Rob Knoth

Raw performance is no longer acceptable in a world of pervasive computing and rising costs of energy

In government, the common saying is that “all politics are local”.  For low power semiconductors, it would be “all computing is mobile”.  The confluence of mobile computing and parallel processing is key to managing the growing demand for power from computing, which will have a huge impact on the energy needs of the planet.  The new watchword is “performance per watt”.

The building block of the internet, personal computing, etc… are processors.  In the past, processor design power houses like Intel and AMD focused on pure performance.   Processors were rated based on clock frequency.   As time went on, the demands on computing began to grow faster than what processor designers could squeeze out of their designs.   Software then advanced to take advantage of parallel processing.  This allowed slower, smaller processors to accomplish bigger tasks faster than a single high horsepower processor.  That was a huge shift in the industry.  The impact of parallel processing on the energy grid is tremendous.  The economy of scale from ganging up a large number of processors means that a small increase in power for one processor results in a major impact on the grid.  This is why low power design for the processors is so important.  Saving a few milliwatts adds up when you have thousands of processors active in a data centers.  Multiply that across the global data centers of Microsoft, Google, Cisco, … and you see the impact.

At the same time that parallel processing became popular, another phenomenon came on the scene: mobile computing.  I’ll continue next week on this.