First Armv9 automotive CPUs aim to power AI-enabled vehicles

Vehicle electronics and software becoming ever more complex

Chip designer Arm is bringing its Neoverse architecture to the automotive sector along with the first Armv9 processors for in-vehicle applications, claiming these will provide better performance for AI.

The chip designer has introduced Automotive Enhanced (AE) versions of two Cortex-A processors, plus a Neoverse V3AE high performance core, a real-time processor and image signal processor, all built for in-vehicle use cases.

As ever, these are designs for Arm's chip manufacturing customers to license for their own silicon products, and so are not likely to filter through into production vehicles for several years.

The automotive market, according to Arm, is going through a period of big change, with vehicle electronics and the software becoming more complex.

To try to tackle this, the chip designer said it is bringing server-class Neoverse tech to this market along with the AI, security and virtualization features of Armv9-based Cortex-A products.

The Neoverse V3AE is aimed at use cases like autonomous driving and high-end driver assistance systems that call for high single-thread performance, so it is lucky that it is based on the V3 core announced last month, billed by Arm at the time as being the "highest single thread performance Neoverse core ever."

This means the core has the same underlying CPU design and performance as in Arm's latest cloud and datacenter designs, to address growing compute demands in vehicles, and adds Automotive Enhanced (AE) features.

Likewise, the Cortex-A720AE and the Cortex-A520AE are based on the Cortex-A720 and Cortex-A520 that the chip designer introduced last year as part of its TCS23 system-on-chip (SoC) platform, but with added AE features.

Those AE features include Transient Fault protection (TFP) and Split-Lock support, which lets a pair of cores be configured to run independently as normal (split mode), or lock stepped for fault tolerant operation.

Cortex-A720AE targets driver assistance and "digital cockpit" systems, while Cortex-A520AE is aimed at use cases where processing workloads as efficiently as possible is required.

Arm has seen a 30 percent performance uplift when moving automotive workloads from the older Cortex-A78AE to Cortex-A720AE, it says.

As with the TCS23 platform, the Cortex-A720AE and Cortex-A520AE can be combined in cluster configurations of up to 14 cores, thanks to the DSU-120AE. This is a version of Arm's DynamIQ Shared Unit glue logic that now has the necessary AE features for automotive systems.

Also introduced is the Cortex-R82AE, a real-time core designed for functional safety applications, which might include monitoring other processors or controlling vehicle functions. Cortex-R82AE brings 64-bit processing to this line for the first time, Arm says.

The final core is the Mali-C720AE image signal processor (ISP), designed for in-vehicle vision systems. Compared with previous designs, this features multiple parallel pipelines that can be tuned differently to support human vision or computer vision use cases. This reduces the time taken to process images, speeding up the overall reaction time of the compute system, Arm claims.

The chip designer has rounded out its portfolio with AE-enabled versions of components such as Network on Chip (NoC) interconnects and Coherent Mesh Network.

Arm also says it is introducing virtual prototyping, starting with these latest Arm AE automotive products. This will allow software developers to start their work before the physical silicon is available, shortening time to market by up to two years, the company claims.

"The additions to the Arm AE IP portfolio will power and enhance the next generation of vehicles," Arm senior director of Product Management for Automotive Tom Conway wrote on the company's blog.

Conway said that the company was making tech to respond to the shifts taking pace in the automotive industry, including AI-powered workloads and enhanced security and safety features. ®

More about

More about

More about


Send us news

Other stories you might like