[vox-tech] When RAM goes bad...

[Karsten M. Self]

on Sun, Mar 06, 2005 at 04:55:14PM -0500, Peter Jay Salzman (p at dirac.org) wrote:
> I suspect one of my machines has bad RAM.

I believe in corporal punishment:  beat your stick!
 
> Out of the blue, Unreal Tournament occaisionally segfaults.
> 
> Then Nero is no longer capable of verifying burned DVDs in Windows.
> When I boot into Linux, the burned files on DVD and resident files on
> the hard drive have the same md5sum, so Nero's verification is faulty.
> The burn looks good.  Verification is ill.
> 
> Played some Quake3 while KDE libs were downloading, and it just
> segfaulted.  It never did that before.  In fact, none of these things
> ever happened before.
> 
> Everything on both OS's points to bad RAM.  The RAM is only 2 or 3
> years old.  Is it unheard of for RAM to die that quickly?

Several possibilities:

  - Various permutations on bad RAM / slots.
  - Bad caps (more below).
  - Bad CPU.
  - Other mobo components.

 
> I've never run memtest86 before, but I got it running right now.
> Aptitude got it, made a boot floppy and it's running.  Looks like it
> may take awhile.

You want memtest86+  It can be run as a boot option via GRUB/LILO.
/usr/share/docs/memtest86+ has the appropriate stanzas.

> I've never come across this piece of bad luck before.  Any other tools
> to look at?  I only knew of memtest86 from this mailing list.

  - memtest86+:  test RAM.  1-12 hours recommended.

  - cpuburn:  stress CPU / cooling system.  5-20 minutes recommended.

  - Several alternatives, including disk tests, in the event of, say,
    bad swap, controllers.
 
> Any other words of wisdom?  Except for the odd hard drive, all my
> machines outlived their usefulness rather than components dying before
> their time.  This is a new one on me.

What hit me last summer:  bad capacitors.  There was a batch released
about 2-3 years ago, and they're starting to fail with increasing
frequency.  Fortunately, you can sometimes visually scan for 'em.  Brown
crud on the largish, tall cylinders (think a knuckle or two of your
pinkie finger) is a sign of things Not Good[tm].

My symptoms were increasingly frequent hard freezes in various
circumstances, ranging from light CPU loads, to overnight w/ no apparent
cause.

  - memtest86+ produced pretty spectacular random character painting
    across thet screen, in certain tests (it runs a range of test, IIRC,
    5 or 7 was where things went south).

  - cpuburn failed out pretty fast.

  - Swapping RAM sticks (had to buy one for this purpose) made no
    change.  Likewise, swapping banks.  Got myself up to 1 GiB on
    account of this....

Visually inspecting mobo showed brown crud.  My vendor kindly repaired
the damage w/o charge on account of a GNU/Linux tips page I've written
about their kit.  Thanks again, CappuccinoPC.  No repeat since repair.
 

Peace.

-- 
Karsten M. Self <kmself at ix.netcom.com>        http://kmself.home.netcom.com/
 What Part of "Gestalt" don't you understand?
    So many men, so many opinions.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://ns1.livepenguin.com/pipermail/vox-tech/attachments/20050306/261490dd/attachment.bin