EDA company’s clients have finished hundreds of new tapeouts using Cadence Cerebrus AI to speed development and make chips that run faster, use less energy, and cost less. For Cadence, AI is all about increasing engineering teams’ productivity, speeding higher quality, lower-cost chips to market.
While the US Chips Act will provide funding to increase domestic design and manufacturing to compete with Taiwan, many projects are still short of experienced engineers to do the critical design work. New AI chip projects have begun that require more performance, lower power, and shorter design times than the 2-3 years typically required for advanced semiconductors. AI has delivered significant benefits to chip designers, notably with reinforcement learning for physical layout optimizations.
Chip teams are starting to realize some pretty amazing results applying AI across the multi-year workflow. At its recent annual user group event, Cadence Design walked the audience through its AI portfolio, and announced that hundreds of commercial designs have taped-out using the company’s back-end design optimization platform, Cadence Cerebrus. Cadence’s “Chips to Systems” strategy of AI assistance models and software seems robust, and is uniquely connected by a common data and model repository called JedAI. Let’s take a closer look.
The Cadence AI Portfolio
I was frankly surprised by the breadth of what Cadence has already achieved in applying AI across the chip design workflow, although details of what models are used by which tools is lacking. Cadence has quietly added AI tools to digital design back-end, debug and verification, analog custom design, printed circuit board design, and multi-physics optimization.
Let’s start with the glue that underpins the Cadence design portfolio; JedAI.
Cadence Joint Enterprise Data and AI Platform (JedAI)
The Cadence JedAI platform is an AI-driven large scale data analytics environment. Cadence JedAI is a relatively new platform that was launched in 2022 and continues to evolve.
Cadence JedAI harnesses the vast amounts of design data generated during the verification and implementation phases of the design process. This platform also enables vertical knowledge transfer which will provide customers an opportunity to build and learn from both vertically in parallel, as well as for forward for future designs.
The Cadence JedAI provides a graphical user interface (GUI) for managing the design flow, and also includes a command-line interface (CLI) for more advanced users who prefer to work in a text-based environment.
Named a winner of Fast Company’s 2022 Next Big Things in Tech Awards, Cadence Cerebrus is an AI-driven solution that helps designers optimize their chip design flow. It uses reinforcement learning to automatically optimize the full flow to meet power, performance, and area (PPA) goals. This can significantly reduce the time it takes to design a chip and improve the overall quality of the design. Cadence Cerebrus can lay out a chip much faster, with fewer engineers, and produce better products. Introduced two years ago, Cadence Cerebrus has generated significant benefits in productivity and chip quality and costs, and Cadence has the data to prove it.
“Our customers are realizing transformative results with our broad Generative AI portfolio spanning across chip-package-board-systems. With Cadence Cerebrus for instance, customers have seen up to 60 percent improvement in timing and lower leakage by nearly 40 percent, while driving up to 10x increased engineering productivity,” said Nimish Modi, Senior Vice President and General Manager, Strategy & New Ventures at Cadence Design Systems.
Renesas provided an example project where Cadence Cerebrus reduced design exploration time to a single engineer over a 10-day period, compared to traditional techniques, which can take many engineers several months to finish the job. And Cadence Cerebrus was able to improve PPA by 20%.
Virtuoso Studio, now with AI
Cadence Virtuoso is a widely-used electronic design automation (EDA) tool suite that includes a range of tools for designing, simulating, and verifying analog and mixed-signal circuits. Virtuoso is particularly well-suited for designing high-performance analog and mixed-signal circuits, such as data converters, power management circuits and RF circuits.
Recently introduced is the new Virtuoso Studio, which infuses generative AI technology into the flow. The foundry-supported solution eases the burden of process migration for schematics and layouts. Tools within the Virtuoso ADE Suite quickly re-center and validate designs post-migration, so customers can achieve aggressive time-to-market goals. Customers can utilize the AI-enabled tools to take existing IP and transform it for their next-generation designs.
