Not to spark endless debate, but that's an extremely narrow view of things. One would hope that new level and model data, dialog, textures, etc. would also be seamlessly streamed from the optical media to a combination of main memory and/or hard disk cache. And depending on the number of simultaneous channels in the audio engine, dozens to hundreds of audio samples are being read, manipulated, and mixed into the music score as well, which could consume significant bandwidth and maintain an unnecessarily large memory footprint if uncompressed. Game audio over the last couple generations has transformed from synthesized sounds overlaid onto static CD background music, to fully dynamic environments of ambient sounds, sampled effects and speech, and orchestral scores all of which blend and morph to match the players in-game situations, and shouldn't be compared to a pre-canned multi-channel movie audio track.