I heard about a hack I never tried, myself, where you buy an oversized hard drive and only use the beginning of it. Hard drives store data from the outside in, so you can improve latency by minimizing arm travel, and improve throughput by storing data where the linear velocity is faster. No idea how big of a difference this would make.
However, latency is about both the latency of the arm movement and the rotational latency. The latter you can't overcome by short-stroking.
However, short-stroking does point to an interesting problem when hard drives were still relevant.
If you didn't size your storage right, you ran out of IOPs long before you ran out of capacity.
For instance, if you run VMs on your SAN / storage array, you may have to stop putting VMs on the storage long before capacity is reached.
It's actually kind of funny that "just using the beginning 25%" of the drive is completely unnecessary provided you've got your partitions set up properly- you just create a partition at the beginning of the disk that has the data you want to access the quickest, and use the end of the disk for data that you don't access coincident with the data you want to get the fastest (a perfect place to put a dual-booted Linux install, for instance).