NVIDIA has unveiled DRIVE Thor, a central vehicle computer set up to make the driving experience safer and easier for self-driving vehicles.
The new offering combines infotainment, automated driving and parking into one cost-effective system.
The DRIVE Thor is said to achieve up to 2,000 teraflops of performance and unify smart functions, including automated driving and assistance, parking, driver and passenger monitoring, digital instrument cluster, in-vehicle infotainment (IVI), and rear-seat entertainment.
It also includes next-generation super chipsets and NVLink-C2C chip interconnect technology, which comes with advanced AI capabilities first introduced in the NVIDIA Hopper Multi-Instance GPU architecture.
The company also focuses on efficiency and effectiveness. DRIVE Thor technology with MIG support for graphics and computing enables IVI and advanced driver assistance systems to operate with field isolation, allowing time-critical concurrent processes to run uninterrupted.
“Advances in accelerated computing and artificial intelligence are moving at the speed of light,” says Jensen Huang, founder and CEO of NVIDIA.
“DRIVE Thor is the superhero of centralized computing, with ultra-fast performance to deliver scalable, secure, software-defined supercomputers on wheels.”
Looking at the market, Geely-owned auto maker ZEEKR has announced that it will integrate DRIVE Thor on the central car computer for the next generation of smart electric vehicles, and start production in early 2025.
“ZEEKR users demand a luxurious experience that includes the latest technology and safety features,” says An Conghui, CEO of ZEEKR.
“NVIDIA DRIVE Thor will support our mission to provide the latest technologies that meet the needs of our customers and ensure ZEEKR remains at the forefront of tomorrow’s innovation.”
Typically using similar technology, dozens of electrical control units are distributed throughout the vehicle to operate individual functions.
With DRIVE Thor, manufacturers can integrate many functions into a single system on a chip (SoC), easing supply constraints and simplifying vehicle design development, resulting in significantly lower cost, lower weight and fewer cables.
Another advantage of DRIVE Thor is its 8-bit floating point (FP8) capability. Developers often lose neural network accuracy when switching from 32-bit FP data to 8-bit integer format. NVIDIA says DRIVE Thor features 2,000 teraflops of FP8 resolution, allowing for 8-bit transitions without sacrificing accuracy.
DRIVE Thor is also the first AV platform to include the Inference Converter Engine, a new component of Tensor Cores within NVIDIA GPUs.
Using this engine, DRIVE Thor can accelerate the inference performance of transformer deep neural networks by up to 9x, which is critical for supporting complex AI workloads associated with autonomous driving.
The DRIVE Thor SoC and AGX board has been developed to comply with ISO 26262 standards, and the software suite is designed to comply with both ISO 26262 and ASPICE.
“The shift to software-defined vehicles with centralized electronic architectures is accelerating, leading to the need for more powerful and more energy-efficient computing platforms,” says Sam Abuelsamid, Principal Research Analyst, Guidehouse Insights.
“Virtualization, high-speed data transfer, and the massive processing performance of NVIDIA DRIVE Thor can enable safer vehicles, better user experiences, and potentially new revenue streams.”
This technology will be available for automakers’ 2025 models and will accelerate production roadmaps by bringing higher performance and advanced features to market in the same timeline.