i am working on a simple time dependent simulation that include a single domain of a conductive gas (custom made T dependent material ) with electric current + heat transfer in fluids interfaces, so the equations are the current conservation and heat transfer equation with a single heat flux boundary + everything else is insulation , and joule heating calculated as a variable (no multiphysics). i used a very fine mesh, and i run the simulation on a single core, and 40 cores (2x20cores) on the hpc, and i noticed that the single core one run similar or even faster than the 40 core one, so i read the blog "Understanding Parallel Computing " https://www.comsol.com/blogs/understanding-parallel-computing/. i realized that my model is a Completely Serial problem, and i was wordering if there there are ways to help comsol parallelize my model and use as much cores as possible since it is stated in the blog "The possible speedup is mainly governed by the speedup possible when solving the partially parallel system of linear equations". in case of a serial problem, would a cheap 4Ghz i3 processor run faster than the hpc xeon 2.8Ghz one? since it can support up to 64Gb of ddr4 ram as well.
↧