In October 2017, NVIDIA introduced a groundbreaking addition to its DrivePX family of autonomous driving computing platforms at the GT Cup conference. The new platform, codenamed "Pegasus," was designed to push the boundaries of self-driving technology and was expected to be available to NVIDIA's R&D partners in Q2 2018.
Pegasus boasts an impressive 320 TOPS (Trillion Operations Per Second) of computing power, which is up to ten times more powerful than its predecessor, the DrivePX2. This performance boost comes from four key processors: two SoCs powered by NVIDIA’s Volta GPU architecture and two dedicated GPUs optimized for vehicle vision and deep learning tasks.
The development of a fully autonomous driving system requires not only immense computational power but also extensive input/output capabilities. As NVIDIA points out, future services like Robotaxi will need to process high-resolution data from multiple sensors—including cameras, LiDAR, and radar—in real time. This includes accurate localization, object detection, and route planning, all of which demand significantly more processing power than today’s most advanced vehicles.
Currently, no existing platform can meet these demands, making Pegasus a game-changer. Its I/O interface supports 16 pairs of sensors, including radars, cameras, and Optima, along with standard automotive interfaces like CAN, FlexRay, and Ethernet. The system also features a total memory bandwidth of 1TB, ensuring seamless data handling.
Beyond raw power, NVIDIA emphasized that Pegasus adheres to the ASIL-D level of the ISO 26262 functional safety standard, a critical requirement for automotive electronics. This means the platform incorporates multiple redundant systems to ensure reliability and safety in real-world conditions.
NVIDIA has also built a strong ecosystem around the DrivePX platform, with over 225 development partners, including companies like Zoox, OptimusRide, and Nutonomy, working on fully automated taxi solutions. In 2018, NVIDIA partnered with DHL and ZF to develop an unmanned logistics fleet using the ProAI system based on DrivePX.
Major automotive suppliers such as Autoliv, BOSCH, and Tesla have also committed to using the DrivePX platform, signaling its growing importance in the industry. To support developers, NVIDIA released the DriveIX SDK, which helps create features like driver face recognition, distraction detection, and interior automation, further enhancing the capabilities of autonomous vehicles.
Despite the advancements, achieving SAE Level 5 autonomy remains a long-term goal. Current systems are still limited to Level 2 or 3, and many technologies—such as 5G communication for vehicle-to-everything (V2X) interactions—are still in development. This makes it challenging to estimate the exact computing needs for full autonomy.
Unmanned vehicles are likely to first appear in mass transit services like taxis and buses, where controlled environments make testing and deployment easier. Public transport providers stand to benefit from reduced labor costs, while early adopters can manage the high initial investment and risks.
Finally, ecosystem management will play a crucial role in the success of autonomous driving. NVIDIA has built a vast developer network and collaborated with leading automakers and research institutions, creating a strong foundation for future innovation. By fostering open collaboration and supporting academic research, NVIDIA continues to shape the future of autonomous mobility.
Differential Mode Inductors,I type Inductors,Common Mode Inductors,Differential Mode Choke Inductor
Xuzhou Jiuli Electronics Co., Ltd , https://www.xzjiulielectronic.com