ARM has introduced a compute sub-system (CSS) featuring 16 high-performance processors aimed at accelerating the development of the next generation of AI chips tailored for automotive applications. This marks the debut of the Zena branding, offering 16 Cortex A720AE automotive processor cores alongside Cortex-R82AE real-time microcontroller cores.
One notable aspect is the optional inclusion of the Mali GPU and AI accelerator, suggesting ARM anticipates chip manufacturers to incorporate additional IP, either within the chip as RTL or as a chiplet. This shift signifies a move towards more AI in software-defined vehicles, consequently escalating processing demands.
According to Dipti Vachani, senior vice president and general manager of the automotive and IoT product lines at ARM, "Over 94% of automakers utilize ARM technology, and the top 15 silicon suppliers leverage our technology. All new entrants in electric vehicles also embrace ARM technology, providing insights into market trends, from software-defined vehicles with centralized compute to AI-defined vehicles."
Vachani further elaborated, "What sets the automotive sector apart is the early adoption of AI compared to other industries. The unique requirements of AI-defined vehicles, characterized by heightened compute demands, have propelled the development of the Zena CSS."
As part of the rebranding initiative for its core IP, ARM has integrated the ARMv9 720AE core, introduced last year, into the CSS. These cores will be accompanied by a safety island for real-time processing through the Correx-R82AE cores and a security enclave crucial for safeguarding vehicle data. The system-wide security and root-of-trust are facilitated by ARM TrustZone, with a Runtime Security Engine enabling secure over-the-air updates.
- ARM to re-brand its core IP
- Faraday licenses its first ARM automotive core
- ARM moves Neoverse V3 core into automotive
Moreover, a higher-performance core, the NeoverseV3AE, is utilized in the upcoming Nvidia Drive Thor chip. Future iterations of the Cortex-A520AE, A720AE, and NeoverseV3AE are anticipated to be marketed as the Zena Pro, Premium, and Ultra cores, aligning with the recent product line rebranding.
By leveraging the CSS, projects require fewer engineers, expediting development timelines. Vachani highlighted, "This translates to a 20% reduction in engineers per project and a 12-month time saving, offering substantial benefits for our customers. This enables customers to concentrate on differentiation, whether they are an OEM, silicon provider, or tier one supplier."
The interface to the GPU or neural processor adheres to the UCIe standard, utilizing PCI Express modulation for simplified integration. This design allows for scalability, enabling the addition of multiple AI accelerators or NPUs to the CSS for high-end processing, such as in self-driving vehicles, while maintaining consistency across vehicle models.
ARM has collaborated with industry leaders like Amazon AWS, Cadence Design Systems, Siemens EDA, and Synopsys to develop tools for running software on the CSS in the cloud, enabling early-stage development before silicon availability. Software plays a pivotal role in development, particularly for AI-defined vehicles, where developers can validate AI workloads and edge inference behavior using virtual platforms and RTL emulation.
The CSS is tailored for the Scalable Open Architecture for Embedded Edge (SOAFEE), with the SOAFEE community creating Blueprints for digital cockpit and entertainment system designs incorporating AI for detection and voice control. These Blueprints adhere to the SystemReady certification and compliance program, utilizing industry-standard APIs atop the Zena CSS.
ARM is actively working on extensions to SystemReady slated for the end of 2025, focusing on operating systems and software stacks for automotive applications that can operate across diverse hardware platforms.