Never miss a story from Computational Fluid Dynamics. Subscribe for in-depth analysis and articles.
Optimizing a silicon chip at the system level is crucial in achieving peak performance, efficiency, and system reliability.
As Moore's Law faces diminishing returns, simply transitioning to the latest process node no longer guarantees substantial power, performance, or cost reduction improvements for a given product iteration. Instead, the focus has shifted towards comprehensive optimization of the overall design, from place and route to thermal and power budget—as a fundamental requirement to achieve significant advancements.
This complex and expensive workflow incorporates multiple disciplines: design, prototyping, testing, refinement across multiple iterations, verification, and eventual manufacturing. It’s a workflow that has required extensive restructuring of design teams, moving away from sequential delivery towards simultaneous design of multiple components within a system.
As more system companies design their chips and semiconductor companies deliver software stacks, optimization has spread to software, too. But while this approach has yielded many remarkable achievements and enabled substantial differentiation of end products, the next great leap in system-level optimization will require us to think bigger, broader, and more holistically than ever before. We need to optimize for the end device, end-user, the environment, and the ever-evolving market—redefining what we mean by "system-level" to encompass everything from the silicon to the entire end product, be it a smartphone, a car, or an entire data center.
This isn’t a new idea—as far back as the 1970s, the aerospace industry was exploring optimization techniques that focused on the entire airplane over individual components in response to airline and Air Force procurement policies shifting from peak performance towards overall lifecycle cost.
With Cadence Optimality Intelligent System Explorer, we brought in-design optimization to chip design for the first time. But if the aerospace industry has been looking at overall systems in terms of jumbo jets and fighter planes, and Cadence software is used to create technology that goes into those airplanes, why should our role stop at chip design? It makes perfect sense to expand "system level" to include an airplane’s entire electronic system, its network of sensors and actuators, the shape and construction material of its fuselage, the airfoil profile of its wings, and the ease of future maintenance and upgrade.
At this scale, we verge on chaos theory—each element interoperates with and, therefore, affects one another. If you optimize the electronic system design and then change the airfoil's shape, might the electronics need to be reoptimized somehow? Even the environmental conditions the airplane will eventually be used in might affect how we design certain systems—for example, to deal with wildly fluctuating temperatures.
I’m speculating, but this is exactly how it works in nature. Animals and plants have evolved to function optimally in isolation and within the wider ecosystem—from the cellular to the planetary. Take the polar bear, which has for hundreds of thousands of years been a perfectly optimized system of energy conservation and thermal regulation, capable of surviving up to half a year without food in temperatures as low as -50°F. Yet today, the global polar bear population is shrinking rapidly, with numbers projected to decline by 30% by 2050. The harsh reality? The polar bear is no longer optimized for the (eco)system in which it lives.
In product design, unlike in nature, we can iterate and evolve far more rapidly. But iterative design in the real world is all-consuming (and even then, you’ll never know if you’re truly, truly optimized). The traditional V-model of systems development places implementation mid-lifecycle, yet when you start to extrapolate out to such a huge scale and to a seemingly infinite number of requirements and variables, building anything in the physical world only limits the system's optimization as a whole.
The solution is simulation in design. But until recently, simulating anything this complex would still have placed even simple designs outside the realms of possibility. However, we’ve spent the past few years applying artificial intelligence (AI) to each domain-specific design tool—enabling our customers to take a quantum leap forward in accelerating the design, simulation, and verification of hugely complex systems.
For engineers, this has been life-changing. For the past 30 years, they’ve followed a process of input, output, analysis, tweaking of parameters, and repeating this process until after some weeks, months, or potentially years, they arrive at something like an optimized design. With AI-powered simulation, this iteration process can take only days or even hours.
Cadence has extended its expertise into fields including Computational Fluid Dynamics (CFD), used to characterize the flow of liquids and gases inside and around physical products, and molecular simulation. Now, by developing AI for the overall system, Cadence is taking steps to break the mold and offer a comprehensive approach to system optimization—one that goes far beyond the silicon chip to consider all aspects of a product’s design and application.
We’re paving the way for a truly integrated optimization approach, something we are uniquely positioned to deliver—one that enables our customers to optimize in many dimensions simultaneously, with all the competitive promise and design excellence it offers. Watch this space.