it is kind of odd simply because we're so used to multiple of 2. i do wonder myself if there is some kind of manufacturing reason to do it. maybe the layering does not lend itself to multiples of 2 so it goes with a different multiple???
There's no real reason why manufacturers need to make memory modules in powers of 2. It may simply be for cost reasons or there's some other application for it that doesn't allow for as much memory density.
I noticed a lot of the 24/48/96 Gbit memory chips at a supplier were for automotive applications, suggesting they're made to different tolerances in some way.