Pay 100 dollars for a cheap 40GB MLC SSD with questionable wear leveling implementation and having equal rewrite rates as a harddrive[1] in small random reads due to bad controller implementation that does not properly remap sectors? Sure, that is totally a bargin.
As for the 100GB/7.2k vs 250GB/5.4k, this depends on the drive manufacturer, specific model and interface. However for the general stuff:
If we assume both drives have proper engineering and come from the same generation with an addition of possibly using some related components/platform...
100GB 7.2k will mean you'll have less rotational latency.
250GB 5.4k might mean you'll see higher sequential rates due to higher aerial density IF and ONLY IF both drives have the same platter count
This becomes a challenge once you mess with how many platters are in the drive and what their density is (Three platters in the 250GB? One in the 100GB?)
As for rotational latency, harddrives do not necessarily read all the data on the first pass nor are they able to get onto the track within the first pass. The head might not be able to get there in time before it needs to wait for the next rotation, or it might have issues servoing onto the track due to bad SNR.
It'd help if I knew more about the harddrives so I could consult with an engineer friend.
[1]A similar problem
faces flash-based solid-state disks (SSDs). SSDs are
organized into “pages” that are erased before being
written, and erase is relatively slow. While small
random reads of SSDs are about 100 times faster than
today’s magnetic disks, simple SSDs (ie., Memoright,
MTron) that read-erase-write each modified page
deliver small random write performance no better than
magnetic disks,
http://repository.cmu.edu/cgi/viewconte ... ontext=pdl