Nvidia Unveils Powerful New Processor for Self-Driving Cars

The new Tegra "Parker" processor is pegged as the power behind the company's Drive PX 2 platform for autonomous vehicles that was introduced at CES in January.

Nvidia cars

Nvidia is driving forward its ambitions in the fast-growing autonomous vehicle space with a powerful new mobile processor that pairs its latest Denver 2.0 CPU and Pascal-based GPUs to power deep-learning applications that will make cars smart enough to recognize and respond to obstacles such as bikers and pedestrians.

Nvidia engineers unveiled the details of the Tegra "Parker" processor this week at the Hot Chips 2016 show, noting the chip is aimed at automotive uses such as self-driving cars and digital cockpits in vehicles. Given the power and capabilities in the chip, it also could drive applications in other areas such as gaming and mobile devices.

Company executives have identified both artificial intelligence and autonomous cars as key growth markets, and have added to Nvidia's portfolio in the automotive space over the past several months. At the 2016 Consumer Electronics Show (CES) in January, CEO Jen-Hsun Huang introduced the Drive PX 2 (pictured), a computer designed for self-driving cars that he said had the massive amount of computing power needed for autonomous vehicles to find their way along complex and congested roads full of obstacles and challenges.

Such vehicles—which essentially are computers on wheels—must be able to collect and analyze in real time huge amounts of data from various controls systems and sensors, react accordingly and learn as they go, Huang said. Drive PX 2 delivers that compute power. Leveraging the parallel processing power of the vendor's GPUs is a key to future self-driving cars, the CEO said.

"Modern artificial intelligence and GPU breakthroughs enable us to finally tackle the daunting challenges of self-driving cars," Huang said in a statement at the time. "Nvidia's GPU is central to advances in deep learning and supercomputing. We are leveraging these to create the brain of future autonomous vehicles that will be continuously alert, and eventually achieve superhuman levels of situational awareness."

At Hot Chips, Nvidia engineers talked about the technology behind the Drive PX 2 platform, which uses two Parker processors and two GPUs based on Nvidia's Pascal architecture. The 16nm Parker includes two next-generation 64-bit Denver 2.0 CPU cores and four 64-bit ARM Cortex A57 CPUs, all of which are designed to work together in a heterogeneous multiprocessor configuration and are interconnected through a proprietary coherent fabric.

The Denver 2.0 CPU is a seven-way superscalar chip that supports the ARMv8 instruction set and is designed for improved performance and power efficiency over its predecessor. According to Nvidia officials, the chip provides 50 percent to 100 percent better multicore CPU performance than other mobile processors.

The new 256-core Pascal GPU in Parker enables the chip to run deep-learning inference algorithms for self-driving functionality and can run multiple high-resolution displays, including instrument and infotainment displays, officials said.

Combined, the CPUs and GPUs deliver up to 1.5 teraflops of performance needed for deep learning and artificial intelligence (AI) capabilities in self-driving cars. The chip includes hardware-level virtualization that supports up to eight virtual machines that can concurrently host multiple systems—such as infotainment and driver assistance systems—a hardware-based safety engine for fault detection and error processing, and multiple IO ports.

According to company officials, Drive PX 2 delivers 24 trillion deep-learning operations per second, providing supercomputer-level computing to self-driving cars. Parker-based cars communicate with cloud-based Pascal-powered supercomputers that can continually update the chips with new algorithms and data to enhance car accuracy and safety.

The Parker chip also supports decode and encode video streams up to 4K resolution at 60 frames per second, enabling automakers to use high-resolution in-vehicle cameras for object detection and 4K display panels in the infotainment systems.

More than 80 carmakers, top-tier suppliers and research centers are using the Drive PX 2 system, they said. That group includes Volvo, which has plans to test the Drive PX 2 system in its XC90 SUVs next year.