Out of memory, solution does not converge, failed to find solution

I still try to solve my photovoltaics-model ;) I thought i might done something wrong and created a new model because i had that NaN/InF Error but well, it still doesnt solve it. To understand my problem better, i just attached a picture of the whole model but also a section.

Because i had several times the error: Out of memory, solution does not converge
i tried to change the solver to iterative and GMRES with Out of Core checked, but after 4 hours of solving the message that returns is:
Failed to find a solution.
Divergence of the linear iterations.
Returned solution is not converged.

I created every layer of the 5 layers (glass/polymer/silicon/polymer/polymer2) with the same plane geometry (quarter symmetry to save ram) wich i extruded in several steps for the thickness of each layer, then draw the framemodel and extruded it from both sides and united it with the virtual option of composite areas.

Then i set up a free triangle net on the top surface with refinement at the borders to the frame. I extruded this step by step thru the layers, made the convert option at the borders with an area-diagonal to fit with the free tet elements from the frame. Also i divided the glass layer with the distribution of 5 elements.
So far everything is ok. No errors or anything.

Thanks for the help ;)

The constraints are: Area load from induced by the circles on top, symmetry on 2 sides to save ram, fixed constraint at a predefined area.

First of all my question: What do i have to change to let it converge?
secondly: Wich solver would be the best? Id rather let it run longer with ram saving due i just got a Quad-core pc here with 8 Gb of Ram.