The end product isn’t what makes a new technology safe. It’s everything that goes into a product — from design to development to manufacturing.
That’s why global safety experts at German agency TÜV SÜD have assessed the development process for the NVIDIA Xavier system-on-a-chip (SoC), the world’s first processor for autonomous driving, and approved it as suitable for building a safe product.
With the completion of this step, Xavier is one approval step away from being the most complicated SoC to demonstrate compliance with international safety standards.
NVIDIA approached the assessment of Xavier as a three-step process. In April, TÜV SÜD deemed the processor’s safety architecture suitable for use in safety applications. With today’s announcement, the assessment of the development process applied to Xavier has been completed.
The final step, assessment of the Xavier implementation, will start soon.
Xavier: The World’s First Autonomous Machine Processor
Building the industry’s most complex SoC didn’t happen overnight. Xavier represents the work of more than 2,000 NVIDIA engineers over four years and an investment of $2 billion in research and development.
To achieve the performance capability to power autonomous driving functions from Level 2+ advanced driver assistance to Level 5 robotaxis, Xavier builds on the lessons learned from our two previous generations of automotive SoCs. Most importantly, this development process has taught us that increased computation performance enables higher overall safety.
And Xavier can tap into enormous performance, more quickly than ever. Xavier uses NVLink interconnect technology so it can be paired with dedicated GPUs at speeds of up to 20GB per second — 10x faster than what’s possible with previous PCI Express connections.
NVIDIA DRIVE AGX Pegasus incorporates two Xavier SoCs and two next-generation NVIDIA Tensor Core GPUs to deliver the computational speed required for truly driverless vehicles. NVLink also resolves the coherency bottleneck found in some of today’s computing architectures.
Building on this iterative development, Xavier incorporates redundancy and diversity into its architecture as well as its engineering processes, ensuring that failures are reacted to and mitigated as required.
Measuring Up
With an architecture built for safety, the next step in establishing safe self-driving technology is ensuring the development processes to build it are just as robust.
Established by the International Organization for Standardization, the world’s chief standards body, ISO 26262 is the definitive global standard for the functional safety — a system’s ability to avoid, identify and manage failures — of road vehicles’ systems, hardware and software.
These standards include the technology as well as the processes that surround it — ensuring a product has been developed in a way that mitigates potential systematic and random hardware faults. That is, SoC development must not only avoid failures whenever possible, but also detect and respond to them when they cannot be avoided.
Using these standards to measure the Xavier development process, the safety experts at TÜV SÜD — which has served as one of the leading technical agencies for product testing and certification for more than 150 years — deemed it suitable to build a safe automotive product.
The Road Ahead
After completing the first two assessments for ISO 26262 compliance, the next and final step is to ensure the final product achieves the intended safety performance. We expect to complete this process soon, making Xavier the world’s first autonomous driving processor to demonstrate ISO 26262 compliance.
Of course, the journey to safe self-driving doesn’t stop there. From our Self-Driving Safety Report recently submitted to the National Highway Traffic Safety Administration to new and robust validation processes, NVIDIA is continuously driving toward a safer autonomous future.