Jensen Huang, NVIDIA founder, and CEO revealed the NVIDIA DRIVE Thor, a highly superior superchip to achieve autonomy in-vehicle computing at a faster rate. The brand new vehicle-grade SoC, or system-on-a-chip, is constructed on the latest CPU and GPU advancements to provide 2,000 teraflops of performance while limiting the entire system’s costs.
DRIVE Thor is the successor to the company’s NVIDIA DRIVE Orin, comprising the latest computing technology to speed up industry deployment of smart-vehicle technology, focusing on automakers’ 2025 models.
The first customer of NVIDIA’s new next-gen platform will be Geely-owned premium EV maker ZEEKR. The company will begin production starting in 2025.
NVIDIA’s DRIVE Thor combines traditionally dispersed vehicle processes, including digital clusters, infotainment, parking, and assisted driving. The company anticipates higher efficiency in development and prompter application iteration.
The DRIVE Thor superchip can be configured by manufacturers in several ways, such as dedicating all of the platform’s 2,000 teraflops to the autonomous driving pipeline. The new platform will also allow for utilizing a fragment for in-cabin AI and infotainment and another for assisting drivers.
Like the NVIDIA DRIVE Orin, DRIVE Thor employs the productivity of the NVIDIA DRIVE software development kit, designed to be ASIL-D functionally safe. DRIVE Thor is built on a scalable architecture, allowing developers to transfer their previous software development to the next-gen platform seamlessly.
Additionally, DRIVE Thor produces an extraordinary leap in deep neural network accuracy along with raw performance.
DRIVE Thor honors the first inclusion of a transformer engine in the autonomous vehicular platform family. The transformer engine created by NVIDIA is a new component of the company’s GPU Tensor Core. Transformer networks process video data as a single perception frame, boosting the compute platform to process additional data over periods.
Using FP8, or 8-bit floating point precision, the system-on-a-chip presents a new data type for the automotive industry. Developers of autonomous vehicles see an accuracy loss when transferring from 32-bit floating point to 8-bit integer data formats. FP8 precision reduces the struggle of this transition, ensuring that developers can move data types without sacrificing accuracy.
NVIDIA DRIVE Thor utilizes updated ARM Poseidon AE cores and, in turn, makes the processor one of the highest performing in the industry.
The system-on-a-chip is adept at computing multi-domains, partitioning tasks for autonomous driving, and in-vehicle infotainment. Isolating the multi-compute domain allows simultaneous time-critical processes to run without interruption. The vehicle can simultaneously run Linux, QNX, and Android operating systems on a single computer. These functions are regulated by tens of electronic control units dispersed throughout a vehicle. Instead of relying on these distributed ECUs, manufacturers can directly reduce vehicle functions using DRIVE Thor’s capability to separate specific tasks.
Everything from vehicle displays to sensors can connect to a single SoC, streamlining automakers’ extremely complicated supply chain. Clients can employ one DRIVE Thor SoC or combine two via the latest NVLink-C2C chip interconnect technology to function as a monolithic platform that operates a single operating system.
This ability supplies automakers with the computing space and flexibility to construct software-defined vehicles that are constantly upgradeable through protected, over-the-air updates.