[ale] Losing stability

[Jim Popovitch]

On Mon, 2006-10-30 at 11:15 -0700, JK wrote:
> Jim Popovitch wrote:
> 
> > On Mon, 2006-10-30 at 12:50 -0500, Matt Kubilus wrote:
> > 
> >>memtest86 is only really useful to prove that the memory is bad, not
> >>that the memory is good.
> > 
> > 
> > It's orientation is binary, right?  If not bad, then good.  Or am I
> > missing something?
> 
> memtest86 essentially writes and reads specific bit
> patterns to every location in RAM, and ensures that the
> data read is identical to the data written.  But it cannot
> test every conceivable bit pattern and access sequence.
> If memtest86 does not find a problem, that only means that
> its tests didn't tickle any condition that causes the
> memory to fail.  Of course that will always be true for
> good RAM, but it might also be true for RAM that is
> damaged in some very subtle way.

I thought memtest86 did an exhaustive series of tests.  If so, I would
think it would be sufficient for most cases.

BTW, IIRC, all memory is damaged in some very subtle way.  There is no
such thing as perfect memory due to the manufacturing process, CPUs for
that matter too. ;-)

-Jim P.