Parallel SQL does not typically parallelize the execution to more than two threads. Parallel SQL works best in cases where SQL interpreter does a lot of processing and also does a lot of scanning of rows from the database engine. Examples of such cases are aggregate calculations from a full table scan or merge joins with full table scans. In those cases, response times can be expected to be 20 to 50 per cent shorter than in default setting. Since the execution will not parallelize to more than two threads per statement, the CPU usage will not increase to more than twice the amount of default setting.