Few things that come to mey mind:
You can check if there is any product join. Check if you can avoid redistribution, avoiding IO and CPU. If can use column-partitioning, also there is a chance that it may reduce CPU usage or it can be the other way round too, increasing more CPU usage, with autocompression.See if you have compression and decompression, needing to balance IO vs CPU . Single-amp operation will not attribute much to CPU.Conditional statements may increase CPU. Partitioning design fault can cause high CPU usage. How many duplicates NUPI per value, per row.It can be caused by checksum too.
