ATK vs VASP benchmarks

[staykov]

I wonder is someone performed benchmark test between VASP and ATK?

my quick try on 24 core machine for bulk SrTiO3 10 10 8 kpoints, scf-only and similar settings gives 54 sec ATK-LCAO, 336 sec ATK-PAW, and 374 sec VASP. I was surprised to see faster performance in the PAW methods for ATK. I guess, bulk is easy and convergence problems might appear for LCAO for complicated nanostructures so there VASP or ATK-PAW might turn more useful. 

[Nordland]

As a rule of thumb and of course, somewhat anecdotal, LCAO performs best for large and complex structures, where PW solutions perform best for smaller systems.