The purpose of serialize option is to reduce the locking/blocking overhead between different sessions, with serialize feature you have the ability to specify that all rows with a given PI value gets mapped to the same session. this will save the job from getting tangled up because different sessions trying to lock on the same rowhash.
You probably are better off without it if you are loading a table with UPI. (apart from hash synonyms there won't be any collisions).
More information is available on the tpump manual.
TPUMP is doing updates to the PPI table on daily basis. Whether the serialize on multi session TPUMP works fine? My basic question is to understand whether the Serialize work on multiple session TPUMP to the PPI table? Please advise