Exactly how does an add-on IDE controller card work? i.e. the Maxtor/Promise controllers. Does the card actually create two new IDE channels or does it somehow "borrow" resources from the original two channels. In other words, are the new channels as efficient and good as if they were built in like the primary and secondary? And what limitations do the channels have? For example, the IDE controller can only access one device on the channel at a time. So therefore your fastest performance between IDE devices comes from cross-controller data transfers. If you have IDE3 and IDE4 controllers, does that mean IDE2 to IDE4 is just as fast as IDE1 to IDE2? Or is IDE4 kind of a "virtual" controller that pulls from the IDE2?