There is a comparison OpenLB vs. OpenFOAM for single phase flows: https://scholar.google.com/citations?view_op=view_citation&hl=en&user=sCl2dpIAAAAJ&citation_for_view=sCl2dpIAAAAJ:blknAaTinKkC on CPUs. OpenLB was found 32 times faster fixing the accurany.
Now, OpenLB is running on GPU clusters which accelarates simulations dramatically. Also the free surface modell runs on GPU clusters now!