A useful performance test should produce a result that can be compared with previous runs. That means the environment, inputs, build artifacts, and reporting path all need to be controlled closely enough that a result says something about the change under test.
Automation helps by making the normal path repeatable. It can also capture context that is easy to miss in manual testing: versions, configuration, hardware, workload parameters, and result metadata.
The goal is not only to run tests faster. The goal is to make the result easier to trust and easier to investigate when it changes.