Here's what Intel has to say about hard drive burst speeds:
http://www.intel.com/support/graphics/sb/CS-010481.htm
Sustained vs. Burst Transfer Rates
Sustained transfers refer to a continued transfer that does not occur from the drive cache. Burst rates refer to data transferred directly to/from the high speed cache. A true indicator of performance is sustained rate; however, most drives are advertised with their faster burst rate.
A typical ATA/100 hard drive bursts at about 100MB/sec from the cache, but has a sustained rate of about 26-42MB/sec, depending on the drive. If you consider a best-case scenario with the 2MB cache full of data, 100MB/sec will quickly deplete the cache and commence transferring at the lower sustained rate.
ATA/66 hard drives are typically less expensive and do not saturate the 66MB/sec bandwidth available. In fact, a hard drive with a sustained transfer rate of 26MB/sec will not even saturate available ATA/33 bandwidth.
In summary, sustained transfer rates should be considered over burst transfer rates for maximum hard drive performance.
The average read (sustained transfer) rate is more important, especially when dealing with larger files. Generally, larger cache memory sizes will give you higher burst transfer speeds. This is one of the reasons why the older Raptors may lose out in burst speed comparisons. 1st gen 36GB Raptors had 8MB caches in addition to their SATA 1.5Gbit/s interfaces. Later ones upped the cache to 16MB, but the SATA interface is unchanged, so compared to other 16MB cache drives with SATA 3.0Gbit/s interfaces they still have a disadvantage. But, as stated above, burst speeds aren't the most important spec to look at.
See the random access numbers? That's due to the fact that the Raptors use a 10,000 rpm rotational speed. Raptors can find data on their platters quicker due to that advantage. Their downside is their cost per GB, not to mention they top out at 150GB in capacity when 1TB (Terabyte or 1000GB) 7,200 rpm drives are available.
Now, here's a chart comparing average transfer (read) rates:
http://www23.tomshardware.com/storage.h ... 0&chart=34
You may also wonder why the 10,000 rpm Raptor's average read speeds aren't much better than a 7,200 rpm drive's. If the platter is moving faster under the heads, data should be being read off the drive faster also, right? The reason is partly due to lowlier drive having higher density platters. Since these high capacity drives achieve their gains by packing bits closer together, the same segment of platter area of a 1TB drive contains much more data than the equivalent segment of a Raptor's platter. So why doesn't Western Digital combine these high capacity platters and 10,000 rpm motors for the best of both worlds? I don't know (although I can speculate).
If cost isn't an object, Raptors in RAID 0 is still the best performing setup overall. Personally, I use a RAID 0 setup with 2
WD Caviar SE16 WD2500KS drives for my data storage (which is most of my video files).
BTW, the MaximumPC article used WD Raptor drives in all the tests.