Memtest86

So I've got several one-time errors in linux. Stray errors when
compiling in gcc (You know, syntax errors, bad characters), but I can
never repeat the errors a second time. Also had filesystem corruption. I
booted and init couldn't find my runlevels. (The acces was forbidden
even for root) Filesysten tree rebuild fixed the thing and I ran a bad
block checker. No problem found. So I thought that maybe I had bad ram.

I run memtest86 with memory sizing set to Bios-all and all the tests.
Get 40000 errors in all. I retry it, removing one stick at a time (two
256) and trying combination of slots. No problems. Only get problem when
both sticks are in. What can I do? Do I need to replace them both.

Also, could the fact that my case is always open (I forgot one side of
it where I was living before.) cause some problem. Can dust cause short
circuit?

Thanks

Jeremie

On Fri, 29 Oct 2004 10:34:51 -0400, Jeremie Lasalle-Ratelle
<lasallej RemoveThis @iro.umontreal.ca> wrote:

 >So I've got several one-time errors in linux. Stray errors when
 >compiling in gcc (You know, syntax errors, bad characters), but I can
 >never repeat the errors a second time. Also had filesystem corruption. I
 >booted and init couldn't find my runlevels. (The acces was forbidden
 >even for root) Filesysten tree rebuild fixed the thing and I ran a bad
 >block checker. No problem found. So I thought that maybe I had bad ram.
 >
 >I run memtest86 with memory sizing set to Bios-all and all the tests.
 >Get 40000 errors in all. I retry it, removing one stick at a time (two
 >256) and trying combination of slots. No problems. Only get problem when
 >both sticks are in. What can I do? Do I need to replace them both.
 >
 >Also, could the fact that my case is always open (I forgot one side of
 >it where I was living before.) cause some problem. Can dust cause short
 >circuit?
 >
 >Thanks
 >
 >Jeremie

This type of problem is usually the motherboard itself. You
didn't mention the specific hardware but generic attempts
would include using first and last memory slots, raising
memory voltage in bios if possible (within reason) and
higher (slower) memory timings. Similarly it might help to
lower the memory bus (FSB since it's often tied to memory
bus speed) speed, but usually this is the less desireable
course since it lowers the CPU speed too, though a bit
slower CPU may be better than buying new memory, depending
on the role, use of the system.

If you can isolate a certain memtest86 test that produces
the most errors it can cut down on retesting time to just
run that test first for a few passes then if it does "pass"
that test, rerun all tests next, or if it doesn't pass, no
need to rerun before changing memory timings or bus speeds,
or whatever else you choose.<!-- ~MESSAGE_AFTER~ -->