Abstract: Ray casting is the bottleneck of many rendering algorithms. Although much work has been done on making ray casting more efficient, most published work is high level. This paper discusses efficiency at a slightly lower level, presenting optimizations for bounding-volume hierarchies that many people use but are rarely described in the literature. A set of guidelines for optimization are presented that avoid some of the common pitfalls. Finally, the effects of the optimizations are shown for a set of models.