According to Nvidia, who is exhibiting its GPU technology at the GTC event, the AI processor will be in automobiles by 2025. Nvidia, a company that designs chips, unveiled a new processor on Tuesday called Drive Thor, which it believes to fuel the revolution in driverless vehicles. According to Danny Shapiro, vice president of Nvidia's automotive division, Thor processors should debut in 2024 for vehicles that will hit the roads in 2025, starting with Chinese automaker Zeekr's 001 EV. To better handle the artificial intelligence software that is essential to self-driving cars, they are built on Nvidia's new Hopper graphics processing unit. It will undoubtedly go to Level 4 or Level 5 self-driving capabilities, in which automobiles can drive themselves even when no human occupants are present or paying attention, according to Shapiro.
Atlan, a chip that Nvidia had planned for 2024, was scrapped in favor of Thor, which runs AI software at 2 quadrillion operations per second, eight times faster than Orin and twice as fast as Atlan. One essential Hopper component is included in Thor: the capacity to speed the transformers, a potent AI technique. Nvidia also anticipates lower-end Thor versions for less ground-breaking driver-assist features like automatic emergency braking and lane keeping.
The demand for processors and other semiconductor chips from automakers for driver assistance, infotainment, and the electronic control units that regulate everything from engine combustion to GPS navigation is large and growing. There are 8,000 semiconductor components in each Porsche Taycan. The new market is paying well for chip designers. Automotive chip orders for Nvidia are $11 billion, while orders for Qualcomm, a major rival, total $19 billion.
Thor will be a sizable processor with 77 billion transistors, if not the largest one available. However, according to Nvidia, it will enable manufacturers to replace a bulkier, more expensive, and more power-hungry collection of smaller processors. For ordinary computing workloads, it also uses Hopper GPUs and CPU cores from Nvidia's 2023 Grace processor. It also uses the Ada Lovelace architecture, the newest GPU technology from Nvidia for gaming and design. According to Huang, the design will make it simpler for automakers to upgrade their vehicles' software over the air. Tesla has long held a significant technological advantage in that field.
According to Huang, Thor will also be utilized for robots and medical devices. Additionally, it will be able to run three operating systems at once for distinct components of the in-car computer environment: Linux, QNX, and Android. According to Nvidia, partitioning technology makes sure that less important activity, like infotainment, doesn't interfere with the vital safety-related activities.
Those chips become even more crucial with driverless vehicles, which have been promised for years but are still in the testing phase. When it comes to driverless vehicles, "the industry has realized that it's a far more difficult undertaking than previously assumed," Shapiro added. Nobody is prepared to unleash these vehicles into the wild until there is more computation because safety is crucial.