This is how my key is set up:
pub 4096R/AAAAAAAA created: 2012-11-14 expires: 2013-11-14 usage: SC
trust: ultimate validity: ultimate
sub 4096R/BBBBBBBB created: 2012-11-14 expires: 2013-11-14 usage: E
[ultimate] (1). xxxx <xxxx@xxxx.com>
I know this is standard, to have the subkey handle decryption. Why is this though? How is this different from me just having a single key that does everything?
There are three main reasons why you may want to have separate keys:
Not all key types can be used for both encryption and signature. OpenPGP (the format that GnuPG implements) primarily supports RSA, ElGamal and DSA; ElGamal is encrypt-only, while DSA is sign-only. RSA can do both; however, there was a time when RSA was still patented, and since that time some PGP implementations have taken the habit of defaulting to ElGamal+DSA, hence necessarily two key pairs.
Encryption keys should be escrowed (i.e. have backup somewhere), while signature keys should not. See this answer for a discussion on this subject. Since a given key cannot be both escrowed and not escrowed, you should need two keys.
Ideally, you should have a single master key, that you keep in a specially well protected keyring, and subkeys for daily tasks. The master key is supposed to be your "root" key with no expiry date; it is large and bulky and possibly inefficient, but you don't use it often. The subkeys may be shorter (thus more efficient) because you define them to have a short lifetime (e.g. one or two years): you don't have to make them fat enough to survive advances in technology one or two decades from now. The master key doubles as "revocation key" in case one of your subkeys gets stolen.
