Redefining Mobile

Mobile devices traditionally have been defined by their power source. In the future, they will likely be defined by where data is processed, which has enormous implications for designs.

The good news for many of these devices is that the work on battery-power devices will pave the way for greater efficiency in those with a plug, allowing much more intelligent decisions about tradeoffs between performance and power. The bad news is that designers of other equipment now have to start wrestling with the same kinds of issues developers of smartphones and notebook computers have been dealing with for several process nodes.

The impetus for these kinds of changes varies by market. In data centers, for example, virtualization has moved from increasing utilization on a single server to utilization across the data center as a whole. The net effect is that entire sections of the data center can be turned off when they aren’t needed, which feeds into the internal cloud concept. That reduces the amount of electricity needed to power and cool these servers. But taken as a whole, that also means racks of servers need to be turned on and off the way portions of an SoC are turned on and off. And while the devices themselves aren’t actually mobile, the data that moves across them is. Energy consumption is a function of the processing, not of the individual machines, and that processing is extremely mobile.

The same concept applies to mesh or grid processing, such as a sensor network on a bridge or road. Most of the time, sensors are in deep sleep mode, but they can be woken up as data moves across these devices and put back to sleep as data stops moving. It can apply to grids of solar panels, as well, which may change in effectiveness as the sun moves and electricity is no longer being harnessed.

What also becomes important in this scenario is the I/O piece of the design. Offloading or importing data requires bandwidth and power to drive it, and more data requires more bandwidth and more power. As data becomes the focal point for what’s fixed and what’s mobile, power will need to adjust to the flow of data across a wide range of devices. The center of the universe is moving from the chip to something far less tangible, where the chip is only one small component in a larger architecture with its own power budget. And if you thought bridging the gap between hardware and software teams in the same company was difficult, this challenge is looking orders of magnitude more difficult.