The above graph is of a scalable parallel computer with 128 nodes. It shows that the region of computation in a parallel computer is bounded by latency, speed, precision and memory. The left side of the curve is bounded by latency, the top of the curve is bounded by peak speed, and the right side of the curve is bouned by precision or memory. All the computation has to be done within this region. Clearly, we can increase this region by reducing latency, increasing the speed and increasing precision or memory of the computer.