Not sure exactly what OP is referring to, but CSS-trees [1] are a classic example of cache-aware indexing structures that fetch entire pages into cache and arrange data so that most of the comparisons happen on cached data. In most cases, they significantly outperform binary trees. Masstree [2] is a more recent example of this.
[1] http://www.vldb.org/conf/1999/P7.pdf
[2] https://pdos.csail.mit.edu/papers/masstree:eurosys12.pdf