Cadence Verisium is an AI-driven verification platform that helps designers optimize verification workloads, boost coverage, and accelerate root cause analysis of bugs. It is built on the Cadence Joint Enterprise Data and AI (JedAI) Platform, which enables Cadence to unify its computational software innovations in data and AI across the full portfolio of Cadence products and solutions. Verisium offers uses the following benefits:
- Increased verification productivity: Verisium can automate and accelerate many verification tasks, freeing up designers to focus on more creative and strategic work.
- Improved verification coverage: Verisium can help designers find bugs that would otherwise be missed, increasing the overall quality of the design.
- Accelerated root cause analysis: Verisium can help designers quickly find the root cause of bugs, reducing the time it takes to fix them.
Verisium is fully integrated with JedAI, and can help automatically assess failing tests and identify the code differences between versions, speeding time to solution, helping identify bug root-cause.
One area that Cadence recently addressed with AI is the job of printed circuit board (PCB) design. Similar to the challenges in back-end chip physical layout, PCB design can be automated and improved with reinforcement learning. Cadence Allegro-X AI was introduced in April 2023 and uses AI to automate the placement and routing of components on PCBs. It offers several benefits, including:
- Reduced design time: Allegro-X AI can automate many of the manual steps involved in PCB design, significantly reducing the time it takes to complete a design.
- Improved design quality: Allegro-X AI can help designers create more optimal PCB layouts, which can lead to improved performance, reliability, and manufacturability.
- Increased design productivity: Allegro-X AI can free up designers to focus on more creative and strategic work, rather than spending time on manual tasks.
- Enhanced design flexibility: Allegro-X AI can help designers explore more design options, which can lead to better overall designs.
Cadence Allegro-X AI has the potential to revolutionize the way PCBs are designed, and it is already being used by leading companies in the electronics industry.
Optimality: Multi-physics optimization
Cadence Optimality is an AI-driven system analysis and optimization solution that enables designers to explore 3D electromagnetic (EM) and high-speed signal and power integrity results and zero in on the optimal design. It is based on the AI capabilities first introduced to the market in Cadence Cerebrus.
Cadence Optimality can be used to optimize a wide range of electronic systems, including integrated circuits, packages, PCBs, and even complete electronic systems. Optimality uses a combination of AI and traditional analysis and optimization algorithms to explore the design space and find the optimal solution. This can significantly reduce the time it takes to optimize a design and improve the overall quality of the design.
Here are some of the benefits enabled by Optimality:
- Increased design productivity: Optimality can automate many of the manual steps involved in system analysis and optimization, which can significantly reduce the time it takes to complete a design.
- Improved design quality: Optimality can help designers create more optimal system layouts, which can lead to improved performance, reliability, and manufacturability.
- Increased design flexibility: Optimality can help designers explore more design options, which can lead to better overall designs.
“Insane” Performance, says Nvidia CEO
During his keynote at Computex in late May, Nvidia CEO Jensen Huang shone a spotlight on how the two companies are partnering closely to accelerate Cadence tools on Nvidia’s groundbreaking GraceHopper accelerated CPU. “Nvidia is a big customer of Cadence. We use all of their tools. And all their tools run on CPUs. And the reason they run on CPUs is that Nvidia’s datasets are very large, and the algorithms are refined over very long period of time,” said Huang. “We’ve been accelerating these applications for some time now, but now with GraceHopper – and we’ve only been working on it for a couple of days and weeks – the performance speed up, and I can’t wait to show it to you, is insane.”
Cadence introduced several new AI-driven technologies at its recent user group meeting, and the pervasive use of AI, primarily reinforcement learning, can lead to a step function in design team productivity. After successful explorations over the last two years with Cadence Cerebrus, the company is confident it knows how to help its customers speed design time and produce products that use less energy, cost less, and perform better than chips designed without the aid of AI tools.
AI will never take over the design process, but it has become a critical capability that helps design teams get the job done and focus on higher-value functions that help their employer succeed in the market.