Continental builds a Nvidia-based supercomputer for autonomous vehicle AI development, extending GPU infrastructure beyond cloud providers.

Continental is building a high-performance supercomputer cluster based on Nvidia DGX AI systems to boost autonomous driving development. The system, which has been operating from a datacenter in Frankfurt am Main, Germany since the beginning of 2020, is intended for vehicle artificial intelligence system training and reduces development time from weeks to hours. The main use cases include deep learning, simulation, and virtual data generation. According to the TOP500 list, which monitors the fastest and most powerful supercomputers in the world, Continental's and Nvidia's high-end system is the most powerful computer in the automotive industry.

The supercomputer is built with more than 50 Nvidia DGX systems connected with the Nvidia Mellanox InfiniBand network. Christian Schumacher, head of Program Management Systems in Continental's Advanced Driver Assistance Systems business unit, emphasizes that "the state-of-the-art system reduces the time to train neural networks, as it allows for at least 14 times more experiments to be run at the same time." The project was implemented in less than a year, and Schumacher notes that "the supercomputer is a masterpiece of IT infrastructure engineering. Every detail has been planned precisely by the team – in order to ensure the full performance and functionality today, with scalability for future extensions." Manuvir Das, head of enterprise computing at Nvidia, states that "Nvidia DGX systems give innovators like Continental AI supercomputing in a cost-effective, enterprise-ready solution that's easy to deploy," and that Continental is "engineering tomorrow's most intelligent vehicles, as well as the IT infrastructure that will be used to design them."

Advanced driver assistance systems use AI to enable vehicles to make decisions, assist drivers, and ultimately operate autonomously. However, traditional software development and machine learning methods have reached their limits as systems become increasingly complex, making deep learning and simulations fundamental to AI-based development. Training neural networks requires millions of images and enormous amounts of data—while a child can recognize a car after seeing a few dozen pictures, a neural network needs several thousand hours of training. Balázs Lóránd, head of Continental's AI Competence Center in Budapest, Hungary, explains that "overall, we are estimating the time needed to fully train a neural network to be reduced from weeks to hours," and notes that "with the supercomputer, we are now able to scale computing power even better according to our needs and leverage the full potential of our developers."

Continental's test vehicle fleet drives around 15,000 test kilometers each day, collecting around 100 terabytes of data—equivalent to 50,000 hours of movies. This recorded data can be replayed to simulate physical test drives and train new systems. With the supercomputer, data can now be generated synthetically, a highly computing power-consuming use case that allows systems to learn from traveling virtually through a simulated environment. This approach enables vehicles to navigate safely through changing and extreme weather conditions and make reliable forecasts of pedestrian movements, paving the way to higher levels of automation.