But how fast can it run a single thread Python loop through a massive dataframe. Don't even suggest vectorization.