Nvidia held its virtual GTC 2022 conference last week, and as expected there is always several new product introductions — yet this year seems like a record.
Nvidia held its virtual GTC 2022 conference last week, and as expected there is always several new product introductions — yet this year seems like a record. This column will provide perspectives on the announcements, with a focus on automotive-related product announcements, including Nvidia’s Drive Hyperion 9 platform, Drive Orin and more.
Before this analysis, it is useful to understand how Nvidia is positioned for success in the automotive industry:
NVIDIA GTC 2022 key announcements
The following table summarizes the announcements during Nvidia CEO Jensen Huang’s GTC keynote and the many press releases that Nvidia published. The table primarily looks at products that impact the auto industry. Nvidia’s investor presentation from Mar. 22 has additional information and many of the keynote slides are available in PDF format.
The H100 Hopper GPU chip replaces the Ampere 100 as Nvidia’s flagship GPU. The H100 will offer 3x to 6x the performance of the A100 depending on floating point format. The H100 has exceptional memory bandwidth at 3 TB/s, and nearly 5 TB/s of external bandwidth. The H100 is expected to be used in future autonomous vehicle (AV) software platforms.
The DPX instruction set built into H100 GPUs will increase dynamic programming algorithm speeds in multiple industries, boosting workflows for robotics, disease diagnosis, quantum simulation, graph analytics and routing optimizations.
Nvidia also unveiled its latest “superchip” Grace, which has two CPU chips connected over a 900 GB/s NVLink chip–to–chip interconnect to make a 144–Arm–core CPU with 1 TB/s of memory bandwidth.
Grace has an interesting architecture that can handle enormous data loads to and from the chip and other systems. This is advantageous for AI training and inferencing at cloud facilities. It is likely to become valuable in AV software platforms in next generation hardware.
Nvidia’s Drive Map will provide survey–level ground truth mapping coverage to 500,000 kms of roadway in North America, Europe and Asia by 2024. This map will be updated and expanded with data from millions of vehicles. It combines the accuracy of DeepMap survey mapping with continued updates of AI–based crowdsourced mapping.
Drive Map uses multiple localization layers of data for use with camera, radar and lidar. The camera localization layer consists of map attributes such as lane dividers, road markings, road boundaries, traffic lights, signs and poles.
The radar localization layer is an aggregate point cloud of radar returns. It is especially useful in reduced lighting and poor weather conditions. Radar localization is also useful in suburban areas where typical map attributes are unavailable.
The lidar voxel layer provides the most precise and reliable representation of the environment. It builds a 3D representation of the surrounding at 5–centimeter resolution.
Drive Map workflows are centered on Omniverse, where real-world map data is loaded and stored. Omniverse is building an Earth–scale representation of the digital twin that is continuously updated and expanded by survey map vehicles and millions of passenger vehicles.
Omniverse tools can generate detailed maps that are converted into a drivable simulation environment that can be used with Drive Sim.
Drive Hyperion 9
Nvidia announced Drive Hyperion 9 as its next generation platform for automated vehicles and AVs. Hyperion 9 is scheduled for 2026 production vehicles. It is built on multiple Drive Atlan SoCs for use in AV driving and in–cabin functionality. Hyperion is compatible across generations, with the same computer form factor and Nvidia DriveWorks APIs.
Partners can leverage current investments in the Drive Orin platform and migrate to Drive Atlan and future versions. Hyperion 9 is a programmable platform with redundant Atlan computers and sensors for Level–3 and Level–4 autonomous driving. It can also provide Drive Concierge features.
Hyperion 8 is the current generation based on Drive Orin SoC and Drive IX software. Nvidia announced the start of production of its Drive Orin AV computer. In March alone, over 25 auto OEMs have adopted the Drive Orin with some customers stating production of software-defined vehicles in 2022.
Mercedes–Benz will ship Hyperion 8 systems in 2024 followed by Jaguar Land Rover in 2025.
Nvidia unveiled Omniverse Cloud at GTC 2022. Omniverse provides designers and developers access to a software platform to create digital twins and similar simulations. Omniverse Cloud users can leverage Omniverse Create, an app for technical designers, to interactively build 3D worlds in real time; and Omniverse View, an app for non-technical users.
Among Omniverse Cloud’s services is Nucleus Cloud, a simple “one–click–to–collaborate” sharing tool to access and edit large 3D scenes from anywhere, without having to transfer massive datasets.
