Are you certain that your system problems were caused by overheating? I'm not saying it wasn't running too hot, I'm sure it was, with clogged up fans, but 50C isn't especially high for a CPU. Also, after an OS refresh it ran OK.

Yeah I'm 99% sure it was overheating. Anyway it's running more stabler now. The psu is still problematic as a heat source. I tried blowing/sucking dust out of it while still mounted in case but may need to open it up, replace fan and completely remove dust.

I probably got the sequence of attempted fixes garbled, but the server was still falling over with the new os until additional cooling measures applied.

Forgot to mention one other thing I'd done was replace thermal compound to cpu/hsf.

Or not ... after about a day and a half of running ok, it went back to the falling over tricks. As a last hurrah I've booted it from systemrescuecd on a usb stick to see how long it stays up on that (monitoring via "watch -n 20 uptime" over ssh).

Maybe it'll do ok on minimal system load, or maybe another coincidence I'd forgotten was the first kernel upgrade on 18.04 LTS in what feels like a year, ~3-weeks ago. 20.04 may even share that kernel, or features therein too new for the 2005 dx5150.

Closer to throwing in the towel on that one.

Anyway, I have a cunning new plan: Swap in the ssd with installed system as a dual-boot to MrsD.'s other discard PC, the newer hp dx5750 (her latest is an i5 w/ 8G), currently hosting seldom-used Win 10 + Office 365.


 

The mystery deepens. Deep in OCD  territory now, for I did not abandon the terminally ill device, but have nursed it along whilst studying the underlying morbidity.

Apparently (apologies) this model (dx5150) is known for its bad caps that are made badder due to *overheating* in the SFF case. If true, the damage was (likely) done during the mad, overheat the whole fucking building episode and, by all accounts, unrecoverable, being I lack SMD soldering skills. Others have suggested replacing the psu, yeah but no.

Anyhoo, the server stays up just so long as I don't do anything with it except (for example) "watch -n 20 uptime" via ssh console. It's been up for ~4-hours today!!!  :-O~~~

When I go and use it for anything useful, such as mounting a share, it typically goes down within an hour or two, which is enough time to get stuff done, I suppose  :-S .

Haven't tried the swapping the new ssd into my other, newer (5750 IIRC) model old Mrs.D pc trick as yet, but it approaches.

Edit: Whelp, I guess it didn't like the report card ^^^ because yesterday it stayed up ~10-hours and I got some use out of it as a dev server before normal shutdown via console at 5pm.