Re: [gentoo-user] [Maybe OT]: Instability of system

[Mick]

On Monday, 21 May 2018 01:20:21 BST Dale wrote:
> Adam Carter wrote:
> > On Mon, May 21, 2018 at 8:35 AM, Dale <rdalek1967@×××××.com
> >
> > <mailto:rdalek1967@×××××.com>> wrote:
> >     Alan Mackenzie wrote:
> >     > Hello, Gentoo.
> >     >
> >     > I'm having problems with my machine hanging or rebooting
> >
> >     spontaneously.
> >
> >     > It's doing this, perhaps, every three or four weeks.  I think
> >
> >     that when
> >
> >     > I'm in X, the system usually reboots, when I'm on a tty, it hangs.
> >     >
> >     > This phenomenon has, up till now, been just below the level at which
> >     > it's annoying enough to do something about.  But my machine just
> >
> >     hung on
> >
> >     > me a few minutes ago, and now it's definitely reached tha threshold
> >     > where spending effort fixing it seems justified.
> >     >
> >     > My actual Gentoo installation is fine, in fact, so good that
> >
> >     I've not
> >
> >     > needed to post to the list for a long time.  :-)
> >     >
> >     > My system is an AMD Ryzen processor on an Asus Prime X370-Pro
> >
> >     mainboard,
> >
> >     > and is just over a year old.  I don't think my RAM is unstable
> >
> >     (though
> >
> >     > it's been a long time since I've run that RAM checking program).
> >
> > I ran memory checking overnight on my unstable (segfaults) AMD 8350
> > system, but no issues were found. Underclocking the RAM to the next
> > lowest speed completely addressed the issue. If i get keen i may
> > re-visit the RAM timing to see if it can be made to run stable at the
> > nominal frequency with more conservative settings on the other parameters.
> >
> > If the firmware hasnt fixed it, the underclocking is cheap test, but
> > yeah power supplies seem to be problematic. I always buy branded ones,
> > but even then only had mediocre results.
>
> In the past, I've had bad ram test OK with those tests.  When those
> tests say ram is bad, it seems to always be accurate but sometimes it
> doesn't catch a bad stick.  I don't know if it doesn't test the whole
> thing or what.  I read somewhere that it sets aside a certain amount of
> ram for the test program itself, after all it has to be in memory to
> run, so it doesn't test that part.  So, even tho it says they are good,
> it may not be certain.  It would make me consider other causes tho.
> I've only had that happen a couple times.
>
> I'd run at the default settings first.  If it still does it, then you
> can go back to your custom settings after you fix the problem.  I
> thought about overclocking at one time but for what I do, it doesn't
> really matter much.  My puter spends most of its time waiting on me not
> the other way around.  I guess if someone has their system doing some
> number crunching, foldingathome or something then it may matter more.
>
> I have Thermaltake power supplies here.  It's not the best out there for
> sure but it is in spec and reasonable price wise.  Some of those el
> cheapos are cheap for a reason.  I bought a case once that came with
> one.  I only paid like $50.00 for the whole thing.  Given a bare case
> that was close usually costs around $40.00 at the time, they couldn't
> have spent much on the P/S.  I removed it first thing and put in a
> proper P/S.  The el cheapo made a good paper weight tho.  The best
> insurance tho is one that someone has tested.  You can still get one bad
> out of the box but it gives you better odds for sure.
>
> Dale
>
> :-)  :-)

I've had firmware updates causing problems like this (on Asus) only to be 
corrected with the next firmware update.  If with default settings your RAM is 
playing up, update the firmware and check again.  I tend to knock back the 
DRAM timings a bit on O/C'ed boards and this helps with stability.  It goes 
without saying if default settings give you an unstable system, you should not 
try overclocking it.  :-)

PS. I use Corsair PSUs and always spend more on them to make sure it is not 
the cheapest model.  A middle of the range modular unit comes with Japanese 
capacitors and has not caused me problems in various builds.  In cheaper 
models I've ended up replacing the capacitors without waiting for the PSU to 
fail first - cheap PSUs always fail on me at the end.
-- 
Regards,
Mick