10.2 CD-ROM Drive Performance

Although CD-ROM drives differ in reliability, the standards they support, and numerous other respects, the most important issue for most buyers is performance. But performance is not accurately described by the simple transfer rate number that most manufacturers use to characterize drive performance. There are actually two important performance metrics:

Data transfer rate

How fast the drive delivers sequential data to the interface. Data transfer rate (DTR) is determined by drive rotation speed, and is rated by a number followed by an X. All other things equal, a 32X drive delivers data at twice the speed of a 16X drive. But note that we have used 16X SCSI drives that transfer data faster than some 32X ATAPI drives, so it's a mistake to depend solely on X-rating. The specifications for some drives list only maximum burst transfer rate, which is always the advertised number, while others list sustained transfer rate, which is far more important to overall drive performance. Fast DTR is most important when you use the drive to transfer large files or many sequential smaller files, e.g., for gaming video.

Average access

How fast the drive accesses random files located anywhere on the CD. Average access time is only loosely tied to DTR, is determined by the quality of the head actuator mechanism, and is rated in milliseconds (ms). Some inexpensive drives have very high nominal DTR ratings but relatively poor average access performance. To make matters more complicated, different manufacturers calculate average access using different methods, so you cannot necessarily compare figures from one manufacturer with those of another.

The following sections describe data transfer rate and average access in detail.

10.2.1 Data Transfer Rates

CD-DA discs record music as a digital data stream. The analog music is examined, or sampled, 44,100 times per second (the sampling rate) using 16-bit samples, for a data rate of 88,200 bytes/s. Multiplied by two channels for stereo, that means the CD stores 176,400 bytes for each second of music. Each second's data is stored as 75 physical sectors, each of 2,352 bytes.

Music data need not be completely error free because an occasional flipped bit will be inaudible, which means that the entire 2,352-byte capacity of each physical sector can be used to store actual music data. The same is not true for computer data, for which every bit must be correct. Accordingly, CDs store computer data using 2,048 bytes/sector, with the remaining 304 bytes in each physical sector allocated to error detection and correction data. This means that a computer CD running at the same speed as an audio CD delivers (75 x 2,048 bytes) per second, or 150 KB/s. This data rate is called 1X, and was the transfer rate supported by early CD-ROM drives.

Later CD-ROM drives transfer data at some integer multiple of this basic 150 KB/s 1X rate. Rather than designating drives by actual KB/s throughput, drive manufacturers use a multiple of the standard 1X rate. A 2X drive transfers (2 x 150 KB/s) or 300 KB/s, a 40X drive transfers data at 6,000 KB/s, and so on.

10.2.2 CLV Versus CAV Versus P-CAV

Unlike hard disks, which record data on a series of concentric tracks, CDs have only one track that spirals from the center of the CD out to the edge, much like a vinyl LP. Because the portions of the track toward the center are shorter than those toward the edge, moving data under the head at a constant rate requires spinning the disc faster as the head moves from the center—where there is less data per revolution—to the edge, where there is more. If an audio CD spun at some compromise constant rate, the audio would sound like the Addams Family's Lurch when the CD was playing the inner portion of the track, and like Alvin the chipmunk when it was playing the outer.

The solution to this was to change the disc rotation rate as the heads progressed from the inner to the outer portions of the track. When you play an audio CD in a CD player (or in your computer's CD-ROM drive), the drive actually speeds up and slows down according to where the heads are currently reading. This technology is called Constant Linear Velocity (CLV).

All audio CD player drives use CLV. CLV is a good choice for audio for two reasons. First, the drive only need spin fast enough to deliver 150 KB/s. Second, music is inherently sequential. That is, music data is typically played from first to last, which requires only gradual changes in rotation speed. Early CD-ROM drives also used CLV, but it soon became apparent that CLV was not the best choice for data CDs, for two reasons. First, market demands meant that the speed of CD-ROM drives had to keep increasing, from 1X to 2X, 4X, 6X, 8X, and ultimately to 12X or 16X. Delivering data at 16X speeds requires spinning the disc much faster than for 1X audio. Second, data CDs are often accessed randomly, which means that the head may have to move quickly from the inner to the outer portion of the track, or vice versa. In order to maintain CLV with such rapid head moves, the drive motor was required to make radical changes in speed. Motors capable of doing this were large, power-hungry, loud, and expensive. That meant that 12X was the realistic limit for CLV CD-ROM drives, although a very few 16X CLV CD-ROM drives were made.

The solution to this problem was to implement Constant Angular Velocity (CAV), which is a fancy term for simply spinning the CD at a constant speed, allowing the data rate to vary according to which portion of the track is being read. The advantage to CAV is that the drive can use a relatively simple, inexpensive motor because that motor runs at a constant rate. The disadvantage is that the data rate varies according to what portion of the disc is being read, which is really no disadvantage at all for data CDs. Actually, CAV drives are also capable of running in CLV mode, which is why you can play an audio CD in any CD-ROM drive. But that's slow CLV. For delivering data, which is their true purpose, CAV drives run at a much higher, but constant speed.

