Why is adding partition_on slower?

[u3Izx9ql7vW4]

I have a basic query that I'm running on a table stored in Timescaledb, and I noticed that it runs slower when I add partition_on argument. The column I'm partitioning on is a unix timestamp (type bigint). It takes 5s to load the dataframe without partition_on, it takes 7s to load with partition_num=2, and 13s to load with partition_num=10.

select *
from mytable
where
close_time > now() - interval ‘500 day’ and
symbol = ‘abc’

Are there any other performance boosts? The database I'm hosting is local, and inside a Docker container. pd.read_csv takes 0.5s, which is substantially faster. I don't expect this type of performance from connectorX but 5s+ is also too long.