Huawei develops AI-chip that will challenge Nvidia 09: 37

Huawei is preparing to test a new processor to teach Artificial Intelligence Ascent 910D. The company hopes that this chip will replace the high -performance solutions of the American Techno giant Nvidia in the local Chinese market. About it reports Broadcast Wall Street Journal (WSJ).

According to sources, Huawei is already negotiating with Chinese technology companies to test the new AI processor. Chip’s first examples are expected to be available for verification at the end of May. Development is at an early stage and a complete assessment of performance will require a series of tests. However, Huawei expects the Ascent 910D to overcome the widely used capabilities to teach NVIDIA H100 CHIP II systems.

This development takes place on the basis of tightening US export restrictions on the supply of advanced NVIDIA developed chips, such as the H20 model. These restrictions have created additional opportunities to strengthen China’s position in the semiconductor industry despite sanctions for Huawei and other local producers. Huawei has become one of the leaders in the creation of indigenous analogs of Nvidia chips, and previously released the flagship smartphone Mate 60 based on the Chinese processor and showed the ability to skip American restrictions. In 2023, this device’s presentation overlap with US Minister Gina Raymondo visited China, caused significant concern in Washington.

This year, Huawei plans to put more than 800,000 rise chips of previous generations (910B and 910c models) to large state telecommunication companies and AI’s special developers, including Bayta, the owner of the Tiktok platform. After improving restrictions on the supply of the NVIDIA H20, some of these customers already discuss the likelihood of increasing the order volume to the Ascend 910c version.

Although Chinese chips are still behind Western analogs for several years, Huawei and other local manufacturers are actively looking for ways to improve the general performance of systems by combining many chips. For example, the Cloudmatrix 384 Information Processing System is presented in April, according to analysts, in some problems, despite potentially higher energy consumption, it is presented to exceed NVIDIA’s flagship solutions.

According to experts, even if the performance of a single Huawei graphics processor is lower than the latest NVIDIA models, the total number of chips used in the system can compensate for this gap. Analysts also accept the existence of problems with energy efficiency, but they believe that China is not critical for its current priorities.

Before Samsung stated The problem with the production of electric vehicles in the United States due to Trump tasks.

What are you thinking?



Source: Gazeta

Popular

More from author