Re: memtest86+ fails at address '00000000000 - 0.0MB'

[Peter Hillier-Brook <phb@hbsys.plus.com>]

Dave Ewart wrote:
> I've got a Debian Woody server (three-year-old dual Xeon thing) and as a
> result of one or two little quirks (failed CRC sum for some gzipped
> files etc.), I decided to run memtest on it.
>
> The memtest produced an error at location '00000000000 - 0.0MB', with
> output roughly as follows after running for about 16 hours (I've
> transcribed this, so it might not be 100% accurate):
>
> Tst Pass Failing Address    Good     Bad    ErrBits   Count Chan
>  8    1  00000000000 0.0MB  uncorrected        0000   ECC    0
>  6    2  00000000000 0.0MB  uncorrected        0000   ECC    0
>  4   12  00000000000 0.0MB  uncorrected        0000   ECC    0
>  6   13  00000000000 0.0MB  uncorrected        0000   ECC    0
>  8   14  00000000000 0.0MB  uncorrected        0000   ECC    0
>
> I replaced *all* the RAM in the box and ran memtest86+ again, and after
> a few hours I'm getting an error at '00000000000 0.0MB' again, with
> results similar to the previous run.
>
> Is this memory address something 'magical', is it really part of the
> main system RAM?
>
> Can anyone suggest what might be wrong here?
>
> Thanks for all comments/suggestions...
>
> Dave.
Dry joint on the memory socket in the motherboard? You've described the 
classic symptoms of a dry joint failing at higher temperatures as the 
system/motherboard warms up.

Peter HB