Why are there non-zero energy differences between the GROMACS ports and the AMBER 8.0 package?
There are two reasons. The first, as discussed below, is due to an inconsistency in the way AMBER orders improper torsions in TRP residues. This bug has been reported to the AMBER team, and our ports do not suffer from this issue. The second difference is seen in non-bonded interactions, which can have absolute values near zero. In these cases, the relative difference does not reflect the very low absolute error for these terms.