[mythtv-users] Re: Building a myth box, wanna look over my components?

Mike Isely isely at isely.net
Wed Dec 15 22:25:19 UTC 2004


On Wed, 15 Dec 2004, Yan-Fa Li wrote:

> Mike Isely wrote:
> > I will do more testing with 2.6.9-ac14 and see if anything has improved.
> > It may take a while.  Even with these problems I can still get uptimes of
> > several weeks so it's pretty hard to get something statistically
> > significant without a lot of testing time.
> >
> > With all that said, clearly I am not going to recommend nforce2 boards so
> > long as I'm still having problems with mine.  But your mileage may vary...
> >
> >   -Mike
>
> Yikes.  You sound like you're having other problems.  All I can suggest
> is fundamentals.

Yeah, I know.  Been there, done that :-(


>
> The holy trinity of PCs: Power Supply/Power Source, Memory, Motherboard

First rule of hardware chasing: suspect the PSU.  I got my start far too
long ago hacking hardware.  Then I learned about software.  Dirty power is
a common cause.


>
> Assuming you have a half decent PSU, because AMD stuff requires min of
> 350W, you could look into wether you have brownout conditions due to
> overloading or perhaps dirty power due to some other appliance or a
> noisy circuit.  I've seen perfectly good hardware act completely wacko
> when the circuit was too heavily loaded.

In my case it's an Antec SLK-3700 case with its requisite Antec PSU.
It's a 350W Antec PSU, and Antec AFAIK is known for good PSUs.  The case
has additional front / rear 120mm fans; it runs very cool overall.  The
video card in this system is a Radeon 9200SE class board so it doesn't
need a lot of power - these days that's a big source of power overload
problems.  A wattmeter on the AC side shows the system pulls in about 100
watts nominally.  So I seriously doubt I'm overloading the PSU.  It could
be a bad PSU, though I find it unlikely.  But I do in fact have a spare
Antec 400W PSU sitting in my junk box so I'll add that to my list of
things to try.


>
> Memory falls to memtest86+.  This catches most, if not all, memory issues.

Been there, done that.  Several months ago I pounded on it for 30 hours
straight with memtest86 - not a single hiccup.  The memory timings are
"nominal", what they should be for the two 512MB PC3200 sticks I have
installed.

I've had previous experience with bad memory.  I find that a Linux system
with bad memory is lucky to survive more than a few hours.  This system
I've run for several weeks at a time - and that includes mass
recompilations of 2.6.x kernels.  So I'm fairly confident the memory is
OK.

Running memtest86 I don't think can test arbitration problems.  While
memtest86 is running, there's no I/O - no DMA or other bus mastering
activity.  So timing problems caused by arbitration issues are not going
to get caught by memtest86.  What memtest86 should be really good at is
finding bad memory or identifying memory access timing screwups.


>
> Motherboard.  I'm not a huge fan of Abit.  Their quality of the years
> has been variable.  Maybe you could check your board for leaking caps.
> Instability in the power circuit would cause issues with stability.
> My first myth system was on a box with bad caps and caused me no end of
> stress and grief.  Replacing the MB and CPU with NF2 fixed it.  Later I
> replaced the caps on the "bad" board and now it runs like a champ.

I'm familiar with the "bad cap" problem.  Fortunately I've avoided that -
during the period of time when the bad caps were circulating I wasn't
buying anything.  I've got a local friend who has had to deal with bad
caps on at least 3 different boards.  He's actually done his own repairs
on such boards, with replacement caps from digikey.

I've actually been a fan of ASUS over the years.  I have an old ASUS P5A
SuperSocket-7 board that I think will just simply never ever quit.  The
K7V Athlon board I got afterwards was similarly flawless.  But then it
took me 2-3 weeks to stabilize an A7V-266E I got next (bad DRAM voltage
jumper).  The current hassle with this A7N8X-E has just been the pits.
The next board may be something other than ASUS.


>
> Good luck tracking down your problem.

Thanks.  I definitely need it.

The plan going forward will probably be this:

1. Put 2.6.9-ac14 on it and let it age a while, with apic and lapic
enabled (if the I/O problem was fixed then this should work safely now).

2. Swap PSUs.

3. Remove PCI devices one at a time (the box is fairly loaded) and see if
that helps.

I've been slowly slogging away at this for quite a long time.  I don't
expect any kind of quick solution.  I think I exhausted the quick
solutions a long time ago :-(

  -Mike


-- 
                        |         Mike Isely          |     PGP fingerprint
    POSITIVELY NO       |                             | 03 54 43 4D 75 E5 CC 92
 UNSOLICITED JUNK MAIL! |   isely @ pobox (dot) com   | 71 16 01 E2 B5 F5 C1 E8
                        |                             |


More information about the mythtv-users mailing list