GAMG vs. PCG in fvSolution
i'm working with two domains of exactly same geometry and case configurations. one domain has 150000 cells, the other about 6*10^6 cells.
in the coarse domain case the GAMG solver needs about 30 iterations for one timestep (nCellsInCoarsestLevel = 1000). on the other hand, if i choose PCG the calculation needs 230 iterations.
in the fine domain case it's the opposite: PCG needs about 1000 iterations but GAMG does need more (maxIter limitation at 1500, various nCellsInCoarsestLevel).
what's your experience with GAMG? any tips what i could try?
Did you find any answer?
I am solving a external flow simulation. When i increase the number of layers and decrease the min thickness close to 9.6e-06,in a GAMG solver the iterations reach close to 1000 and crashes but on the other hand PCG works very well.
Is there any reason behind.
Thanks for opening this topic. I have two suggestions:
1) Some people here on the forum say nCellsInCoarsestLevel shouldn't be above 50. On my cases I always make sure to make a "nCellsInCoarsestLevel-study".
2) I think your post would be even more meaningful if you would tell us the calculation time/iteration or per timestep. The number of iterations alone is not so meaningful from my point of view.
|All times are GMT -4. The time now is 00:20.|