulate tribal knowledge of CPU/RAM stability tests. Hardware can be "fun". A ton of variables. Even things like bios versions can cause instabilities.