In 2008 sp1 a NlogN solver is added which should reduce simulation time significantly with large problems (> thousands unknowns).
Trying it with a simulation having 5000 unknowns (single layer, no holes but thick conductor with side currents) I note that the loading matrix phase (N2) is largely dominant in respect the simulation phase with all the methods, N3to NlogN).
So I should expect that for larger problems, using either the iterative solver or the NlogN solver the simulation time should increase as N2because loading time reamains dominant.
Is It correct?
Trying it with a simulation having 5000 unknowns (single layer, no holes but thick conductor with side currents) I note that the loading matrix phase (N2) is largely dominant in respect the simulation phase with all the methods, N3to NlogN).
So I should expect that for larger problems, using either the iterative solver or the NlogN solver the simulation time should increase as N2because loading time reamains dominant.
Is It correct?
May 14, 2008 2:52 AM