Here's an idea. It could be that one or more of your components are not properly constrained. Have you tried using the same fixtures in a static study? This may help you to find out which component(s) are not constrained (assuming that to be the cause of the error). Turn on gravity and run a static study to see if static equilibrium is achieved. If not, apply a fixed restraint to one vertex on each part and that will show you which part is not constrained properly. Then, apply bonded contact, or adjust mates, or otherwise fix it up so that it works.
Other than that, 5 million dof seems pretty huge. It may be that Cosmos cannot handle a problem that big. Is there any way you can sub-divide the problem to make it smaller?
Okay, that's two ideas. Good luck!