[mythtv-users] Whats the most stable system at the monent?

[Eric S]

I'm goin going to chime in hear and talk about some of the more
obscure things and agree with Kyle on gcc compilers,  gcc4+ is still
pretty cutting edge.

First of all, if you don't leave your Myth box on 24/7/365 then you
cannot even begin to comment on stability, those that do should be
measuring stability and uptime of computer and backend in weeks or
months, not days.  If you are in the days category take a hard look at
your system and make some changes.

Some other items which are often overlooked:

Clean, constant power . . . . put the backend on a UPS, even if its
just a small one to even out local power flux.   Also be sure your
power supply is up to the task, multiple tuners and 3 to 4 7200rpm
hard drives, and modern CPUs can pull some serious current.

Cooling . . . .  computers and components running excessively warm are
more prone to errors which will eventually lead to stability issues.  
One of the reasons I like my remote backend is that I can "over-cool"
it with 4 case fans blowing across harddrives and the motherboard and
not be concerned about the fan noise tucked away in its corner of the
basement.

Memory - don't discount possible memory problems if you are having
stabilty issues, run MEMTEST86 . . . .

Motherboard Chipsets (and implementation) . . . . Value chipsets can
be great, some are better than others, but in my personal experience I
have found that real Intel chipsets (845/865/older 810) seem to offer
a higher level of overall stability than others.   Chipsets that use
older generic driver modules often seems to have more bugs eliminated
as well.    Many people seems to have decent stabiltiy with nForce
chipsets, but I would suspect anyone running VIA or SiS Chipsets would
see a stability improvement with a move to a Intel or nForce chipset
motherboard.

My $0.02, YMMV, 
-Eric



On Fri, 21 Jan 2005 11:31:57 -0500, Kyle Rose <krose+mythtv at krose.org> wrote:
> 
> I disagree.  One big variable is the compiler: gcc-4.0, for instance,
> is still pretty experimental and still often produces buggy code.  I
> would recommend avoiding any distros (e.g., the Debian gcc4 AMD64
> repository) that compile packages with gcc-4.0, because who knows what
> kind of bug it might introduce into (say) libc.
> 
> Cheers,
> Kyle