Original link: https://www.williamlong.info/archives/6911.html
On August 31, the U.S. government ordered chip maker NVIDIA to stop selling some high-performance GPUs to China. According to a Reuters report, another AMD (Super Micro Semiconductor) also claimed to have received a related ban.
On August 31, NVIDIA issued an announcement saying that as early as August 26 this year, the U.S. government notified NVIDIA that it had implemented new licensing requirements for the company’s future A100 exports to China (including Hong Kong) and Russia, and the upcoming H100 chip. ,Effective immediately.
The new licensing requirements may affect the company’s ability to complete H100 development in a timely manner or to support existing A100 customers, the announcement said. In addition, the company is reaching out to customers in China and seeking to meet their planned or future purchases of the company’s data center products with products that are exempt from the new license requirements. If a customer requires a product covered by the new licensing requirements, the company may apply for a license for the customer, but there is no assurance that the U.S. government will grant any waiver or license to any customer, or that the U.S. government will take prompt action.
Zhu Jing, deputy secretary-general of the Beijing Semiconductor Industry Association, told the Economic Observer.com that according to the news, the blocked products are high-end GPUs with sufficient double-precision computing power. For NVIDIA, it is based on the A100 computing power, which is more advanced than it. All graphics card products are limited.
According to a Reuters report, an AMD spokesman told Reuters it had received new licensing requirements that would prevent its MI250 (6nm process, high-precision computing chip) chips from being exported to China, but the company believed its MI100 chips would not affected.
The reporter asked AMD and NVIDIA for further inquiries, but no response has been received as of press time. Shares of AMD and NVIDIA fell 2.38% and 2.42%, respectively, in after-hours trading.
Zhu Jing said that this time the United States should curb domestic advanced computing (including supercomputing and intelligent computing), and China’s ongoing strategy of east and west computing will also be temporarily affected because it needs the support of computing power infrastructure. In addition, the affected area will temporarily extend to the Internet field.
High performance GPU limits
On the evening of August 31, there was market news that AMD informed it that it had received an urgent notice from the headquarters: 1. 2. Suspend the shipment of GPU cards MI100 and MI200 to all data centers in China; 3. Statistics of Ml 100 shipments in China; Statistics of MI 200 shipped customers in China and delivery details AMD analysis may be that the US government wants to restrict the sales of high-performance GPU cards in China, especially for Chinese HPC double-precision high-performance cards.
The MI100 in the above news is a commercial product, and the MI200 will be commercialized by the end of 2022, all using the 7nm process technology.
There is also market news that NVIDIA China has received a request from the headquarters: suspend the shipment of data center GPU cards A100 and H100 to all customers and all agents in China, and other GPU cards will not be affected; the existing stock A100 of each server OEM GPU cards can continue to be delivered to their respective industry customers at present, and NVIDIA China has not issued any Letter to OEMs at present; NVIDIA headquarters is still analyzing the policy requirements of the US government, and it is expected that it will take 2-3 days for customers in China to communicate with customers in China. Communication caliber of partners.
In the news, NVIDIA’s A100 has been commercialized, using a 7nm process technology, and the H100 was just launched in March this year, using a 4nm process technology.
Rumors also include: Nvidia notifies Chinese Internet customers of high-end Al training GPU chips, H100 and A100 are restricted to Chinese Internet companies, and companies such as Tencent, Baidu, and Byte have all verified that they have received the notification. There may be clear sanctions news in the follow-up, which is currently a pre-communication.
The reporter asked Alibaba Cloud and Tencent Cloud for verification, but the other party has not responded as of press time.
Point to China Supercomputing
Zhu Jing said that according to the news, the blocked products are high-end GPUs with sufficient double-precision computing capabilities, and low-end GPUs are not affected. High-end GPUs with high double-precision computing capabilities are mainly used in high-performance computing fields, including scientific computing, CAE (computer-aided engineering), and medical care.
The supercomputing center is the national supercomputing center. It consists of thousands or even more processors with super-high computing power. It is known as the “Mount Everest in computers” and mainly meets the needs of the country’s high-tech fields and cutting-edge technology research.
In contrast, ordinary data centers are oriented to all scenarios that require information technology support, including a large number of Internet applications. China’s telecom operators and Internet companies have built their own data centers. Zhu Jing said that enterprise-level data centers often purchase A100 and H100 products in NVIDIA news. These products are high-end GPUs with sufficient double-precision computing capabilities. If the above-mentioned interruption of supply is implemented, the scope of the impact will be relatively large.
In the past, the United States has “hands-on” with China three times. In 2015, four Chinese institutions related to China’s “Tianhe-2” project were included in the “Entity List” by the United States; in 2019, Haiguang, Zhongke Suguang, Wuxi Jiangnan In 2021, 7 supercomputing institutions including Feiteng and Shenwei entered the entity list.
Zhu Jing said that from the above process, it can be seen that for China’s supercomputing, the United States has upgraded its attack methods and scope. If the news is true, this time the attack method will be upgraded from “supercomputing related units entering the entity list” to “direct ban on sales of related products that can provide services for supercomputing”, resulting in the expansion of the affected area from the supercomputing field to the Internet field.
Zhu Jing said that the cutoff of high-end GPU supply still seems to be a further blockade of China’s supercomputing and intelligent computing, but the scope of the impact has been greatly enlarged, and the cutoff of technical points must also consider the implication of upstream and downstream.
Source: Economic Observer
This article is reprinted from: https://www.williamlong.info/archives/6911.html
This site is for inclusion only, and the copyright belongs to the original author.