So it is the same model? That was not clear to me.
With the original file from CopperGenie, the same as message as CopperGenie.
That is too less. The original file from him needs for CCX ~ 25GByte, for the run only. For loading and postprocessing it needs a little bit more. From my experience this needs ~ around the factor 1.25 to 1+1/3 more. Just a little bit too big with my 32GByte. I have swapping deactivated per system, because a have a SSD and the lifetime of a SSD depends, how often it is rewritten (not reread). It is a general a bad idea, to swap to a SSD (except, it is any kind of special and made for it). Swapping to a HDD would need endless time. If CCX swaps, i do not know, i can imaging, that it does not, because for a calculation like FEM, RAM is a general requirement. Maybe there is an option, which must be activated.
No idea, i did not tested it.
But just be clear, the original mesh is way too fine for a 2nd order mesh, especially because everything is that fine, which is completely unnecessary. For FEM you should generally optimize the mesh to your hardware and needs and general circumstances. Also a 1st order mesh needs much much lesser RAM with the same count of nodes, if then the calculation is valid, is an other case (depends on the loading and their directions, geometry, and so on. As i posted the file above, i have optimized the 2nd order mesh in that way, that it need not really much RAM, but runs (except a peak in the mesh, which comes from a particular not good geometry (sharp edge fillet)) pretty well.
Greetings
user1234