CBR 320 kbps MP3 keeps the bitrate constant. VBR tries to keep the quality constant. It can go up to the 320 kbps bitrate when needed, but most of the time the same quality doesn't need that much bitrate. VBR Extreme is the best VBR MP3 quality in MC and also in general. The general consensus is that the LAME encoder is the best MP3 encoder, especially at high bitrates. MC uses LAME. The MP3 encoder in iTunes is not usually found to be as high quality. Apple has given priority to its own AAC format.
I would say that LAME CBR 320 kbps and iTunes AAC 320 kbps are on par. LAME VBR extreme is as good most of the time (99,9% or so).
Sometimes VBR encoding can make wrong choices on difficult passages. I read this thread at Hydrogen Audio forums:
http://www.hydrogenaudio.org/forums/index.php?showtopic=37003. After quickly testing the sample file and hearing the obvious encoding error I decided to encode the sample with several lossy encoders using a few different settings and bitrates. I used LAME 3.90.3 (CBR and VBR), LAME 3.97 beta 1 (CBR and VBR), Vorbis aoTuv beta 4 (always VBR), Musepack 1.15v alpha (always VBR) and iTunes 5 (CBR and VBR).
Here are my 27 test files in a zip package (6.9 MB):
herding_calls_sample.zip. The original lossless file is available at
HA.
If you have a high quality sound system you could try to find differences. With lesser systems it might be impossible to distinguish them. Each sample is 11 s long, so it is easy to make a playlist of them and let MC play through the list. You could select random play and close your eyes. Keep in mind that this is only a rare extreme sample. Most of the time finding differences is more difficult.
By the way, Apple added AAC VBR to iTunes 5. However it is not a real VBR. It works like LAME ABR (average bitrate). iTunes AAC VBR doesn't change the average bitrate according to the complexity of the encoded tracks.