Andes Technology and MetaSilicon Collaborate to Build the World’s First...
By Sanjay Bali, VP of Strategy and Product Management, Synopsys EDA Group
2022 was a big year for the electronics industry. From the continued growth of AI (both in end devices and in chip design itself) to the emergence of more ways to design in the cloud, the level of innovation we saw was impressive. And it was necessary, as we experienced how complex semiconductors have become, with engineers striving to meet the progressively challenging task of optimizing power, performance, and area (PPA) in chips with as many as trillions of transistors.
In addition to systemic and scale complexity, a variety of other challenges were laid bare, from cyclical swings in product demand to a shrinking pool of engineering talent to the growing impact of all this energy consumption on our planet. Given this backdrop, what might 2023 bring to this increasingly vital industry?
For decades we have seen how engineering ingenuity has persevered to extend Moore’s law and extract more computing power from a single chip—in spite of the limitations of physics. To create products like autonomous cars and advanced robotics. Ahead of us, we see the market demanding more sophisticated features from chip technologies that power our world.
In 2023, we can expect to see these trends continue to unfold, with a few emerging technologies taking hold to further shape the industry. In this article, originally published on the “From Silicon to Software” blog, you’ll learn about three key technologies that are poised to transform electronic design.
1. Silicon Lifecycle Management Hits the Mainstream
Big data analytics has made inroads in a variety of industries, from science to finance. Now, it’s time for the chip industry to turn a trove of information collected from every phase of the device lifecycle into actionable insights. Design teams in market segments such as high-performance computing (HPC) and automotive are well-versed in silicon lifecycle management (SLM). In the HPC space, SLM technologies help designers of data center server SoCs meet service level agreement uptimes—99.99% (four nines) or 99.999% (five nines) availability over a certain time period are considered reasonable. For automotive designers, SLM technologies continuously assess factors such as aging and degradation, paving the way to a more predictive approach to maintenance and replacement of in-vehicle electronic systems. However, given an increasing emphasis in reliability, availability, and serviceability (RAS) for mission-critical applications across many other market segments, SLM is poised to become more widely used in the coming year.
With process variability (particularly at advanced nodes), as well as environmental and aging effects, silicon designs are under a great deal of stress. At the same time, they are expected to perform at ever higher levels and, in cases like automotive, last for longer periods of time.
SLM provides the mechanisms for monitoring and analyzing semiconductor devices in real time while in design, production, system bring-up, and in the field. On-chip sensors and monitors collect data continuously, sending it to AI-enabled analytics engines in the cloud to enable optimizations at each stage in the silicon lifecycle. Predictive and prescriptive analytics can bring greater efficiency to the design process while improving quality of the designs themselves.
2. Greater Integration of Cloud Technologies
In 2022, we saw new ways to design and verify semiconductor devices in the cloud. Flexible, pay-per-use solutions that allow on-demand access to electronic design automation (EDA) tools (either as a software-as-a-service (SaaS) model or a bring-your-own-cloud model) simplify software licensing, allow use of any tool at any scale, and free designers to let their designs dictate how they’ll use design and verification tools. Further extending the SaaS approach, flows for analog design, digital design, and verification provide on-demand, easily accessible, and scalable access to tools when they are needed.
While a variety of industries have long used cloud technologies, the semiconductor industry is now turning to the cloud, taking advantage of its elasticity, flexibility, and scalability to design chips better, faster, and cheaper. Cloud solutions also facilitate design of next-gen systems based on real-time data using digital twins. A virtual model or clone, a digital twin allows engineers to apply adjustments, adaptations, and optimizations based on inputs gathered from sensors in the field and relayed via the cloud to the processing system in the lab.
It’s no longer a matter of whether the cloud will become mainstream for semiconductor designs, but a matter of when—and 2023 looks to be an inflection point for wider adoption. SaaS cloud solutions, in particular, make it very feasible for companies with limited IT budgets to pay for EDA resources as needed when needed. In response, we can look to the industry to roll out more collaborative cloud-based design and verification tools.
3. Broader Adoption of Multi-Die Systems
By integrating heterogenous chiplets into a single package, a multi-die architecture can deliver performance, power, cost, and yield benefits on a scale that’s not possible with monolithic SoCs. While not new to the semiconductor landscape, multi-die systems look to have a big year in 2023. The ecosystem around multi-die architectures is fast maturing, reducing previous barriers to entry and easing the path toward wider adoption. A greater prevalence of holistic, unified, and mature tools is simplifying design, verification, testing, signoff, and silicon lifecycle management of multi-die systems. Standardized IP for robust, secure die-to-die connectivity lower integration risks while accelerating standards adoption. Foundries as well as outsourced semiconductor assembly and test (OSAT) vendors have offerings such as advanced packaging, ecosystem alliances, and flows for multi-die systems. And standards like Universal Chiplet Interconnect Express (UCIe), which offers a complete stack for the die-to-die interface, are making it easier to bring disparate components together while meeting aggressive bandwidth, latency, and energy efficiency demands. The convergence of all these developments bodes well for moving multi-die into the mainstream.
An expansion in multi-die system designs could trigger a shift toward multi-foundry strategies, since design teams can opt to use, for example, an advanced node for the core and an older process node for the peripherals in their systems. The supply chain challenges we saw in 2022 are also causing chip designers to consider multi-foundry strategies for greater predictability.
Related to the move to multi-die is the next frontier in semiconductor technology: silicon photonics. Using the power of light to transmit and process data, photonic ICs can be integrated into multi-die systems to deliver energy-efficient bandwidth scaling. As multi-die systems become more entrenched, photonic ICs could also see an uplift.
4. Addressing Tech Talent Gaps
Engineers have never been the type to back away from a good challenge. As we’ve experienced in years past and will see again in 2023, their ingenuity will continue to lead the way to powerful chips that can take on the most demanding of compute workloads. Investing in talent is a must and programs such as the Synopsys Academic & Research Alliances (SARA), that provide educational opportunities, access to advanced technologies, and collaborative support to students, educators, researchers, and entrepreneurs, can help nurture the next generation of engineers.
As new technology challenges emerge, Synopsys is primed to help design teams realize their silicon goals, whether they’re designing monolithic, energy-efficient SoCs, 2.5/3D multi-die systems, or something in between. Our full stack facilitates design, verification, testing, manufacturing, silicon lifecycle management, and software security.
While no one can really predict the future, we can see the trends and patterns that may shape the coming year. AI and big data analytics are becoming more prevalent in our everyday lives (and even in chip design), while cars are being infused with greater levels of intelligence. Hyperscalers are demanding more from their chips to find solutions to problems as vast as vaccine discovery, grid-scale energy storage, and world hunger—and in doing so, they’re shaping the chip industry itself. Multi-die systems are fast becoming an answer to the slowing of Moore’s law, and with these architectures come greater opportunity to deliver what our data-driven, bandwidth-hungry world needs.
It’s a transformative time in the electronics industry, as we see chips playing a critical role in innovations that are tackling some of the world’s biggest challenges. The three key technologies anticipated to make a big mark in 2023—silicon lifecycle management, the cloud, and multi-die systems—are primed to move the momentum toward a smarter, more equitable world.
Join our community of thousands of semiconductor professionals. Keep up to date with all latest news and activities. Don’t miss out!