[Beowulf] Repenting for sins against Dell (on good Friday, no less)

Greg Lindahl lindahl at pbm.com
Sun Apr 19 13:06:06 PDT 2009


> dmidecode | grep -A7 "Physical Memory Array" | grep "Error Correction" 
> 
> If that says "None" then there's no ECC goodness for you..

This only works on some of my systems -- a big pile of Intel
SuperMicro boxes w/Phoenix BIOS says "None" when I'm sure there's ECC,
because I've seen some single-bit errors recovered. A brand new
SuperMicro Nehalem box with a Megatrends BIOS does indicate ECC.

On desktops, Intel 975 mobo correctly says ECC, while ASUS mobos with
Megatrends BIOS say "None" while I think it's on.

My bet is that this is part of the BIOS which is customized by the
OEM, so we shouldn't blame Phoenix or Megatrends, and instead blame
ASUS and SuperMicro. Poor QA at the least.

I did discover one of my AMD servers was incorrectly configured, so
this exercise was worthwhile.

-- greg





More information about the Beowulf mailing list