Redefining A System…And Why It Matters

Methodologies will be altered radically as we move into stacked die and the IoT.


The concept of a system on chip has been around since the mid-1990s, but the actual mass implementation of this scheme is almost synonymous with the rise of the smart phone over the past decade. Put in perspective, it isn’t that old, and it’s about to change.

Prior to the SoC, an electronic system was largely a collection of components on a PCB that included memory, a processor, and some sort of I/O and networking. Putting it all together on a single die was an engineering feat, but it was one that was based on the components and approaches that were already well defined. As one industry expert observed, if Nokia remained the market leader in cell phones we would have had a smaller phone.

Instead, we have multi-function smart phones with more performance and capabilities than most PCs had a decade ago. But the PC era is over. We’re well into the smart phone era. And what comes next may not even be tied to any single device. It may be more about following the data, as Cadencefellow Chris Rowen termed it in a speech last year, than the electronics themselves. When that finally happens, it would mark the single biggest shift in the history of computing.

The Internet of Things concept—an incredibly bad description for what’s about to take place—will revolutionize how we think of a system. Where exactly do your songs live in iTunes, for example? Do you really own Waze? And where is your social media data really being kept?

This has a direct bearing on the electronics you choose to access this data. It doesn’t matter if your on-chip processor can run at 3 GHz when all you’re doing is listening to music that’s being stored in the cloud. It also doesn’t matter if you have two cores of 16 if all you’re doing is streaming a video.

What’s changing here is as fundamental as the creation of the SoC in the first place. We are migrating, in the words of SonicsCTO Drew Wingard, from a processor-defined world to one that is application-focused. This has ramifications for everything from semiconductor design to EDA tools, and it opens the door to many more players and ideas because the fundamental building blocks will be platforms, derivatives, and more standardized approaches to connecting everything together more quickly.

We are moving into a world where the next big thing is how you slice up existing platforms and technologies into customized little things—basically using aggregation of standardize parts to specialize in unique ways. And all of this has to be done efficiently, meaning quickly, with minimal added cost, and with multiple options that can fit a particular slice of a market. Some may require extremely low power, while in others the emphasis will be on performance or the addition of very unique sensors.

Add all of this up and it presents one of the biggest opportunities ever for semiconductors and EDA. They will be the engines and the building blocks for these changes, driving economies of scale and innovation well beyond their own narrowly defined physical world into a much broader universe of interconnected systems where data defines the system.