Nvidia unveils new artificial intelligence chip for self-driving cars

Nvidia has introduced a Drive Thor chip with artificial intelligence for self-driving cars. The processor is based on the new Hopper graphics architecture and is able to speed up the operation of artificial intelligence algorithms using special transformation engines.

Nvidia plans to launch the drive Thor in production in 2024, and next year self-driving cars equipped with it will enter the market.

"These processors can scale to full autonomy. We are talking about Level 4 or 5 autonomy, in which cars can be driven without human intervention or without their presence at all," said Danny Shapiro, vice president of Nvidia.

The Drive Thor chip will take the place of Drive Atlan in the company's model range. The drive Atlan processor was planned to launch in 2024, but now the company has abandoned these plans. Thor's performance is 2000 teraflops-twice as fast as Atlan's, and eight times faster than the existing Orin processor.

With 77 billion transistors, Thor will be one of the most sophisticated processors on the market. In addition to using Hopper GPUs for normal computing tasks, it borrows CPU cores from the NVIDIA Grace processor. It also uses gaming GPU technologies — the Ada Lovelace architecture. Thanks to this, automakers will be able to replace more expensive and energy-intensive small chips with one powerful system on a chip, as well as significantly simplify the process of updating automotive software by air.

The Thor processor can also be used for robots and medical equipment. It will be able to run three operating systems simultaneously — Linux, QNX, and Android — for various components of the car's computing environment. This technology of splitting the computing power of the processor ensures that lower-priority tasks, such as the operation of the infotainment system, do not interfere with priority security tasks.