We perform MD simulations with Amber on a protein dimer in different circumstances. In implicit solvent on 350 K the dimer dissociates after 2 ns, while in water box, NPT run it is still stable after 400 ns. What could be the explanation for this?
First, your initial conditions are different at least velocities. Second, in explicit simulation, several thousands additional atoms of water molecules and ions hold off sampling of two chain conformations.