China autonomous driving leads Apple, tests NVIDIA

Chinese electric vehicle maker NIO and smart phone maker Xiaomi have ended the year with the announcement of new products challenging NVIDIA and Apple. This marks yet another advance in China’s efforts to develop its own semiconductor technology and eliminate dependence on imports.

On December 23, NIO revealed its first autonomous driving chip, which it claims is superior to the NVIDIA Drive Orin chips it is using now. The Shenji NX9031 SoC (System-on-Chip) will be used in the company’s new ET9 executive sedan, which was also presented at NIO Day 2023. Held in Xi’an, the event attracted more than 10,000 participants and attendees.

The ET9 is a long-wheelbase, four-door luxury EV priced at about 800,000 yuan, or $113,000 at the current exchange rate. Orders for the ET9 can be placed in China now, but deliveries are not scheduled to begin until the first quarter of 2025.

The first 5-nm Chinese automotive IC, the NX9031 will probably by fabricated by TSMC, Samsung Electronics or Intel Foundry Services. In theory, it could be made by SMIC, but that would be inefficient and more expensive. Under the sanctions against the Chinese semiconductor industry, SMIC does not have access to the advanced EUV lithography systems that IC foundries outside China use to make chips with 5-nm and finer processes.

The NX9031 features an Arm 32-core CPU (central processing unit), an NPU (neural processing unit), a graphics core and more than 50 billion transistors. It comes with low power, double data-rate LPDDR5X DRAM memory (probably Korean-made) and can process lidar (light detection and ranging) data.

An NPU, which mimics the human brain, deduces optimal solutions from available data. Lidar uses laser light to survey the environment, providing autonomous vehicles with 3D images of the road and traffic. The NX9031 will work with SkyOS, Nio’s vehicle operating system, which covers vehicle control, assisted driving, cockpit systems and mobile connectivity.

According to NIO CEO William Li, the computing power of the NX9031 is comparable to four NVIDIA Drive Orin SoCs, the standard configuration currently used in NIO electric vehicles. These 7-nm chips are capable of up to 254 TOPS (trillion operations per second) each, or 1,016 TOPS in total. That is sufficient for smart driving today, says Vice President Bai Jian, but not for the next generation of autonomous vehicles.

NVIDIA has come to the same conclusion. Drive Thor, the successor to Drive Orin, is scheduled to go into production in 2025. It will have twice as much computing power as Drive Orin. What does this mean? NVIDIA Vice President Danny Shapiro explains:

“If we look at a car today, advanced driver assistance systems, parking, driver monitoring, camera mirrors, digital instrument cluster and infotainment are all different computers distributed throughout the vehicle. In 2025, these functions will no longer be separate computers. Rather, Drive Thor will enable manufacturers to efficiently consolidate these functions into a single system, reducing overall system cost.”

NIO established its IC design team in 2020. Led by Bai Jian, its purpose is to develop an independent smart driving capability including sensors, autonomous driving algorithms and now SoCs. Bai was previously an executive at Chinese smart phone makers OPPO and Xiaomi.

Xiaomi beats Apple to the draw

On December 28, Xiaomi revealed its first electric vehicle before thousands of people at the China National Convention Center in Beijing. At the event, CEO Lei Jun said that his company’s goal is to produce a dream car that can rival Porsche and Tesla.

The Xiaomi SU7 (SU stands for Speed Ultra), a four-door electric sedan designed by professionals who previously worked for BMW and Mercedes-Benz, is manufactured by BAIC (Beijing Automotive Industry Corp). It is equipped with NVIDIA Drive Orin SoCs for assisted and autonomous driving and Xiaomi’s own operating system. Videos show the SU7 avoiding obstacles on the street and parking without a driver.  

Deliveries of two- and four-wheel drive models are expected to begin within the next few months. Prices have not yet been announced, but benchmarking against the Tesla’s Model S and Porsche’s Taycan Turbo suggests that they might be as high as 700,000 to 900,000 yuan range – about the same as the NIO ET9.

Xiaomi plans to invest about $10 billion in its auto business in the next 10 years. “By working hard over the next 15 to 20 years,” says CEO Lei, “we will become one of the world’s top five automakers, striving to lift China’s overall automobile industry.”

Xiaomi’s plans appear to put Apple to shame. Last September, MacRumours reported that a semi-autonomous Apple Car equipped with neural processors still might be launched in 2026 – but that there have been so many delays since the project began in 2014 and so little information provided by management that the timeline is up in the air

Apple CEO Tim Cook has said “We’re focusing on autonomous systems. It’s a core technology that we view as very important. We sort of see it as the mother of all AI projects. It’s probably one of the most difficult AI projects actually to work on.” But it was in 2017 when he said that.

In any case, Xiaomi has beaten Apple to the draw, and so has Huawei, which announced its entry into the electric vehicle market in 2021. Now it is reported that Huawei plans to establish hundreds of new EV sales and service outlets in China over the next two years.

Under sanctions that have proved far from crippling, Huawei has developed its own automotive ICs and operating system. The Ascend chipset in its Mobile Data Center is capable of 352 TOPS, which enables Level 4 High Driving Automation. That means completely autonomous operation under certain conditions such as a defined route, highway driving or parking. 

In the meantime, NVIDIA continues to supply more than a dozen Chinese automakers including NIO, Xiaomi, BYD, DENZA, Human Horizons, Ji Yue, Xpeng and ZEEKR (which is owned by Geely). ZEEKR was the first Chinese automaker to announce that it would adopt Drive Thor. Xpeng has reportedly asked NVIDIA to provide it with a custom designed version of Drive Thor.

At the end of November, it was reported that NVIDIA is looking to hire more than 20 specialists for its autonomous driving development team in China. That team is led by Wu Xinzhou, who joined NVIDIA last August. Before that, he was vice president of autonomous driving at Xpeng.

NVIDIA seems likely to maintain a strong position in the China’s autonomous driving market – unless the US Commerce Department decides that Drive Thor is too good for China. If that happens, Chinese companies should be able to pick up the slack.

Follow this writer on X: @ScottFo83517667


This website uses cookies. By continuing to use this site, you accept our use of cookies.