We will look deeper into the applications to discover why they cannot exploit the communication - computation energy gap. To perform this analysis, we rely on empirical observations on the Skiff platform as well as the execution-driven simulator known as SimpleScalar [7]. Though SimpleScalar is inherently an out-of-order, superscalar simulator, it has been modified to read statically linked ARM binaries and model the five-stage, in-order pipeline of the SA-110x [2]. As SimpleScalar is beta software we will handle the statistics it reports with caution, using them to explain the traits of the compression applications rather than to describe their precise execution on a Skiff. Namely, high instruction counts and high cost of memory access lead to poor energy efficiency.