AMBER Archive (2000)

Subject: Shake and a new computer system

From: R. M. Fesinmeyer (rmf_at_u.washington.edu)
Date: Tue May 30 2000 - 12:50:25 CDT

My research group recently picked up a new linux system and I've been having
some problems running Amber on it. The system is a dual PIII 650 on an Asus
P2B-D board w/256 MB RAM running RedHat 6.2 (and the corresponding SMP
kernel, v.2.2.14). Amber6 was compiled with gcc 2.91.66 using the
Machine.g77 config file. No attempt has been made to make the application
SMP capable, as we expect to be running multiple trajectories at any given
time.

With this setup, I've been using Sander to perform MD runs for ~20 residue
peptides in water boxes. My 498K/80 bar run terminated after 240ps with the
following error:

COORDINATE RESETTING (SHAKE) WAS NOT ACCOMPLISHED
     WITHIN 3000 ITERATIONS

Similar runs at higher temperatures or lower pressures result in failure
after a shorter period of time. This makes sense, however I'm surprised
that I did not see these errors on our previous linux machine (although we
never attempted trajectories much over 50ps on it).

Does this sound like a hardware/software issue, or is it simply a
consequence of my methodologies(see below)? In either case, suggestions (or
reference, if appropriate) are welcome.

Thanks very much for your help.
Matt

The following is a representative Sander file used for these trajectories.

&cntrl
        imin=0, irest=0, ntx=1, tempi=0
        ntt=1, temp0=498, tautp=0.2,
        ntp=1, taup=0.2,
        ntb=2, ntc=2, ntf=2,
        dt=0.001, nstlim=1000000,
        ntwe=100, ntwx=100, ntpr=50,
&end