It is a common misconception that more data processing in SQL puts a higher load on the database. A typical database spends 96% of its CPU time on logging, locking, latching and marshalling [1][2] rather than processing SQL. By sending less data to the middle tier and performing fewer round trips, the use of stored procedures means that the database can actually do more real work.

[1] https://dzone.com/articles/mit-prof-stonebraker-%E2%80%9C

[2] https://drive.google.com/file/d/0B7jyeB8kxFPjU0VySkF3UHhoVnM...