Subject: Re: Memory parity error indications with ECC RAM
To: Jonathan Stone <jonathan@DSG.Stanford.EDU>
From: Dr R.S. Brooks <>
List: tech-kern
Date: 05/02/2002 16:17:45
On Tue, 30 Apr 2002, Jonathan Stone wrote:

>In message <>,
>Rex McMaster writes:
>>On Wednesday 01 May 2002 01:57, Dr R.S. Brooks wrote:
>>> We have two systems (one a new machine destined to be a webcache here,
>>> the second a "new" system for home built out of bits bought on eBay).
>>> Both have ECC RAM and Intel chipsets, and both are showing problems
>>> which may indicate memory errors.
>>> Please can anyone confirm that all corrected memory errors do result in
>>> an error log message?
>>I am unsure about whether the i386 port reports ECC errors (can someone
>>clarify this?) - I could only find references to ECC in the alpha port when
>>tracing a problem a couple of months back, which appeared to be memory based.
>>What are the symtoms you are seeing?
>Intel "server" boards (N440BX or newer) will report and log ECC via
>IPMI.  You can find the log via the server entry in the BIOS screens.
>The IPMI-aware systems I've seen can be configured to report SERR,
>PERR, or both, via NMI.   Lower-end systems should report via NMI.
>What kind of motherboard, exactly, is the suspect?

The one which did turn out to be memory error is a Supermicro P6DGU.
I had installed 2 x 128MB and 1 x 512MB DIMMs, of which the 512MB one
was seriously faulty (but not detected by the BIOS memory test).
When I ran MEMTEST86 it produced errors by the yard, but incredibly
I could boot 1.5.2 and even run KDE2 without any apparent problems.
It was only when I tried to build a kernel that problems appeared.

The new system(s) now appear to be SCSI problems related to the Intel
STL2 motherboard.


Roger Brooks (Systems Programmer),          |  Email:
Computing Services Dept,                    |  Tel:   +44 151 794 4441
The University of Liverpool,                |  Fax:   +44 151 794 4442
PO Box 147, Liverpool L69 3BX, UK           |