On March 19th, NVIDIA announced that its latest generation automotive super chip/vehicle computing platform, DRIVE Thor, released earlier this year, has garnered even more attention from customers, including five major Chinese automakers.
DRIVE Thor, tailored for the burgeoning demand for generative AI applications in the automotive industry, is an upgraded version of the previous DRIVE Orin, consolidating all functionalities into a centralized platform.
It offers rich cabin features, highly reliable high-level automated driving, and autonomous driving capabilities, catering extensively to electric vehicles, trucks, autonomous taxis, autonomous buses, and unmanned delivery vehicles, among others.
The specific technical details of the DRIVE Thor chip haven't been widely disclosed. Previously, all we knew was that it housed 770 billion transistors, featured a Grace CPU coupled with Hopper/Ada GPUs, supported FP8 data format, and boasted a floating-point performance of up to 2 PFlops, eight times that of Orin.
However, according to the latest remarks by Huang Renxun, DRIVE Thor has actually been upgraded to include GPUs based on the Blackwell architecture, specifically tailored for Transformer, LLM, and AIGC workloads. It seems the details weren't convenient to disclose earlier.
DRIVE Thor is expected to enter mass production as early as next year.
In the realm of new energy vehicle clientele, besides the announcements from companies like Xpeng and NIO about adopting DRIVE Thor, three additional Chinese automakers are joining the fray:
BYD:
Not content with just integrating the next-gen electric models on the DRIVE Thor platform, BYD is also planning to leverage NVIDIA AI infrastructure for cloud-based AI development and training technologies. They aim to utilize NVIDIA Isaac and NVIDIA Omniverse platforms to develop tools and applications for virtual factory planning and retail configurators.
GAC Aion HiPhi:
Their next-gen electric vehicles will embrace the DRIVE Thor platform, with mass production slated to commence in 2025, promising up to L4 level autonomous driving capabilities. Their current flagship model, the HiPhi GT, already boasts DRIVE Orin, supporting L2+ level high-speed autonomous driving.
Xpeng:
Positioning the DRIVE Thor platform as the "AI brain" for their next-gen electric vehicles, Xpeng will complement it with their in-house XNGP intelligent driving assistance system. This combination promises features like autonomous driving, parking assistance, and driver/passenger monitoring.
Moreover, WeRide is collaborating with Lenovo's Vehicle Computing unit to develop multiple commercial L4 level autonomous driving solutions based on the DRIVE Thor platform.
These solutions will be integrated into Lenovo's inaugural autonomous driving domain controller, the AD1, tailored for various urban-centric use cases, featuring functional safety, redundant safety design, and fusion scalability.
