Semiconductor technology specialist Arm is launching a new central processing unit (CPU) platform it says will cut engineering demand for automakers 20% and trim vehicle-development time 12 months as the industry moves past the software-defined-vehicle stage and toward artificial-intelligence-defined vehicle architectures.
The AI-defined vehicle, a concept that appeared to surface for the first time late last year, is the next leap eyed in automotive electronics. Like SDV platforms, AI-defined vehicles will feature centralized, high-powered computers stocked with updatable software, but in this case the software will be AI-driven and allow for more advanced automated-driving functionality and sophisticated interaction between the vehicle and its occupants.
The Arm Zena Compute Subsystems platform is designed specifically for automotive applications. It consists primarily of 16 Arm CPUs, a safety island and a security enclave that can be paired with any GPUs (graphics processing units) and NPUs (neural processing units), and other components of an OEM’s choosing, to produce a complete system on a chip (SoC) that is both scalable and flexible. The identical platform can be deployed for advanced automated-driving functions, as well as for in-vehicle infotainment and vehicle system controls.
“The AI-defined vehicle will be the future of Arm,” Dipti Vachani, senior vice president and general manager of Automotive, says during a media backgrounder tied to the launch of the Zena CSS for testing. “(And) the Arm platform will be necessary to deploy the AI-defined vehicle.”
News of cutbacks to development programs for next-generation battery-electric-vehicle platforms – once earmarked to lead the transition to SDVs and beyond – and fits and starts in OEM efforts to develop more sophisticated software-driven electronics architectures, have led to some uncertainty as to when the transition to smart vehicles may occur. But Vachani believes Arm has made the right bet with the Zena CSS technology that is geared to driving such a transformation.
“The movement to AI in the car is absolutely happening,” she says, noting every automaker is looking to move up ADAS capability toward more fully autonomous vehicles and integrate AI into the user experience inside the cabin. “So, AI workloads are becoming table stakes.
“No one is debating whether or not AI is going into the car,” she adds. “That’s going to happen. It’s not if, but when. It’s more a conversation of, ‘What’s it going to look like? What’s my brand stand for and how do I want my consumers to experience AI in my car?’”
That dilemma is where Zena CSS can shine, Vachani says, because the platform allows for OEMs to customize their SoCs any way they want and simply add multiple boards to meet their processing power demands, leaving more time for their engineers to focus on differentiating the technology according to their brand objectives.
OEMs are demanding three things when it comes to development of these new AI-defined vehicle architectures, the Arm executive says: They want the flexibility to innovate and the ability to scale the technology across vehicle platforms and their various brands, as well as shorter development times in order to speed new products to market.
“It’s amazing to me that we’re (now) launching more new technology into the car than we see in other industries,” Vachani says, noting automakers long have been criticized for lagging behind in releasing new software features.
Zena CSS comes pre-validated and integrated with other SoC technology, the key to allowing automakers to potentially cut a year off development time. It is compatible with SOAFEE (Scalable Open Architecture for Embedded Edge) standards, so it will work seamlessly with software and systems that meet that protocol, Arm says, giving automakers access to a smorgasbord of software and services offered by third-party developers that can be incorporated more easily into their future vehicles.
Zena “is huge for the SOAFEE ecosystem,” Vachani says.
Arm says Zena CSS has already been delivered to select partners for testing and it expects to ship demonstration units more broadly in September. The company declines to speculate when it might see production applications.
Suraj Gajendra, vice president of automotive products and software solutions, sums up Arm’s play this way: “The movement to AI is absolutely happening. We are ahead of the game here. We have sort of predicted where this is going, and there needs (to be) some foundations that have to be set on how technology will evolve to support these use cases in the future. And that’s what Arm has been part of.”Arm’s semiconductor IP is licensed by a number of chipmakers and has been a factor in the automotive market for more than 20 years. The U.K.-based company says its technology is deployed by 94% of the world’s automakers and the top 15 automotive processor suppliers. Its revenues have nearly tripled in the past five years, Vachani says.