Over 150,000 copies of Omniverse have been downloaded. Nvidia is primarily focused on industrial applications of Omniverse such as robotics, automotive, factories and similar developments.
Nvidia announced OVX, a computing system architecture designed to power large–scale digital twins. OVX will be available later in 2022.
OVX servers consists of eight Nvidia A40 GPUs, three Nvidia ConnectX–6 Dx 200 GB/s NICs, 1TB system memory and 16TB NVMe flash storage. The backbone of OVX is its networking capabilities including the NVIDIA Spectrum–4 high–performance networking infrastructure platform that was announced. The Spectrum–4 has 100 billion transistors.
NVIDIA automotive overview
The following table has a short overview of Nvidia’s platforms that are used in the auto industry. The design wins, current revenue and future revenue potential are also included:
Key auto platforms
Nvidia has a growing portfolio of software platforms for developing and deploying software–defined vehicles. A summary of auto platforms is listed in the above table, and most were described in the previous sections.
Nvidia Drive Chauffeur is built on Nvidia Drive Orin and the Nvidia Drive SDK. It features perception, mapping, planning layers and diverse deep neural networks (DNN). The DNNs are trained on real–world driving data and synthetic data for both highway and urban traffic scenarios. These perception outputs can be used for both autonomous driving and mapping to deliver a personal chauffeur for your daily drive.
Drive Concierge is a digital assistant built on Nvidia Drive Orin, Drive IX, and Omniverse Avatar to deliver an interactive user experience. It uses conversational AI, natural language understanding and recommendation engines to meet passenger’s requests. It works closely with Drive Chauffeur to enhance the experience inside the car. Drive IX is the integrated vision, voice, and graphics user experience.
EV, robotaxi and autonomous truck design wins
Nvidia has done especially well with Chinese battery electric vehicle (BEV) startups such as Nio, Li Auto and XPeng. These startups designed software–defined vehicles and are leveraging Nvidia’s software platforms. Some of these startups have monthly BEV sales in the $10,000 to $20,000 range with strong growth.
Regarding robotaxi design, Nvidia is the clear leader in supplying AV chips and AV software platforms to the robotaxi segment. Most of the robotaxi companies are using Nvidia Drive platforms with eight companies publicly listed as Nvidia customers. They are in testing mode with low current volumes and are years away from market takeoff.
Similar to robotaxi design, Nvidia is the leading AV chip and software platform supplier for autonomous trucks, though they are currently in testing and trial phases only — hence low volume users. Volumes in thousands per customers may happen around 2025.
OEM design wins
The traditional OEM brands have future volume potential for using Nvidia Drive platforms in new software–defined vehicle models as they are introduced. However, these OEMs will take five or more years to transition all their models to software–defined vehicles.
Mercedes–Benz will do so by 2024 and Jaguar Land Rover will get there by 2025. Eight more OEMs will use Nvidia Drive platforms for some of their domain ECUs, but with little public information.
Nvidia auto revenue
Auto revenue is a small portion of Nvidia’s total revenue — only 2.1 percent in 2021 for a total of $556 million. Its auto revenue declined from $700 million in 2020 due to OEM production restrictions as a result of chip shortages.
Nvidia’s revenue pipeline for the next six years is impressive at $11 billion compared to $8 billion in 2021. This indicates that Nvidia is achieving many design wins for its Drive platforms, such as those within its auto product segments including cars, trucks, robotaxis and EVs.
Nvidia’s opportunities in the auto industry are increasing rapidly. Its six-year automotive revenue pipeline is currently at $11 billion, an increase of $3 billion from a year ago. There are multiple reasons for this rapid growth:
Nvidia is using its software–centric chips and growing collection of software platforms to realize strong growth in the auto industry’s software–defined vehicle era. Nvidia is especially competent in creating SDKs that its customers can use to create their own application software platforms — for automotive and other industries.
This article was originally published on EE Times.
Egil Juliussen has over 35 years’ experience in the high-tech and automotive industries. Most recently he was director of research at the automotive technology group of IHS Markit. His latest research was focused on autonomous vehicles and mobility-as-a-service. He was co-founder of Telematics Research Group, which was acquired by iSuppli (IHS acquired iSuppli in 2010); before that he co-founded Future Computing and Computer Industry Almanac. Previously, Dr. Juliussen was with Texas Instruments where he was a strategic and product planner for microprocessors and PCs. He is the author of over 700 papers, reports and conference presentations. He received B.S., M.S., and Ph.D. degrees in electrical engineering from Purdue University, and is a member of SAE and IEEE.