When Oracle hash-joins two data sets, the join-columns of the (typically?) smaller data set is first hashed and stored (in memory or on disk).
Then, Oracle iterates over each record of the larger result set and also hashes the each of the join-columns and looks that hash up in the stored away hashed set.
Hashing the smaller set first has three benefits:
Because a write operation is involved, less data needs to be written
The smaller size increases the likelyhood that the hashed data fits into memory
The probablity of a hash collision is smaller
A hash join is especially performant if the the smaller input set is smaller than n (for example n < 25000).