Thinking Much Bigger

Why system-level design is about to get a huge boost.


For the better part of the past decade the focus has been on integrating an increasing number of smaller components on a piece of silicon. It’s time to start thinking much bigger.

While there is still plenty of work to be done building more powerful processors, or networks of connected processors on a chip or in a package, new opportunities are opening up in markets such as automotive, medical, industrial control, the cloud, and even inside of data centers that can leverage the same skill sets. Instead of integrating everything onto an SoC, the next opportunities may be integrating systems of connected SoCs.

EDA vendors have been talking about this opportunity for years. But integrating racks of servers or electronic systems in a car have largely been closed off to most semiconductor tools vendors in the past. The design process and flows that had been in place for decades in those markets worked well enough while the focus was on performance, or where electronics played only a limited role such as in car infotainment systems. As most EDA companies have discovered, it’s difficult to sell tools into a market if something isn’t broken.

But as the amount of data increases by orders of magnitude, existing systems are breaking down. New architectures already are being proposed for memories, sensors, processors, and where computing should be done. That includes scattering processing throughout the network, throughout various types of memory, and moving data more quickly with silicon photonics. These are the same kinds of problems chipmakers have been wrestling with for years on a nano scale. The issues of power/heat, throughput and signal integrity are nothing new, and there are tools already in place that can improve the flow of data and the layout of processing elements within these mammoth data centers and cloud operations.

In the automotive and industrial markets, what’s broken is the old way of doing things because most of the electronics are brand new. The push toward assisted driving and autonomous vehicles has been a boon for EDA vendors, which are being called on for tools to design and verify chips and electronic systems and wiring harnesses. But this is just scratching the surface. As electronics become an integral part of vehicles, the layout of those systems, optimizing the movement of data, and ensuring the integrity of that data are the next step in this development effort.

For industrial operations, the challenge has been adding electronic controls into systems that may be 50 or 100 years old to improve uptime and yield. Again, these are data flow issues with constraints such as power and various types of noise, which are the same issues chip engineers have been dealing with for decades. And for medical—particularly medical equipment—the issues revolve around data integrity, throughput and security.

Rather than shrinking everything to a nano scale, flows and data maps need to be enlarged. The scale needs to shift from nanometers to meters, and in some cases kilometers. Nevertheless, the problems being addressed by those tools are very familiar, while the opportunities for increased revenue are significantly higher. These are systems of systems, rather than systems on a chip, and improving power, performance and area across all of these systems will have huge payback for tools vendors that can tap into these markets.