Statistics collection recommendations – Teradata 14.0 - comment by carrie

Robin,
 
By any chance are you collecting statistics with the USING SAMPLE option? If sampled stats are being collected on a NPPI table, when you change to a PPI table the optimizer may decide to sample each partition for the stats being collected on the primary index and the partitioning columns.  This can take more time compared to a NPPI table where sampling is only at the beginning of the table.   Check your output and see if it is those two columns that are contributing to greater resource usage.
 
Adding statistics collection on the column PARTITION is not likely to add much time or resources.  The elapsed time for collecting on PARTITION is usually a second or two, as the process just reads the cylinder indexes, not the base table rows.
 
Thanks, -Carrie