Rob Knoth

"I'm on my way..."

Last week I wrote about the rise of performance per watt as the key metric for computing, and how parallel processing helped bring this to the forefront for designers.  This week, I examine the other driving force behind performance per watt: mobile.

Since Star Trek and Dick Tracy, the goal of technology is freedom and mobility.  At the same time that processing power began to plateau for single cores, mobile devices began to come into their own.  This trend of focusing on consuming (and now creating) media on the go has become the driving force in technology.  Wireless data rates have long since eclipsed what old dialups could offer.  Video is now the common language of mobile platforms, and all of that demands massive processing power without limiting the battery life of the device.   Over the past several years, desktop growth has remained stagnant, while mobile has grown year-on-year.

That twin need of powerful computing and long uptime requires low power design to be present throughout the design of a system.  Everything from the display to the processor must look at how much power is consumed relative to the end user’s benefit.  Techniques in semiconductor design that used to be esoteric (shutting off sections that are not in use) have become mainstream.  New techniques like Dynamic Voltage and Frequency Scaling (DVFS) are becoming even more popular to keep the focus on maximizing performance per watt.

Combining together powerful mobile processors, like the Snapdragon or Atom, with programming and system techniques like parallel processing, the impact of computing on power consumption can be limited.  The same kind of cost-benefit analysis needs to ripple throughout our world to help technology continue to deliver life changing impact without taxing our already strained system.

(For more indepth analysis, please see the briefing on microprocessors in this months’ MIT TechReview.)