Not quite. You can perform the tests yourself on bigger table. The bigger the OFFSET the bigger the delay on returning data. This is due to fact that it's easier for database to pick only 20 top rows than order whole table to get 20 from the middle. Using simple thought experiment: what is the sorting cost if you pick only top 20 elements? How would you implement this? It's complexity is less than n*20, right? In the meantime worst case scenario of sorting whole table could be even n^2. Mature databases uses the fact of LIMIT used for sorting optimisation.