ECC memory problems on AXPpci33/Multia
I have a 166MHz AXPpci33 board, which I gather is essentially the same
hardware as a UDB/Multia, without the ethernet and video. I got four
standard PClone 4Mx36 (16Mb parity) SIMMS for it. I had ECC traps fairly
frequently with these, so I wrote a memory test program which allocates
a virtual memory block the same size as physical memory and walks
through it, writing a value and reading it back. I can tell that this
walks through most of physical memory, because it only starts swapping
to disk on the last four or five megabytes.
Using this program, I determined that one SIMM was the source of the
problem, because the fault addresses followed the SIMM when I swapped
banks, and there were no traps when it was not installed. I got a
replacement from the dealer. When I tested this, I found that it was
better than the original, but it still coughed up a few bit errors. When
I ran the test all night with two of the three SIMMs which I thought
were good, one ECC trap occurred.
What are other people's experiences with this board? My feeling is that
ANY memory errors at all is too many, because if it can have a
correctable single-bit error, it can also have a non-correctable double
bit error or a non-detectable triple bit error. Is this unrealistic? Do
you just have to expect a few memory errors occasionally?
My inclination at this point is to try to get a refund for all four of
these SIMMs and go somewhere else. Supposedly, these are 70ns parts, but
maybe the timing is right on the edge of what the AXPpci33 will accept.
Would 60ns parts be more reliable? Does anyone have recommendations as to
brand names or suppliers?
Any thoughts or experiences would be appreciated.
Ian Willmott
Northern Telecom
Ottawa Ontario Canada
willmott@nortel.ca
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-alpha-request@lists.debian.org .
Trouble? e-mail to templin@bucknell.edu .
Reply to: