the buffer pool contains pages, 16kb by default
it is effectively a read-and-write-through cache of these pages
if you read the page (in most cases) it will be added to the buffer pool
if you modify the page, it will be loaded into the buffer pool and then MODIFIED THERE
at some later point, that modified page will be flushed to disk
(if you don't access it for a while)
therefore, in general, pages you read a lot (hot pages) will be in the buffer pool unless your working / hot set is much larger than the buffer pool is
large tables are less efficiently stored in terms of full pages than smaller tables
but other than that, there is not much of a difference.
trust me, partitioning is not pointless
as I've already said, it allows you to prune on the read path
it decreases the size (and thereby improves the performance of) btrees
it allows you to drop chunks of the table in a DDL operation (DROP PARTITION) instead of DML (DELETE TONS OF ROWS)
it is mostly orthogonal to the behavior of the buffer pool you're talking about
but that doesn't make it "pointless"