Close Menu
AI

Technology wars: Huawei’s AI chip capabilities are under intense scrutiny after market leader Nvidia names the company as a potential rival

5gantennas.orgBy No Comments4 Mins Read


The Ascend 910B is considered to be the successor to the Ascend 910, which was launched three months later by Huawei in August 2019. Place on transaction blacklist According to the U.S. Department of Commerce. Dylan Patel, principal analyst at San Francisco-based semiconductor research firm Semianalysis, said the chip could compete with Nvidia’s A100 in terms of enhanced AI algorithms.
The Chinese flag flies near the logo above the headquarters of Semiconductor Manufacturing International Co. in Shanghai. The country’s top contract chipmaker is said to be the supplier of Huawei Technologies’ new artificial intelligence chipset, the Ascend 910B.Photo: Bloomberg
US sanctions since 2019 have restricted Huawei’s semiconductor development, dealing a severe blow to the company. smartphone work.but shenzhenThe company has been quietly beefing up its chip business by partnering with a variety of domestic suppliers, said people familiar with the matter who asked not to be identified due to the sensitivity of the matter.
Kirin 9000 series processors developed by HiSilicon, the chip design division of Huawei Technologies. Photographed from a Mate 60 Pro 5G smartphone in Ottawa, Canada’s capital, on September 3, 2023.Photo: Bloomberg
Upon the domestic launch of the Ascend 910, Huawei touted the chip at the time as the “world’s most powerful AI processor” made by the world’s top contract chipmaker. Taiwan Semiconductor Manufacturing Co., Ltd.using a 7nm process.
Huawei’s new AI chip appears to have arrived around the same time as the release of the Mate 60 Pro last August.China’s online search and AI giant Baidu orders 1,600 Huawei Ascend 910B chips According to a Reuters report in November 2023, citing sources, in the same month.
Two weeks before Reuters published its report, Chinese AI companies i fly tech announced the Feixing One computing platform based on Huawei’s Ascend chip. This means: iFlytek Spark 3.0The company’s latest large-scale language model (LLM) may have been developed on an AI chip. LLM is a technology used to train generative AI services such as: Chat GPT.

Huawei declined to comment on the matter.

Jensen Huang, co-founder, president and CEO of American chip design company Nvidia, attends a session at the World Government Summit in Dubai, United Arab Emirates, on February 12, 2024.Photo: Reuters
In an interview with technology media Wired last month, NVIDIA’s CEO said: jensen fan He described Huawei as “a really, really good company.”

“No matter what semiconductor processing technology we have, there are limits, but it is possible to build very large systems by integrating many of these chips,” Huang said.

Reuters reported last month that Huawei and SMIC have allocated more production capacity to AI chips as generative AI gains increased attention and US sanctions tighten over the past year.

One GPU distributor, who asked not to be named due to the sensitivity of the issue, said the Ascend 910B is “available to order, but supply is very tight at this time.”

Amid industry transformation, Huawei and ZTE showcase AI prowess at MWC Barcelona

The server used for AI training and equipped with eight Ascend 910B cards costs about 1.5 million yuan (US$208,395), according to another person familiar with the matter, which was quoted through black market channels. The price is about the same as the A100 server. also refused to be named.

Huawei has not commented on the Ascend 910B, despite reiterating that AI is a “key strategy” ahead of the company’s display at the four-day trade show. MWC Barcelona Last week in Spain.

Many analysts and industry experts are hesitant to comment on the Nvidia-Huawei showdown, but the U.S. chip designer has depth in GPUs and hopes developers will unlock the semiconductor’s full potential. He pointed out that it benefits from the software ecosystem CUDA, a computing platform that enables. .

“CUDA is resilient, and Nvidia has done all this hard work on its own and is reaping the benefits,” said Brian Colello, technology equity strategist at Morningstar. “Huawei and its software partners will need to build a comparable ecosystem.” [to Nvidia’s CUDA] When it comes to tools to build AI models. ”

SMIC and Huawei to be top beneficiaries of Chinese government funding this year

Despite lagging behind CUDA’s 2 million-strong registered developer list, Huawei is a key player in Ascend’s hardware and software connection platform, essential to unlocking AI computing power. It has its own computing architecture for neural networks.

Colello said Huawei may have to make something similar. [big] Invest in mainland China to strengthen software capabilities. He added that while Huawei will focus on chip design, other companies will likely work on software libraries.

“Huawei’s strength is not in its software stack,” said a Shanghai-based tech investor, who requested anonymity. “U.S. sanctions are limiting chip performance and production yields.”



Source link

Share.

Related Posts

Leave A Reply