[Nottingham] Arch always on system freezing

Martin martin at ml1.co.uk
Wed Oct 31 20:45:48 UTC 2018


Brian,

That's a good test list and a good find!

Good you have it fixed with the OC-style voltage tweaking...


Aside: I've had CPU heatsinks that appeared to be 'clean enough' yet in
reality were choked with dust behind the fan or deeper in the cooling
fins. That has become more of a problem/feature with the later heatsinks
that have some overly fine pitch fins that appear to be below the
critical spacing that sees dust get jammed fast rather than blown through...

And beware the old degrading electrolytic capacitors that inexorably
degrade that then reduces the supply noise margin until you suffer
phantom random reboots... (More volts can be a temporary fix.)


Thanks for noting the fix.

Good luck!

Cheers,
Martin



On 27/10/18 16:30, Brian Pickford via Nottingham wrote:
> x= tested and ruled out
> x Dust
> x Failed cooling fan
> x Internal/CPU overheating (including the PSU or power adapter)
> x Try running Memtest86 for an hour or two...
> x Any disk errors still...
> x Full disk partition
> 1/2 x here Disk failure
> I ran mprime and had a failure:
> [Sat Oct 27 14:37:44 2018]
> FATAL ERROR: Rounding was 0.46875, expected less than 0.4
> Hardware failure detected, consult stress.txt file.
> Self-test 640K passed!
> Self-test 640K passed!
> Self-test 640K passed!
> Self-test 640K passed!
> Self-test 640K passed!
> Self-test 640K passed!
> Self-test 640K passed!
> [Sat Oct 27 14:45:25 2018]
> Self-test 8K passed!
> FATAL ERROR: Rounding was 0.46875, expected less than 0.4
> Hardware failure detected, consult stress.txt file.
> 
> the CPU temp climbed fro its usual 50C to 91C after 15 mins. This is
> quite high, but stable. The error occurred in the first 5 mins while
> still under 80C
> I've upped the North and South bridge voltages 2 notches along with V
> core then ran the memtest, no errors after an hour running at 77C
> 
> A short SMART self-test passed without error ... I'll see if stability
> improves with the slightly higher voltages
> 
> Thanks, Brian



More information about the Nottingham mailing list