In metal complexes with only a few metal centers, using ECPs gives a very negligible speedup. Using LANL2x or SDDx for these is generally a bad idea for publicationgrade calculations, not because they use an ECP but because the non-ECP part is so small, e.g. having no polarization functions which are important since frontier orbitals of such systems often involve the metal basis functions, and these orbitals are the most important for redox reactions and excitations