Wu Ji will push 250TOPS computing power on-board smart chip

“We believe that large computing power and versatility will be two important trends in smart driving chips or smart driving chips.” A few days ago, Wang Ping, CEO of Cambrian Traveling Song (Nanjing) Technology Co., Ltd., will supply new energy and smart cars globally in 2021. The Chain Innovation Conference announced that it will launch the first SoC smart chip product based on 7nm advanced manufacturing process and 250TOPS computing power in 2022. In the second half of 2023, it will pass various vehicle regulatory certifications and achieve vehicle SOP.

It is important to know that in the “Smart Vehicle Innovation Development Strategy” jointly issued by China’s multiple ministries and commissions last year, it was clearly stated that by 2025, intelligent vehicles with conditional self-driving will reach mass production, and the market for highly-autonomous intelligent vehicles will be marketed under specific circumstances.化 Application. Data shows that the market penetration rate of my country’s L2 intelligent networked passenger cars has exceeded 15%, and the development of automobile intelligence has become unstoppable. With the continuous advancement of automotive intelligence, the industry’s demand for automotive chips, especially automotive AI chips, is increasing.

Wang Ping analyzed and pointed out that smart cars currently have higher and higher requirements for computing power. On the one hand, the number of sensors on new cars has increased sharply, and the data collected in the future will be extremely large. On the other hand, the algorithms are more mature and more complex. There is a greater demand for power. From the low-level L2, L3 to L4, the demand for computing power has increased geometrically. At this stage, new cars have embedded computing power to more than 1000 TOPS.

Based on this, he believes that in the future, large computing power and versatility are two important trends for smart driving chips or smart driving chips.

But at the same time, he further analyzed and pointed out that in order to realize such a large computing power universal intelligent driving chip, four challenges need to be solved:

The first one is the challenge of chip system architecture. The chip system architecture is very complex. Chips above 200TOPS have very high requirements for memory access and need to support ultra-high bandwidth. Therefore, the complexity of system architecture design has greatly increased. In this regard, my country’s design talents are very scarce.

The second is the challenge of general-purpose software AI software stack. In the process of continuous evolution, the laser point cloud algorithm and the multi-sensor fusion algorithm are still in rapid iteration. Therefore, the constantly changing algorithm requirements need to pass the OTA mode, and a general hardware architecture and software stack are required to support the continuous upgrade of the algorithm.

The third is the challenge of large-scale chip engineering. The size of a chip with a large computing power is very large, and a large-size chip brings severe challenges to packaging, power supply and thermal management, cost control, and yield.

The fourth is the challenge of advanced technology platforms. At this stage, my country does not yet have advanced automotive-grade chip manufacturing processes such as 7 nanometers and 5 nanometers.

We must know that affected by multiple factors such as the epidemic and the later American cold wave, the Japanese earthquake, the Malaysian epidemic, etc., the core shortage of the automotive industry has swept the world, and mainstream domestic car companies such as Weilai and SAIC Volkswagen have also not been spared. The small chip has become one of the biggest obstacles currently restricting the production of car companies. Based on this, seeking domestic substitution has become the main theme in the automotive semiconductor field this year, but a cruel reality is that China’s independent automotive chip industry only accounted for 4.5% of the global scale in 2019, and the rate of automotive chip self-research in the domestic automotive industry Less than 10%.

As a result, related companies such as SMIC, Ziguang Guowei, Cambrian, Horizon, and Xinchi have attracted much attention.

In January of this year, Cambrian established a wholly-owned subsidiary, Cambrian Xingge (Nanjing) Technology Co., Ltd. (hereinafter referred to as Xingge Technology), to formally develop automotive smart chip related businesses. Subsequently, on June 22, Cambrian announced that Xingge Technology intends to increase its registered capital by 170 million yuan and introduce investors. The Cambrian senior management team will also inject capital. The main purpose of the capital increase is to develop in-vehicle smart chips.

On July 8, at the WAIC Smart Chip Forum, Chen Tianshi, the founder and CEO of Cambrian, the first stock of AI chips on the Sci-tech Innovation Board, responded for the first time to the industry’s conjecture about Cambrian’s entry into automotive smart chips, and disclosed that they are designing a computer. Lichao’s 200TOPS intelligent driving chip, which inherits the Cambrian integrated, unified and mature software tool chain, adopts a 7nm process and has an independent safety island.

This time, detailed information about the Cambrian car smart chip was officially disclosed.

Facing the future, Wang Ping said that the Cambrian will continue to promote the collaboration of cloud-side vehicles. It will also launch a large computing power, open and universal smart driving chip that can support the large computing power requirements of complex models of high-level smart driving in the future, and it can also support the continuous iteration of algorithm models. Assist the realization of intelligent driving with high efficiency and rapid iteration.

*Copyright statement: This article is an original article by Gasgoo. If you want to reprint, please abide by the relevant regulations of the reprint instructions. Anyone who violates the reprinted instructions will be held accountable by Gasgoo Motors in accordance with the law!

The Links:   G101EVN014 FZ1600R17HP4-B2