There's a complication involved in caching write requests to the disk that
doesn't exist when doing reads: should the write be cached (put into the disk's buffer) or
forced to go through directly to the disk platters? If you don't cache the write you
effectively remove the buffering that is in place to isolate the system from the slow mechanics of the drive. If you do cache the write you improve
performance, but what happens to the write if the power is cut off to the disk before it
can be put out to the platters? This is not a simple question, and doesn't have a simple
answer; see here for a full discussion on the
subject.
Most people won't worry about the potential risks of write caching, especially when
they find out that it improves performance. 
That's fine; but you should try to at
least find your drive manufacturer's "write policy policy" when shopping for a
drive; it's a good thing to know. A lot of drives use write caching without explicitly
saying so.
It should be noted that write caching improves performance pretty much only in a random
write environment: writing small blocks to the disk. The cache size is small relative to
the size of the disk, so write caching won't improve performance much on a long sequential
write. (On a long sustained write the buffer will fill up, and thus force writes to the
platters to occur in order to provide space in the cache for later blocks in the
sequence.)
Next: Thermal Recalibration
