If you want a fast convergence, probably yes. Just a technicality, if you are using OpenFOAM implementation of GAMG, you should set the nCellsCoarsestLevel more or less equal to the square root of the number of cells.

If you are looking at scalability, GAMG is not the best choice and conjugate gradient methods tend to scale better.

For a large case, I think GAMG is still the way to go, even if it scales less because the reduction of the number of iterations is typically very large compared to other methods.

Dear Alberto,

interesting. It is almost 2 years old but do you know if there is some interesting literature / paper / publication out based on GAMG in FOAM? I am solving stress equation in my own solver which took a lot of outer iterations for the convergence (lot of explicit terms). However, I am not sure if I can speedup the case because I always just use the standard GAMG settings.

Kind regards,

Tobias

Hi Tobias Holzmann ,

I know this is a quite old question. But I'm just curious if you were able to find something interesting on the topic you mentioned. I am currently developing solid stress model in OpenFOAM myself and having the exact same problem you had. Since I have thermal stress, plastic deformation and phase transformation in the model, the stress equation has many explicit terms and it takes a lot of outer iterations to converge. I'm also using the standard GAMG settings.

Thanks,

Duo Huang