ARM moves Neoverse V3 core into automotive

ARM moves Neoverse V3 core into automotive

Business news |
By Nick Flaherty

ARM has developed a version of its Neoverse data centre processor core for the automotive market, along with a compute sub-system to speed up chip designs.

There are also the first ARMv9 cores for automotive and the first 64bit real time R-class core.

The V3AE is the first core with server-class performance for automotive with functional safety and is aimed at AI-accelerated autonomous and driver assistance ADAS workloads. The first v9-based Cortex-A processors purpose-built for automotive are the Cortex-A720AE for a broad range of software-defined vehicle (SDV) applications and the Cortex-A520AE with a focus on power efficiency with functional safety features.

The Cortex-R82AE is ARM’s highest performing real-time processor for functional safety and the first to use 64bit. The Mali-C720AE configurable ISP is optimized for both computer and human vision use cases.

Zonal controllers must be able to support a mix of different functions combined, ranging from conventional signal-based
control, which is often the realm of AUTOSAR classic software stacks, to service-based software, which often use POSIX-compliant operating

The Cortex-R82AE combination of support for both a physical memory system architecture (PMSA) and its VMSA means software stacks
can be supported on a single processor, while providing the ability to combine software through the processor’s real-time virtualization support.

Compared to Cortex-R52+, one 64-bit Cortex-R82AE core can deliver more than 50 percent higher performance for single-thread workloads
at the same execution speed. This is achieved by its high performance in order pipeline. Twice the the number of Cortex-R82AE cores can now be integrated in a single cluster – eight compared with four in Cortex-R52+ – there is an even greater performance increase.

There are also performance improvements in the FPU and SIMD extensions, which have been updated to offer higher performance for
AI and ML-based workloads through the availability of dot product operations.

The ARM Cortex R82AE real time core

The ARM R82AE real time core

The cores are available in a compute sub-system (CSS) with memory, interconnect and peripherals, and there are virtual models to allow designers to start work on the software before the silicon is available.

“Mercedes-Benz is the first car manufacturer to receive an internationally valid type approval for L3 automated driving. Efficient, high-performance semiconductors are key to push the boundaries in development, if you want to reach the next level. Therefore, I am highly delighted to see Arm launch their latest Automotive Enhanced IP portfolio alongside the launch of virtual platforms enabling software development and verification. These types of innovations will support the development of future automated driving systems,” said Georges Massing, Vice President MB.OS Automated Driving & E/E Integration, Mercedes-Benz.

“Innovation, speed and cost-effectiveness are key requirements for ZF and our customers in the development of software-defined and automated vehicles. This applies in particular to the new E/E architectures with their high-performance computers. As a system provider for this new generation of vehicles, ZF offers ProAI, the most powerful high-performance computer for automated driving. Simultaneous engineering is an important prerequisite for innovative, fast and cost-optimized developments and thereby ARM’s new virtual platforms and software solutions will optimally support these requirements.” said Oliver Briemle, Vice President AD Components & Connectivity at German supplier, ZF Group.

The V3AE is already used by Nvidia for its Thor chip for driverless cars.

“Nvidia Drive Thor brings the highest levels of CUDA parallel compute and cutting-edge generative AI capabilities to autonomous machines, including vehicles and robots. The Thor SoC integrates ARM’s Neoverse V3AE to provide the industry-leading, single-thread CPU performance necessary for intelligent cabin experiences and self-driving capabilities built for safety and security,” said Ashish Karandikar, Vice President, Hardware Engineering at Nvidia.


If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles