Not sure that you realize how many sequences are in UniProtKB. I don't think there is a way of doing this in a reasonable amount of time by classic sequence comparisons.
One way I can think of is to extract all sequence IDs, and find which ones match between the two databases. Everything that doesn't match are different entries. This assumes that a sequence ID from the two database contains the same sequence, which may or may not be true.
Yet another way is to calculate some kind of a hash or cyclic redundancy check for each sequence, and compare those instead of sequences. In this case matching CRC values does indicate identical sequences, but this takes longer to calculate than a simple ID comparison suggested above. I suggest you check out CRC64 from this page: