Memory Testing... or problems

Dean Waldow waldow at rainier.chem.plu.edu
Tue Jun 27 01:18:00 PDT 2000


Greetings,

I think this is a little off topic but related to building clusters...

I have been trying to finalize our cluster and bought a board/cpu to
prototype.  I bought an asus p3v-4x board and a 667 PIII/133 cpu with
128MB PC133 ram.  I loaded linux and then ftp-ed my binary code to
benchmark.  I was quite surprised with it finished in 1/3 the time of
what I expected.  As it turned out, the program was basically returning
non-sense. (It is a monte carlo simulation and some how switched labels
on beads and lost beads in the lattice...)  I checked to make sure the
file transferred correctly and it did (via binary) at least in exact
size. I tried a different distro of linux.  I tried compiling on the new
machine. I finally looked at the 10 ram sticks I got and discovered that
they gave me about 4 different brands.  I had randomly pulled one out of
the bag which turned out to be a one of a kind.  I switched to another
pc133 I had running well on another machine (a p3b-f) and my benchmark
code ran fine using same binary. Having thought I figured it out, I
tried another version of the benchmark compiled with the portland group
compilers and the non-sense returned.  

My question is regarding testing memory since my main conclusion was
that it must be a memory problem if switching memory fixes (or partially
fixes) the problem.  However, my 'simple minded' thought was that if it
was bad memory linux would have problems too... I don't notice any.  

I have found Robert Brown's memtest.tar which tests for timing /
performance if I am understanding it correctly.  Are there other
recommended programs which might test memory integrity or basically look
for bad memory.  

Alternatively, am I missing the boat and might this be a mainboard or
other problem?  I have tried switching components with another linux box
with an asus p3b-f mainboard.  I can't get the p3b-f linux box to run
the binaries incorrectly. 

Thanks for any suggestions for what might be a pretty 'newbie' question.

Dean
-- 
-----------------------------------------------------------------------------
Dean Waldow, Associate Professor      (253) 535-7533 
Department of Chemistry               (253) 536-5055 (FAX)
Pacific Lutheran University           waldowda at plu.edu
Tacoma, WA  98447   USA               http://www.chem.plu.edu/waldow.html
-----------------------------------------------------------------------------
---> CIRRUS and the Chemistry homepage: http://www.chem.plu.edu/         <---
-----------------------------------------------------------------------------




More information about the Beowulf mailing list