Instead of hashing all files unconditionally, the new algorithm first checks if there are multiple files of the same size. Files with a unique size cannot have duplicates and so don't have to be hashed at all.