Nvidia high -end GPU supply is limited?Autonomous driver controller related enterprises: not involved the limited chip

21st Century Business Herald reporter Zhao Yunfan Shanghai report

On September 1st, the self -driving concept company Desesiwei (002920.SZ) broke a decrease of 4.44%, a new low in the past month; China Science and Technology (300496) fell 3.22%, a new low in two months. The decline in the stock price of autonomous driving companies is suspected to be affected by the news within the day.

On the evening of August 31, the graphic chip giant Nvidia announced that the company had received an official US notice that if two high -end GPU chips of China and Russia exported two high -end GPU chips, A100 and H100 Essence

The application of the CPU+GPU chip architecture in the high computing autonomous driving scenario is more common. Will the supply of related companies be affected?

The reporter learned from Disaixi Securities Department that the company's products do not involve relevant restrictions.

Another autonomous driving company told reporters that restrictions are mainly in large intelligent AI systems and some supercomputing centers, which will not affect the import of car SOC chips.

Benefiting from the development of the intelligent automobile industry and the rapid start of the domain controller, the stock price of companies such as Desesiwei has skyrocketed in the past two years. Focus on investors.

The main domain controller products of Disaixi include the two models of IPU03 and IPU04. They all adopt the Nvida Orin architecture SOC, and the built -in GPU+CPU architecture.

At present, IPU03 products have been equipped with mass production models such as Xiaopeng P5/P7. The IPU04 has been equipped with the ideal L9, and the model has been mass -produced in the second half of this year.

The reporter learned that the Desesi IPU03 model is mainly equipped with Nvida Xavier model SOC chip, and Xavier's AI computing power is provided by the built -in GPU core chip, which adopts the older system architecture.

The IPU04 chip is mainly equipped with Nvidia Orin-X chip. The built-in AI computing power GPU uses a more advanced AMPERE architecture. The computing power is nearly six times higher than that of Xavier.

According to previous information, the limited GPU chip AP100 uses the AMPERE architecture, while the H100 uses a more advanced Hooper architecture.

According to a document submitted to the US Securities and Exchange Commission (SEC), Nvidia also stated that the permit requirements also include all Nvidia high -end integrated circuits in the future. The threshold of A100, and any system including these high -end circuits, is included in the scope of the permit restriction.

In this regard, a person in the chip industry told reporters that from the limit, restrictions are mainly in a relatively advanced super algorithm center, and the possibility of commercial GPU chips is limited.

- END -