Synopsys Collaborates with SiMa.ai on Automotive AI IP

EDA giant Synopsys and AI accelerator startup SiMa.ai will develop a joint solution for automotive customers based on SiMa’s AI hardware IP and software, though the exact nature of this solution has not yet been fully defined, both companies told EE Times.

SiMa currently offers its proprietary second-generation AI accelerator as part of the company’s Modalix SoC for multimodal and generative AI in robotics, industrial automation, computer vision, medical device, aerospace and defense applications.

Automotive OEMs and Tier-1s are adopting new technologies like software defined vehicles, where features can be upgraded in the field via software updates, Tom De Schutter, senior VP of product management and system solutions at Synopsys, told EE Times.

“For example, with AI on the on the edge in the car, you could potentially upgrade the large language model,” he said. “[For that] you need a strong, programmable machine learning accelerator, that’s able to handle the tasks you want from an AI point of view, but you also want something that can be upgraded throughout the life cycle of that product.”

Synopsys’ automotive customers are best placed to understand the needs of the application as they apply to different models of vehicles, and influence semiconductor makers’ automotive offerings, De Schutter said.

“This is where Synopsys saw an opportunity to help bridge from silicon to system, to help OEMs influence the semiconductor side, through virtual models of the end electronics,” he added.

Virtual models—or digital twins—of hardware will enable architectural exploration to take place earlier in the design cycle, so that customer requirements can be fed back earlier to semiconductor manufacturers. Creating virtual environments with multiple electronic components in the car means software development and validation can take place in the context of multiple components working together.

“Customers are saying that having just the IP doesn’t help them, they need an environment where they can explore how the IP will behave in the final environment, completely up front, so they can work with the semiconductor partner and say, I need this IP, I need this many versions of it, these configurations of it,” De Schutter said. “They also want to start the software development well before the car comes out. Customers said the only true solution is being able to do all this pre-silicon.”

Hardware IP can also be mapped to emulators for final performance and power validation before going to silicon, to assure OEMs that the silicon will meet their needs. This is getting more important in the compute-heavy age of AI, De Schutter said.

Specific vertical

This is a new type of collaboration for Synopsys, De Schutter said, that will focus on solutions for a specific vertical.

“We do expect that [our virtualization technologies] will eventually be required in other markets, but we see that automotive is the frontrunner, and [that market] is also having trouble dealing with that change,” he explained. “Obviously, problems also mean an opportunity to provide more of a solution.”

Automotive customers requested AI hardware IP that could handle multimodal models and was sufficiently reprogrammable, De Schutter said, and SiMa’s accelerator fit the bill.

“In a sense, it was driven by customer discussions, customer demand,” he said. “For us it’s not something exclusive, or something we only want to do with SiMa.ai, but we did see customer demand for that as the first off.”

Asked whether working together with SiMa would clash with Synopsys’ own ARC NPU IP, which can be configured up to 250 TOPS and can be ASIL-D compatible, De Schutter said the focus is on providing options for its 50 plus OEM and Tier-1 automotive customers.

The exact nature of the eventual Synopsys/SiMa solution is still under discussion, De Schutter said.

“Right now the focus is on the enablement of the architecture exploration, and the early software development, so we can see how is that received by OEMs,” he said. “This combination is a unique offering to explore [microarchitectures], and help customers make those decisions, both on which IP to use, but also within that IP, how to configure it for their specific needs.”

De Schutter added that Synopsys is intent on enabling automotive customers to work on the entire compute subsystem, and that the work with SiMa will be complementary to Arm’s Total Design program, for example.

Under discussion

Automotive OEMs and Tier-1s are thinking about whether they should make their own AI chips, Harry Kroeger, head of automotive at SiMa.ai, told EE Times.

“Every automotive player I know is thinking: should I be a customer of a big company where my business is completely unimportant, or should I take my destiny into my own hands and do something different?” he said.

EDA vendors like Synopsys enable homegrown silicon development, virtualization and testing in these companies, he added.

“We’ve shown the world how powerful our [accelerator] is, how power efficient it is, which are the key ingredients for automotive,” he said. “Bringing this together is great for us, of course, it’s exciting because it opens up new ways to go into the market.”

Kroeger also emphasized that the exact nature of the joint solution the two companies will work on is still under discussion, but SiMa will be seeking input from its automotive customers in person at CES 2025, he added.

“That will help us define our path in more detail,” Kroeger said. “The automotive world needs exactly the product we’re trying to enable here, and this is the first step we are taking.”

SiMa does not currently offer its AI accelerator as an IP product, only as part of its Modalix SoC. Simply offering proprietary IP to end customers would be a difficult way to work, Kroeger said.

“It’s not just the accelerator part, getting a high performing, super power-efficient ML SoC together means you need to have a lot of other pieces connected to that,” he said. “Theoretically a company could say, give me your IP and I’ll build it into my product, but then how do they test that? How do they virtualise it? How do they make a test run on a chip that doesn’t exist yet?”

Synopsys was impressed with SiMa’s technology, including its software stack, Kroeger said. The collaboration will support chiplet-based designs as well as monolithic SoCs, he added.

Specific vertical

Under discussion

Suggested Reading