hi everyone i'm modeling a large number of volume piles and as a result the program changes from pardiso to picos for lack of memory (8 gb) and then picos gives slow convergence ... i tried one core iterative method but it didn't help either is there anyway to force the solver type as pardiso even it'd take longer time taking into account the lack of memory? ... and if not how much memory are we talking about to use pardiso for such big models ?
thanks in advance
abdarhman dbass said:is there anyway to force the solver type
There is no way to force the solver type to stay Pardiso.
Pardiso uses somewhere near 4-5 times more than picos, but it is not a rule.
You have two options:1) upgrade memory;
2) find mesh with fewer elements.
abdarhman dbass said: how much memory are we talking about
It depends on dofs and parallel tasks. There is no limit.
Dear Abdarhman,
Indeed, as mentioned by the other comments from dep25 niiosp, PLAXIS automatically checks the number of the degrees of freedom to be solved and if Pardiso would require many GB of memory it switches to Picos solver to be more efficient.
There is no generic rule as the number of elements, the type of analysis, the boundary conditions can affect the total number of the degrees of freedom to be solved, so there is no simple answer to that.
Note that slow convergence can be an issue of many things in PLAXIS. I would first start by checking all the material parameters and then some of the following Output plots to indicate what the problem might be:
A combination of these plots will usually give insight into what is happening.
thank you all very much this was helpful