> I have tried to compile the amber8 at linux cluster
> with intel compiler and mpich, however, the
> performance is very poor for sander, usually around
> 30% or less CPU usage for each CPU. If I use only one
> CPU, the usage is above 90%. I am not sure this is
> common for parallel sander in linux? or is it problem
> from mpich?
> Thanks,
> Chen
how big is your system? i observe such behaviour for small system
simulations (~1000 atoms) on linux cluster with ethernet connection
between nodes. if your system is small, try something big (10000 atoms)
and you'll see whether this is your problem.
regards,
--
Lubos
_._"
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Tue Apr 27 2004 - 18:53:01 PDT