Because the data rate on a CAV drive varies according to which portion of the track is being read, there's no single number that describes the drive's transfer rate. Accordingly, such drives are called variable speed or *-Max (as in "48X-Max") drives, and are rated using the fastest data transfer rate (that on the outer portion of the track). When this was written, 48X-Max CD-ROM drives were commonplace, and 72X drives were available. The upshot is that a 40X-Max drive may read the longer, outer portions of the track at 40X, but may read the shorter, inner portions of the track at only 17X, with an "average" speed for a full CD of 27X, and a somewhat lower rate for a partially full CD.

Finally, some drives use Partial CAV (P-CAV). A P-CAV drive, rather than using a constant rotation rate, changes between some small number of discrete rotation rates (typically two) according to which arbitrarily defined zone of the CD the head is currently reading. That is, a P-CAV drive may rotate at one rate when reading the outer half of the track, and at some higher rate when reading the inner half. The advantage of P-CAV drives is that they deliver data at a more nearly constant transfer rate than CAV drives.

10.2.3 TrueX Drives

TrueX drives are CLV drives with a difference. Conventional CD-ROM drives read data with one low-powered laser beam. TrueX drives use a diffractor to split a higher-powered laser beam into seven separate beams, which are directed at seven separate parts of the track simultaneously. Using a multi-beam detector array, a TrueX drive reads those seven signals and combines them into one high-speed data stream. The result is that a TrueX drive running at 9.5X CLV provides a data transfer rate equivalent to a conventional 52X drive. The leisurely 9.5X spin rate of a TrueX drive means that it is quiet in comparison to high-speed CAV drives, which spin at very high rates and are correspondingly loud.

There are some drawbacks to TrueX drives, however. First, like any CLV drive, a TrueX drive needs a robust (and expensive) drive motor to support the constant changes in rotation rate required by a CLV drive. Second, CLV drives typically have relatively poor random access performance, because the drive motor must constantly adjust and stabilize the spin rate as the head seeks across the disk. Third, during heavy random seeks, TrueX drives vibrate noticeably, which makes us wonder how durable the drive will prove in heavy-use environments. Fourth, the higher-powered laser used by TrueX drives can cause discs and the drive itself to become very warm during sustained operations. We suspect that the higher operating temperature of TrueX drives may have a deleterious effect on media, particularly CD-R discs, although we have no evidence to support that suspicion.

TrueX drives are based on technology developed by Zen Research (http://www.zenresearch.com) and are manufactured exclusively by Kenwood (http://www.kenwoodtech.com). In addition to a 52X SCSI unit, Kenwood also manufactures ATAPI units running at 42X, 52X, 62X, and 72X. Although TrueX drives are still sold, their high cost and the advent of fast DVD-ROM drives meant they never achieved significant market share. We recommend avoiding them.

10.2.4 Average Access

Although it bears superficial resemblance to the hard drive rating with the same name, average access for a CLV CD-ROM drive is much more complex to calculate, and is subject to manipulation by drive manufacturers who wish to boost their performance figures. Hard drives spin at a constant rate, and average access is calculated as average seek time (the time required for the heads to move over the proper track) plus latency (the time required for the disk to spin the one half revolution required on average to move the correct sector under the heads).

Average access for CLV CD-ROM drives was originally calculated using a similar 1/3 stroke method, assuming that the drive would be used mainly for reading large multimedia files. In about 1993, some manufacturers began substituting "random access" for 1/3 stroke testing. This method was subject to abuse, because manufacturers could define the size of the zone they used for testing. Some chose very small zones to boost their average access ratings, with the result that some drives were advertised with average access times under 60 ms. Worse still, some makers began promoting seek time as a performance measure. Seek time is a useless performance measure for a CLV drive, because it ignores the fact that a CLV drive needs to speed up or slow down the disc to the speed required for data to be read. The time required for this step—roughly analogous to latency, but subject to much wider variation—is determined by the quality of the motor used.

Newer drives, which use CAV, are less subject to these manipulation methods because the disc spins at a constant rate. However, manufacturers are still free to define the zone they use for random access testing, which means that you cannot safely compare different drives unless you know the testing method used to rate them. The upshot is that if you have an older CLV drive that has a very good average access rating, you should suspect that it is artificially inflated. A newer drive, even one with a substantially slower rated average access, will likely outperform the older drive by a significant margin. As a point of reference, the fastest CD-ROM drive currently on the market, the Plextor UltraPlex Wide, is rated at 85 ms average access. Testing that drive against inexpensive ATAPI drives that have a faster published average access reveals that the Plextor is in fact much faster at retrieving random data from a CD. Therefore, take any published average access rating with a grain of salt and do your own testing if possible.