You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
@Fred Chen , any hints on why ? My suspicion is that wafer capacity for Kirin 9XXX has been diverted to AI targeted wafers. SMIC 7nm/5nm continues to be very limited and expensive (low yields, complex process)
@Fred Chen , any hints on why ? My suspicion is that wafer capacity for Kirin 9XXX has been diverted to AI targeted wafers. SMIC 7nm/5nm continues to be very limited and expensive (low yields, complex process)
Zhang Ping-An, head of Cloud Computing at Huawei, said "3nm, 5nm definitely could not be obtained" by China, which seems to conflict with many earlier reports unless they could be referring to TSMC's 3nm and 5nm, but he also said even solving 7nm was indeed an achievement. So possibly, they are stuck, and we may not see process advances later this year.
March 24 (Reuters) - China has introduced guidelines to phase out U.S. microprocessors from Intel and AMD from government personal computers and servers, the Financial Times reported on Sunday.
The procurement guidance also seeks to sideline Microsoft's Windows operating system and foreign-made database software in favour of domestic options, the report said.
Government agencies above the township level have been told to include criteria requiring "safe and reliable" processors and operating systems when making purchases, the newspaper said.
China's industry ministry in late December issued a statement with three separate lists of CPUs, operating systems and centralised database deemed "safe and reliable" for three years after the publication date, all from Chinese companies, Reuters checks showed.
The State Council Information Office, which handles media queries for the council, China's cabinet, did not immediately respond to a faxed request for comment.
I’m betting that they are encountering this issue - they can‘t make enough Kirin 9XXX’s if they want to make Huawei 910B AI chips.
Huawei just retasked a factory to prioritize AI over its bestselling phone
Demand for Huawei’s AI chip has grown.
Huawei makes both its Ascend AI chip and the Kirin chip, which powers the Mate 60, in one facility. However, production in the plant has been low, people familiar with the matter told Reuters, so the company now plans to prioritize the AI chip.
It wants to improve production of its Ascend AI chips.
www.theverge.com
The 910B is fairly impressive when compared to last generation A100’s, per recent claims, but I haven’t see any specifics on training benchmarks. But I’m guessing large die size combined with challenging yields are sucking up all the wafer capacity.
According to a report from South China Morning Post, Wang Tao, Executive Director, Chairman of the ICT Infrastructure Managing Board at Huawei, recently stated at the World Semiconductor Conference and Nanjing International Semiconductor Expo that Huawei’s advanced Ascend 910B AI chip achieves an efficiency of up to 80% compared to NVIDIA’s A100 when training large-scale language models. In terms of specific test performance, it surpasses NVIDIA’s A100 AI GPU by 20%.
According to a report from South China Morning Post, Wang Tao, Executive Director, Chairman of the ICT Infrastructure Managing Board at Huawei, recent...