The primary index for a table controls the distribution and retrieval of the data for that table across the AMPs. Both distribution and retrieval of the data is controlled using the database hashing algorithm.
If the primary index is defined as a partitioned primary index (PPI), the data is partitioned, based on some set of columns, on each AMP, and ordered by the hash of the primary index columns within the partition.
Data accessed based on a primary index is always a one-AMP operation because a row and its index are stored on the same AMP. This is true whether the primary index is unique or nonunique, and whether it is partitioned or nonpartitioned.