[StillBored]

I've done this kind of stuff, and between OS buffering, and making sure the ripping software is writing large blocks (say 4-32MB at a time) its possible to run drives at basically full bandwidth with something less than a dozen streams. There is going to be more inner/outer track bandwidth variation than the perf falloff going from 1 to 6 streams with large blocks (say 4-32M sequences). There are a lot of reasons for this, but a lot has to do with data placement effectively combining multiple streams into data writes to the same sequential track.

More interesting is that even "sequential" read/writes already have seek times built in because HD's aren't spiral track, so head switching, and track to track seek (and the associated rotational/finding the servo track) are inherent in sequential IO perf. So most filessytem placement/schedulers aren't going to place 3 files being written at the same time on opposite sides of a disk, so those head switch times and track times have nearly immeasurable increases because the drive itself is also storing a large part of a track write and moving 3 tracks and a head, is basically the same as just moving a head.