Some other factors to think about:
Is there only one such process inserting rows, or multiple inserts working in parallel?
If you have a single-partition table with no clustered index, or clustered index on an key that keeps increasing (like datetime), then all the inserts are going to the same last page of the table, which can result in contention on that page. Round-robin partitioning can help out then as inserts are randomly distributed among the partitions and each partition has its own last page.
Having a clustered index on a key with a fairly random distribution results in the insertion point being randomly distributed in the table space. You get occasional page splits, but very infrequent contention for an exclusive lock on the same page.
What is the lock scheme of the table?