Earlier today, we told you about NVIDIA's new Ampere architecture, which forms the basis of the A100 GPU and the DGX A100 AI compute system. However, Ampere is also scaling down to be used in NVIDIA's DRIVE AGX platform for automakers.
NVIDIA's next-generation DRIVE AGX range is incredibly scalable through the use of the Orin SoC and 7nm Ampere GPUs. Starting at the very low-end, NVIDIA is promoting a windshield-mounted advanced driver assistance system (ADAS) that offers 10 TOPS of compute performance while consuming just 5 watts.
From there, NVIDIA can scale to a Level 2+ DRIVE AGX autonomous vehicle platform with 200 TOPS compute within a 45W power profile. Finally, there's the next-generation Drive AGX Pegasus platform for Level 5 autonomous robotaxis that combines two Orin SoCs with two Ampere GPUs providing a total of 2,000 TOPS compute at 800W (compared to 320 TOPS for the current Turing-based DRIVE AGX Pegasus). Vehicles powered by this advanced Pegasus platform would need absolutely no human intervention and could help revolutionize the transportation industry.
"The architecture offers the largest leap in performance within the eight generations of NVIDIA GPUs — boosting performance by up to 6x," NVIDIA writes in a blog post.
Interestingly enough, NVIDIA says that Orin will begin sampling next year, and it won't actually go into production until late 2022. So, auto manufacturers looking to adopt this new technology will have quite a wait until it will find its way into the hands of paying customers.
NVIDIA also announced EGX A100, which is an Ampere-based edge AI system for compute-intensive workloads and 5G applications (among other workloads). EGX A100 will be deployed in retail stores, hospitals and on factory floors according to NVIDIA, and that latter sector has been of keen interest to BMW in particular.
BMW announced today that it will be deploying EGX A100, and EGX Jetson AGX Xavier edge computers at its factories along with AI robots using NVIDIA's Isaac robotics platform. The German automaker is hoping to not only speed up the assembly process of vehicles and make it more efficient, but also help workers when it comes to sorting through the dizzying array of customization options available with current BMWs.
"BMW Group vehicles are offered to customers with an average of 100 different options, resulting in 99 percent of customer orders being uniquely different for each other. This creates an immense challenge for factory logistics," NVIDIA explains. "To optimize the enormous complexity of this material flow, autonomous AI-powered logistics robots now assist the current production process in order to assemble highly customized vehicles on the same production line."
According to NVIDIA, EGX A100 will be ready for customers by the end of 2020, while EGX Jetson Xavier NX is available now.