Obviously, the interface's job is to allow for the expedient transfer of data between
the hard disk and the system. Since the hard disk is an important performance component in
the PC, the performance of the hard disk interface is probably the most frequently
examined aspect of an interface's overall quality. In fact, it is generally given far too much
emphasis in my opinion, particularly by companies looking to sell hardware based on the
speed of the interface.
I'm certainly not suggesting that the performance of the interface is not important.
However, it's important to keep it in perspective. The main problem when it comes to
looking at interface performance is that on modern systems, the speed of the interface is
not the limiting factor to overall hard disk performance. If a given hard disk can't read
data from its platters fast enough to saturate an interface of a given speed, going to a
faster interface yields improvement only on reads of data already in the drive's internal buffer--which makes virtually no difference in
overall, real-world performance. Despite this, companies try to claim that (for example)
drives using Ultra DMA/100 are "50% faster"
than those using Ultra DMA/66. In fact, at the time of this writing (late 2000) no IDE/ATA
drive can saturate even a 66 MB/s interface, so going to 100 MB/s is essentially
pointless. For a more complete discussion of this issue, see this page on the interface transfer rate
specification.
Actually, I have devoted an entire section of the site to discussing hard disk
performance, so you should look there for a lot more
detail on this important subject. Of particular interest will be the page I referenced
above on the interface transfer rate specification, as well as the general discussion of interface performance factors.
Next: Command Overhead
