MultiMech is fully parallelized and has developed numerous techniques to improve the simulation runtimes for a multiscale simulation. However, incurring some computational expense for the gains in accuracy is expected.
The software is shown to be comparable in speed to other analytical multiscale softwares, but far more accurate and flexible. Further, it is between 100 and 1,000x faster than the gold standard "Finite-Element-Squared" methods.
The RVE size and damage modeled greatly effects the simulation time. As a benchmark, a simulation of 20,000 global multiscale integration points can run on a 32GB workstation with 8 cores in under 2 hours. As the available memory and available cores increase,the runtime will go down.
See this blog post for more information on the different approaches to composite analysis.