In PPI (Partitioned Primary Index) tables, rows are first organized by partition number at each AMP, then by hash value within each partition. This two-level sorting (partition then hash) enables efficient partition elimination in queries. Option B correctly states that rows are first sorted by Partition Number.