Guidelines on Optimizing I/O Operations, CPU Usage, and Disk Space Usage for Column-Partitioned Tables and Join Indexes
Other factors such as those from the following list can also play a role in reducing the number of I/O operations required to resolve a query.
Trading I/O for CPU might enhance the performance of many workloads on an
See “Comparing the Number of I/O Operations Required to Answer the Same SELECT Request” on page 313 for more information about I/O operations and column‑partitioned tables.
While there are cases where CPU usage decreases for queries made on a column‑partitioned table, CPU usage can also increase for some functions such as INSERT operations undertaken on a column‑partitioned table.
For a CPU bound system, column partitioning might not provide any benefit, and might even degrade performance. An exception is the case where a subset of the workload that is I/O bound, even if overall the system is CPU bound, in which case column partitioning could be beneficial. Experiment with running your CPU‑bound workloads against both nonpartitioned tables and column‑partitioned tables to determine what the differences are.