Most people are running these at 4 bits per parameter for speed and RAM reasons. That means the model would take just about all of the RAM. But instead of swap

Most people are running these at 4 bits per parameter for speed and RAM reasons. That means the model would take just about all of the RAM. But instead of swap (writing data to disk and then reading it again later), I would expect a good implementation to only run into cache eviction (deleting data from RAM and then reading it back from disk later), which should be a lot faster and cause less wear and tear on SSDs.