Hello, let me share my experience.
I didn't do large system brenchmark, but
in minimizing human growth hormone 3hhra using WAT216 10A, one PIII 450 takes
245s
while a PIII500+PIII450 uses only 130s!
But for MD, there will randomly an error of
net_recv fd=??? error.
This make me must turn off MPI for MD -_-...
I've searched for the beowulf page that this is an error concerted with
MPICH,
not amber. But this error also is concerted with LAM-MPI too, as this error
appear in clusters with either MPICH or LAM-MPI, using intel CPU.
But didn't hear of that on other archs.
Does anyone have this error in his/her beowulf too?
Received on Tue Jun 13 2000 - 08:55:47 PDT
