Don't forget, just because you have short-stroked a drive, does not mean you can't add the unused portion of the drive as a seperate partition.
You're right that the drive could have been partitioned to achieve the same effect... but the way the reviewer "shorted" the drives was by lowering their LBA count in the drive settings. To the controller and OS, the drive now appears smaller and any extra capacity the drive had is now inaccessible.
Handy is also correct that if partitions are used to "short" a drive, both partitions should not be accessed simultaneously in order to avoid contention.
One major flaw in the review's testing is that the tests measured the entire (logical) drive surface. Testing 20GB on the 20GB drive, 44GB on the 44GB drive and 450GB on the 450GB drive. This basically assumes that you will fill your entire drive to 100% capacity - which is unrealistic.
A better comparison to gain insight into performance differences between the different setups would be fill each drive with a set amount of data. Say a 20GB on all three drives. I'm guessing the performance differences would be negligible. Though you might find that 20GB on an actual 20GB drive is very slow and 20GB on a 44GB drive is acceptable, and 20GB on a 450GB is very fast. However, the performance gain is not related to short stroking the drive, but rather in having better data consolidation/using a lower percentage of the drive surface.
The end conclusion that one could then draw is that for best performance, one should not exceed more than x% of a drive's usable capacity. Where x is likely 10%. There is no need to artificially cap the drive at this limit, just realize that performance will degrade past this point. Through further testing one could find that performance suffers considerably at y% of the drive's capacity. From my experience, y would likely be 50%.
