Re: [gentoo-user] Random reboots. Where to start?

[Paul Hartman <paul.hartman+gentoo@gmail.com>]

On Fri, Feb 25, 2011 at 9:33 AM, Dale <rdalek1967@gmail.com> wrote:
> Well, I think my machine is possessed or something.  I'm getting random
> reboots here.  When it does this, it is like hitting the reset button.  It
> is sitting on the grub screen when it does this.  I noticed the first time
> the other day and this was before adding the extra memory.  I seemed to be
> stable at 4Gbs but I seem to be rebooting at random.  I ran memtest
> yesterday, it checked fine.  It didn't find a error but it looked like it
> was only testing part of it.  Memtest recognizes all 16Gbs on the last run
> but it didn't seem to be testing it all.  Is there a trick to getting it to
> test the whole thing?

When you say "memtest" what memtest are you using, exactly? The one
from the kernel?

I prefer memtest86+ as it is updated and has support for the latest
CPUs and memory configurations. You can install it from portage and
add an entry to your Grub menu and don't need to mess with bootable CD
or USB or anything like that.

You can test specific ranges, if you suspect the new RAM is causing
trouble. Full memory testing of all patterns with 16 gigs of RAM can
take forever, but in my experience tests 5 and 8 in memtest86+ are
typically the only tests that actually produce errors on modern
systems. If you're in a hurry you can just run test 5 and that'll give
you many more passes in a shorter time. I would at least want to run
this kind of test for 12 hours with no errors before trusting the
machine. 24 or 48 hours if you can afford the wait. :)

If it does not always recognize the full 16GB i would suspect you need
to increase the voltage to your RAM. You may also (or instead) need to
reduce the memory speed.

On my previous motherboard, an Abit, with Patriot DDR2 RAM, it could
handle 4GB of RAM (2x2GB) no problems, running at recommended voltage
and full speed. When I doubled that to 8GB (4x2GB) it crashed often,
but not constantly. It could not pass an overnight memory test. I
ultimately had to raise the voltage by 0.3 and reduce the speed from
800MHz to 667MHz. I ran memtest86+ for 3 days and it had no errors.
After that it worked like a champ for 2 years, no problems.

Also, if you're using DDR3 which contains XMP data (timing and voltage
presets, basically) beware that it can sometimes be wrong. I have used
2 different brands of RAM whose XMP data did not match the values
printed on the packaging. The manufacturers both times suggested I use
what's printed on the packaging and ignore what the chip itself tells
me.

And of course on my recent Core i7 920 build, I spent a month trying
to get OCZ Gold RAM to work properly with my Gigabyte motherboard.
After 2 DOA sticks exchanged and a month of trying everything I could
possibly think of it still failed memory tests (sometimes it would
only fail after 5 or 6 hours of testing) and I gave up and returned it
to the store for a refund. I ordered some Corsair XMS3 RAM online
instead, it worked right away with the recommended settings, no
messing around, and I've been running happily ever after. :)