Dear Amber mailing list,
I have been successfully running molecular dynamics simulations on
vitamin B12 (a hexacoordinate cobalt complex) systems using pmemd on
CPUs for some time.
When I tried to rerun exactly the same simulations on pmemd.cuda.SPDP
(on NVIDIA Tesla M2090 GPUs using amber 12 with bugfixes 1 to 7
applied) the system blows up with the following error message:
ERROR: max pairlist cutoff must be less than unit cell max sphere radius!
This occurs after a few steps (usually 1 or 2).
In the zipped folder attached I have included all of the necessary
.lib .frcmod and .pdb files and a tleap script to generate a simple
solvated hexacoordinate B12 system.
I have also attached the input and output files and submission scripts
used to run minimizations and short unrestrained MD simulations
(successful on pmemd and unsuccessful on pmemd.cuda.SPDP).
I created the .lib and .frcmod files myself based on parameters in the
literature. To see if the problem was with my parameters I tried to
simulate a solvated heme molecule (a very similar system to B12) using
the parameters from the Bryce group database
http://www.pharmacy.manchester.ac.uk/bryce/amber/ (which by the way
appear to be missing nonbonded parameters for a few atom types).
Exactly the same error occurred with heme as with B12
I found that if I removed the cobalt atom entirely the problem
disappears. The problem also disappears if I remove most of the bonds
to the cobalt atom.
Many thanks in advance to anyone who can shed some light on this.
Patrick von Glehn
PhD student
Harvey and Mulholland groups
University of Bristol
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Aug 20 2012 - 17:00:05 PDT
