Is Apache Spark SQL generally faster with a DELETE and INSERT, or a MERGE?

Question

If one only wants to insert new rows, not update existing ones: Is Apache Spark SQL generally faster with DROP TABLE and INSERT, or a MERGE & INSERT? Or some third option.

There's quite a bit written on traditional SQL, but I don't see much for Spark SQL.

If it depends...depends on what?

(I am working within Azure Synapse, writing to a Delta table, if that makes any difference0.

Answer 1

DROP TABLE and INSERT would be relatively faster as MERGE and INSERT of Delta Lake involves JOIN operations (Inner Join and outer join), which makes the MERGE slower and costly