Dear amber users,
I have installed amber 8 on a cluster (x86_64 GNU/Linux) and I performed RAMD
calculations (with the sander module / RAMD patch) on 2 processors. I have also
installed the serial version of amber8 on my local machine (with the compilers
i686) and carried out RAMD calculations (sander module/RAMD patch, on one
processor). I obtained different results (see attached files) according to the
calculations have been performed either on local machine (serial version of
SANDER/amber8) or on the cluster. With the serial version, my calculations are
often incomplete and there are problems on the energy of the protein-ligand
complex. The calculations stop after a maximum simulation of 2000 steps while
on the cluster (with the sander parallel version) these problems do not arise,
and the simulations last longer.
I would like to ask you whether this difference is due essentially to
compilation problem? or whether in fact the RAMD patch works better in the
parallel version of amber8? or if this is due to another problem ?
Thank you for your help
Waad RHAIEM
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun May 09 2010 - 17:30:06 PDT
