[gmx-users] Segmentation Fault, unless using bond constraints

[Joanne Hanna]

Hi

I have generated two topology files for AMP. One by modifying the ffG43a2 ATP rtp and the second using PRODRG. The main difference between these files is the number of dihedrals defined about the ribose ring.
I am 'testing' the stability of these topologies by solvating and energy minimising (steep POSRES amp, steep, cg) and then doing MD runs (POSRES AMP (no press coupling), no POSRES (pressure coupling)). 
My question is that as long as i define some type of bond constraint, either hbonds or all-bonds, everything runs fine. If I do not define any bond constraints the MD run fails to pass the first step and fails with a segmentation fault.
I was wondering if anyone could suggest why this was? Should this give me concern that my topology is incorrect? 
I am slowly heating the system (tau_t = 2) and my starting structure is minimised to convergence with decent energy values.

Thanks
Jo