I am not sure if this will help you, but I had similar problems when using
4 CPUs in TI runs in amber 9. We bought a new machine with 8 core and I
never had any issue in the TI runs using this new machine (either using 4
or 8 CPUs). The problem in my case was that it was stopping the runs
without giving an error in the outputs. There was error messages in the
log files, though. As far as I understood, it was about the mpich2
parallel implementation. I would suggest you to install the latest MPI
implementation and see if that solves it. Also, check the log files to
see what kind of MPI error you are getting. Hope this helps. Good luck.
Ilyas Yildirim, Ph.D.
---------------------------------------------------------------
= Hutchison Hall B#10 - Department of Chemistry =
= - University of Rochester =
= 585-275-6766 (office) - =
=
http://www.pas.rochester.edu/~yildirim/ =
---------------------------------------------------------------
On Tue, 7 Apr 2009, Hannes Wallnoefer wrote:
> Hi all,
>
> I am doing TI calculations in Amber10 with the soft core potential method. I
> followed the tutorial and just changed the forcefield (99SB instead of ff03).
> The minimization and the equilibration was ok in any case, but the production
> run for the complexes stopped after step 81900. This happens reproduciblly with
> any lambda value and in any of the three steps (charge removal, vdW
> transformation, and recharging). If i restart the produvtion run after step
> 80000 it makes another 81900 steps and again stops. Stop doesn´t mean that it
> stalls, but it simply doesn´t write any output any more, although all of the
> CPU´s (i use sander.MPI with 16 cores) have a load of 100%.
> If i reduce the number of steps to 80000 for any proudction run i get a similar
> result as described in the tutorial and everything seems to run without any
> problem.
> I also tried another example than the T4 lysozyme from the tutorial and observe
> the same behaviour.
> Does anybody have an idea what happens? Is there something like an internal step
> number limit (81900 seems a little bit strange)? Could there be a problem with
> the compilation? Is there another important information i should provide?
>
> Best regards,
> Hannes
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
>
Received on Wed Apr 08 2009 - 01:19:28 PDT
