[gmx-users] system reshaped on itanium II

[Jens Krüger]

Hello,

I observed a strange behavior of my simulation systems on a new Itanium
II Cluster. Whatever system I use, it behaves well on a normal P4, but
when transferred to the Cluster, the box expands along the z-axis and
the water molecules "regroup" from typical brick shape to an
octahedral-bipyramidal shape. After ca. 50 ps the whole system crashes.
I varied all parameters and rebuilt the systems from scratch but the
problem still remains.

What I have tried so far:
- changed the number of nodes
- varied pcoupltype (none, isotropic, ...) with different coupling
intervalls
- varied Tcoupl
- varied coulombtype (cut-off and PME) with different parameters
- tried all variations of grompp with or without -shuffle -sort
- double checked, if I am really using a cubic/tric system and NOT a
octahedron
- used clean protein/water, DPPC/water and protein/POPE/water, each
running stable on a Linux  P4
- energy minimization cg and steep

All of them did not had any influence on that strange failure.

Has anybody an idea what might by the reason?

Thanks,

Jens
--
========================================================
Jens Krueger
mercutio at uni-paderborn.de
dienstlich:
Warburger Str. 100 * 33098 Paderborn * Tel.:05251-602183
privat:
Warburger Str. 64 * 33098 Paderborn * Tel.:05251-2025608
========================================================