Why the industry needs to look way beyond the next process node.
Design Automation has been a crucial part of the semiconductor industry for more than 30 years. Without it, keeping up with Moore’s Law would have been impossible. The Design Automation industry accepted the high-risk responsibility for developing the sophisticated software and algorithms at the pace necessary to corral Moore’s Law.
In a sense, Design Automation had its own Moore’s law. The challenge was to enable the design of next-generation chips in the same time and at the same cost as those of the last generation. Design Automation took on the challenge for decades and succeeded. The promise of Moore’s Law was maintained.
Conventional wisdom suggests that the next process node is coming within 24 to 30 months, though it’s probably closer to five years. Assuming this is accurate and Moore’s Law is slowing down, the challenge faced by Design Automation has fundamentally changed. Rather than slow down to match the pace of new node geometries, design tool providers are being challenged to support a new diversity in application areas, supported by more variations of the current process nodes than had existed.
Design Automation needs to pivot like the U.S. presidential candidates moving away from their party primaries to the general election. A pivot from supporting leading-edge designs on an impossible schedule to supporting a vast variety of designs, each of which may be using an existing process node in new and unique ways.
SoC design—in the sense of real system design, not just more powerful chip silicon—is changing faster than incremental shrinks in node geometry. Other system dimensions will be attacked and challenged, from power to communication to packaging and prototyping to software and security.
Slowing down Moore’s Law challenges the manufacturers of the current process nodes to differentiate, making it difficult to predict what happens next. It won’t be the usual simple shrink. This offers more opportunities, not less, to meet challenges that were not predicted at the beginning of the life of the node.
As in the past, chip designers will depend on Design Automation to guide them through this new and challenging territory, starting from the front end. That, after all, is Design Automation’s job.
The variety of electronic system design will be a new challenge, and the globally spreading design ecosystem will benefit from Design Automation’s all-important brain trust, experience and know-how. This will mean identifying what needs to be supported at the existing node, which new and innovative products will drive the need to modify the characteristics of existing nodes and requirements of new design methodologies.
Moving meaningfully, just not at the leading edge, is essential. Engineers building the next-generation SoC are not asking for more transistors or more end-user functionality. Instead, they are focusing on other variables, like security, to state an obvious example, because the dynamics of security are changing and the evolution of security for the lifetime of the SoC is fundamental.
Unlike previous chip designs, engineers today need more than just size and time granularity. Modern process technologies will add features specifically to meet the needs of modern designs. This will include special processing and materials to help with power consumption and security as well as a host of other requirements driven by the increase in new applications especially related to the Internet of Things. Certain designs will be defined by these new process features, such as 2.5 or 3D or other IoT-specific process variations, bringing new challenges to the tasks of design and verification.
The electronic design community itself will be called upon to change and expand. The software design community may be showing the way. It has changed dramatically from the days of the PC to the days of the smartphone. Many designs will be created by communities of individuals and small entities and not by companies in the traditional sense of the term. Community matters and communities will matter more.
One only has to look at the emergence of community software sharing sites such as GitHub to see new development paradigms in place and working well on a large scale.
Which brings us to what Design Automation methodology and tools could look like. One beneficiary of the new design community could be Open Source –– a free starting point where complicated tools are built on top of Open Source platforms. It’s been a successful and effective business model in other market segments.
Many Semiconductor companies have adopted this model rather than buy expensive or unavailable tools. Design Automation could work with an Open Source design ecosystem and, in a way, it’s already structured to do this. After all, the UNIX and Linux operating systems are Open Source and the technology industry has been building on top of both for years.
Change the business model and build value. A Design Automation industry pivot to support a larger variety of designs, without sacrificing the needs of the cutting edge, would give the design ecosystem the support it needs to build the next-generation SoC. I mean real systems and full systems, not just hardware on chip at the current process node.
Lucio Lanza will moderate a discussion titled, “Lanza’s Tech Vision Challenge: Daring to Move to Open Source,” Monday, June 6, at 2pm during the Design Automation Conference (DAC) in Austin, Texas. It will be held on the Exhibit Floor in the DAC Pavilion (Booth #1839).