Internal drives: a performance plateau?
Exactly how fast have hard drives gotten? Let's put things in perspective and look at a typical hard drive. One of Maxtor's low-end models, which the company sells to a personal-video-recorder manufacturer, handles up to five MPEG-2 video streams simultaneously. That's a lot of data streaming on and off the disk at once-- altogether, roughly 6.5MB per second continuously. There's a clear sign that you're unlikely to hit the performance wall of any current hard drive even if you make a hobby of video or music editing.
Not that all risk in buying a hard drive has been eliminated. It's still possible to spend too much money, particularly if you think you know something about hard-disk technology. And if you're shopping or an external, portable drive, you need to check the specs carefully to avoid getting stuck with a noticeably slow drive.
We subjected all drives to WinBench's application-based WinMark disk benchmarks, as well as the low-level HD Tach, primarily as a cross-check. These are the same tests we use to test hard drives for reviews. We ran WinBench's business disk WinMark, high-end disk WinMark, disk-transfer-rate, disk-access-time, and disk-CPU-utilization tests, all of which isolate performance of the hard disk subsystem. The HD Tach benchmark is designed to test a PC's hard disk subsystem. We ran HD Tach's read-burst-speed, read-speed (maximum), and write-speed (maximum) tests, which evaluate the full range of performance of a hard disk. As if that wasn't enough testing, we also performed hands-on copying, file-loading, and media-file-conversion tests to see if we could "feel" speed differences or measure them with a stopwatch. For internal desktop drives, here's what our tests reveal about the specs that do and do not matter.
Spindle speed: Also called rotational speed; expressed in rotations per minute (rpm), this number indicates how fast a hard drive's platters spin. Of all hard-drive specs, this one is the best predictor of actual performance. To see the effects, we installed 7,200rpm and 5,400rpm EIDE drives from Maxtor in identical 2GHz Dell Pentium 4 computers--and later threw in one of the fastest drives available, a 15,000rpm SCSI model from Seagate. And guess what? Although the WinMark disk tests indicated that the 7,200rpm drive was between 17 and 25 percent faster than the 5,400rpm drive, we observed no difference whatsoever in our hands-on, side-by- side tests. The 15,000rpm drive did seem snappier than the other two, but the observable difference was nowhere near the margin suggested by the benchmark scores (which made the drive seem as much as 60 percent faster). Certainly, the extra kick can't justify the 15,000rpm drive's price, which is roughly four times that of the 7,200rpm EIDE drive.
Seek time: This relatively meaningless number describes how long it takes, on average, for a hard drive's read/write heads to find a random track. A better number-- which you won't always find listed--is the access time, which takes into account the latency incurred by the head waiting for the platter to spin before a track can be accessed. A drive's access time may help predict performance in server applications, where simultaneous calls from multiple users cause the read/write heads to fly all over the platters. But on the desktop? You'll never spot the difference.
Cache size: Every hard drive has a chunk of memory to hold frequently accessed data--usually 2MB worth. To determine the effects of a larger cache, we couldn't resist testing the new Western Digital Caviar 120GB Special Edition, which comes with an 8MB buffer. The conclusion? You guessed it: no discernible boost over similar 2MB drives in desktop applications.
