You can also nominally double the amount of space in the fast/outer zone with a drive that's twice as large (not quite the same, but I'd be surprised if you could tell the difference unless you have very large partitions).
Partitioning is as effective as firmware-based short-stroking, unless the drive firmware has has some very peculiar properties. The objective is to reduce the distance the heads must travel to access a given amount of information. In general, the drive's not going to wander over the rest of the disk unless told to (unless, e.g., the controller or drive firmware is set for patrol-read, but even that is likely to be noise in most cases).
Decreasing the stripe size so more files are split across disks is of dubious benefit, and likely counter-productive. In most cases, you've simply slowed down the completion of the IO until all disks complete seek+xfer.
E.g., given a single reasonably fast drive with a 7ms average seek time, an average xfer rate of 50MBs, and an xfer size of 64KB, the total IO time = 7ms (seek) + 1.28ms (xfer) = 8.28ms. If you split that across two drives, and if the two drives are not with 0.64ms seek time of each other, you lose--that is, if the second drive doesn't complete its seek and start its xfer within 0.64ms of when the first drive started its xfer. Decreasing the stripe size and increasing the number of drives can exacerbate the situation.
Ensuring drives obtain that level of coordination is difficult, and comes with a price (and not what you're going to see with typical mobo RAID). Which is why larger stripe sizes are generally preferred--unless you have a very random IO pattern*--as seek time is orders of magnitude greater than xfer time for typical IOs.
* Edit: and the size of the IO is typically the same as the stripe size, so you really do get the effective benefit of multiple independent heads/IOs.