The single biggest disadvantage of recording with native audio programs is latency. Latency is the time that the audio has to be worked on in the computer. When recording, a musician needs to hear the music played back so he or she can play the right part or parts. Latency is the time that it takes a signal going into the card to come back out. In other words, the analog signal getting to the soundcard has to be converted to digital, stored in main memory, written to disk, mixed together with any audio playing back, then get converted back to analog and head out the sound card again. Using old school analog mixers and tape decks, the latency was microscopic, essentially a function of the speed of the electrons going through the wire and the length of the wires. In practice, there is no latency with analog gear.

Not so when working with digital audio. Depending on the driver and card, latency can be anywhere from under 1 millisecond to over 750 milliseconds. Generally, cards and drivers that use the host system for most of the work will have latencies somewhere between 30ms and 500ms. Recently, host based systems with fast CPUs and well tuned drivers have brought latencies down under 10ms on Windows systems. In any case, anything above a few milliseconds is far too high to record and play back comfortably at the same time. Different cards and drivers try to get around this problem in various ways, but all else being equal, get the card with the lowest latency you can.

Low latency is certainly the most important thing a driver can have, but there are other things to consider. Some drivers allow more than one application to send audio streams to the card at the same time, while others "take over" the card entirely and will not share the outputs. Synchronization in many cases is also very important for sound card drivers. Sync is an issue when you need to have more than one application playing back at the same time, or even more than one computer with applications playing back in time. In my own studio, for instance, I use at least two computers when I record, one for audio playback, and one for MIDI sequencing. In order for this to work, obviously, the two computers need to agree on where exactly they are in a song. Different driver formats offer greater or lesser sync capability.

Finally, some drivers allow a host application, usually your sequencer, to control specialized capabilities on the sound card-things like DSP processors, synthesizer, sample rate converters, and such like.

One more thing before we look at the five main driver types. When we call them drivers, that's not totally accurate. A driver is a .sys or .vxd file that talks to hardware and/or the OS. What we are talking about here are really APIs. However, the names are used interchangeably, for better or for worse, so I will follow that convention.