As with every component of the PC, the hard disk is but one part of an integrated
whole. It is not possible to measure the performance of the hard disk in isolation, since
running any benchmark program involves using the processor, memory and other parts of the
PC. The only way to isolate the hard disk would be if you were to use specialized test
hardware, connected directly to the hard disk itself, and then you'd have a hard time
being sure that the results really related to "real world" PC performance at
all.
Many benchmarks are designed to try to isolate the
hard disk from the rest of the system to test "only" the hard disk. Some are
more successful than others in doing this. Unfortunately, many of them don't take all the
factors into account and end up (for example) testing the system's hard disk cache instead
of the hard disk itself. They are getting smarter over time, however, but still, virtually
every hard disk benchmark I have ever seen has allowed the rest of the system to impact
the number. You take the hard disk out of a Pentium II 300 PC and put it into a Pentium
III 600 system, run the benchmark again, and the score goes up.
As with other components, the best way to compare two hard disks is still a comparative
benchmark. Set up the system they are intended to be used in, test one with a benchmark
that represents the intended use, and then replace it with the second drive, retest and
compare. This eliminates much of the "background noise" that is associated with
absolute benchmark numbers. Most better review sites do exactly this, maintaining constant
"test benches" with hardware that does not change between tests of various
drives.
Another thing that the matter of "component vs. system" means is that there
is no way to (legitimately) compare directly two drives that run under different
interfaces or in different systems entirely. In particular, you can't directly compare the
speed of an IDE/ATA drive to that of a SCSI drive and be sure that you are measuring only
the drive. Some would rightly point out that the distinction isn't really that
important however, since the "end result" of system performance is really most
important.
As discussed here, hard disks have a
very substantial impact on the performance of the system as a whole.
Next: Ranking Performance Specifications and Factors
