Well, the short answer is that it's still sorting out in terms of which actual form factors will prevail. What is perhaps more significant than the actual form factors (the dimensions of specific boards) is now the connectors, which are becoming independent of form factors. This could not have happened with bus-oriented modules, which had to have specific backplane connectors and the same dimensions to fit into a rack.

One of the perhaps unexpected results is the blurring of what constitutes a COM module vis a vis a single board computer (SBC). The idea of a COM module is that you can design a custom I/O module independent of the compute engine and then upgrade the CPU as newer versions become available without having to redesign or re-implement the I/O. However, recently COM modules have appeared with additional I/O connectors on board and SBCs have shown up with the newer connectors like the SUMIT connectors on them. These are symptoms of the ongoing effort to optimize the divide between CPU and I/O for the highest performance at lowest cost.

When these issues get worked out, there will be a winnowing of the various form factor/connector combinations. But the further implication is the effect of the low-power processors. By low power, I mean they draw very low current and dissipate very little heat, but have quite high 32-bit computing performance. This increases what I could refer to as the embedded "granularity." That means that with their small size, low power, high performance and network connectivity, they can be embedded in ever smaller, tighter spaces than were ever possible. This opens up a vast array of possible new applications that can take advantage of embedded intelligence.

Submitted by Tom Williams on May 07, 2009