Hello,
The answer to your question depends on many parameters:
- the performances of your PC (number of processors and cores per processor, performances of your cores, RAM, how old...),
- time step and size of the elements (eg to respect Courant number criterion),
- total physical time you want to simulate,
- parameters you can choose in the steering file (in particular the schemes for advection, the solvers you choose to solve different steps)...
What is certain is that for 3D, it is strongly recommended to use parallel computation.
When you talk about 50,000 elements, you mean 2D elements that you extrude on the vertical or directly 3D elements?
On clusters we can use here, many calculations have been carried out to test the scalability of Telemac-3D, and they show very good performances.
If I remember, I used a 320,000 3D elements (32,000 2D elements x 11 planes) as a simple test case to model 25h time period of time for a coastal model (2 tides) and on my desktop PC with 2 quad-core processors, it takes me around 1h CPU time (a few minutes only with 64 or 96 cores on a cluster of ours).
Regards,
Chi-Tuan