[DFTB-Plus-User] Inconsistent behaviour of dftb+ on different machines

[Matteo Tiberti]

Hi to everyone,
I'm trying to perform some optimizations using dftb+, but I'm experiencing
an inconsistent behavior of the program on different machines.
I'm running my calculations on a new dual-Xeon workstation and on an older
Intel server. Both of them are 64 bit machines and I'm using the x86_64
versions of the precompiled binaries on both.
While my optimization completes successfully in few cycles on the older
machine, running an identical calculation using the same pre-compiled
binaries on the newer one results in energy fluctuations and the calculation
never converges. This only happen, anyway, when using a relatively large
systems; when the system has less than about 300 atoms dftb+ works fine on
both computers. I don't think it's a memory limit problem, as the RAM is
never saturated.
The problem happens both with dftb+ 1.1 and 1.0.1, with the pre-compiled
binary downloaded from the site and with a version compiled from source
using gfortran and the standard gnu libraries.
Are you experiencing anything similar? Do you have any suggestions on how to
address this issue?
We also tried contacting the developers, but I still haven't received an
answer so far.

Thank you for your attention
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20110914/234db730/attachment.htm>