Following Huawei, last week the Commerce Department announced the inclusion of four Chinese companies and a Chinese research institute
Among the five units mentioned above, Haiguang is more noticeable because AMD announced an X86 authorization agreement with Haiguang Investment Company in 2016, which authorized the most advanced Zen architecture at that time to Chinese companies. After that, two joint ventures of Chengdu Haiguang Integrated Circuit and Chengdu Haiguang Microelectronics Technology were established. Domestic X86 processors were also listed this year, covering 4 cores to 32 cores. Core, desktop, server and other market segments.
After being sanctioned by the United States, Haiguang's development prospects are unknown, and the AMD recently issued a statement saying
The U.S. sanctions will obviously affect AMD's cooperation with domestic companies, but the impact of these two rounds of sanctions on AMD will not be significant, but may be more serious for Intel. Mitch Steves, RBC analyst at Royal Bank of Canada, recently published a report that the United States has a much greater impact on Intel than on AMD as it suppresses Huawei, Dawn and other Chinese high-performance computing companies.
His reason is that AMD will launch EPYC 2 processors this year, the 64-core 128-thread Roman processors, mainly for the high-end market, which has little to do with dawn.
But Intel is different. They dominate the low-end server chip market, accounting for 95% of the server processor market, while Huawei and Dawning contributed 10% of Intel's server chip market. This is unbearable for Intel. Q3 revenue guidance will decline.
RBC analysts have yet to respond to Intel's comments, but from past examples, these two things may indeed have a greater impact on Intel. In 2015, the United States announced sanctions against China's high-performance computing units. At that time, sanctions were imposed on four major domestic over-computing centers, including Guangzhou over-computing center, which also prevented Tianhe 2 from using Intel chips to upgrade.
Tianhe 2 used Intel to strong E5-269212 core processor and Xeon Phi 31S1P accelerator card. It has a total of 3.12 million cores. Its theoretical performance is 54.9 PFLOPS (trillions). Linpack's peak performance is 33.86 PTFLOPS. Its performance has not changed from June 2013 to June 2016. It has won six consecutive TOP 500 championships.
The blockade of the United States made it impossible to use Intel products for the subsequent upgrade of Tianhe 2. At the end of 2017, Tianhe 2 began to be replaced by domestic ones. The acceleration card was upgraded to domestic Matrix 2000. Each acceleration card used four Matrix 2000 chips. Each Matrix 2000 consisted of 128 cores with a frequency of 1.2 GHz. It can perform 16 double-precision operations per cycle and the peak value of the processor. Performance is 2.45TFLOPS.
In addition to replacing the accelerator, the Tianhe 2A overrun after upgrading Tianhe 2 also improves the key components such as network, memory and storage. While the overall power consumption decreases slightly, the performance improves from 54.9 PFLOPS to 94.97 FLOPS.
For Intel, the application of their Xeon Phi acceleration card in Tianhe 2 over-counting should have been an excellent example of cooperation. Tianhe 2 is the six-sector TOP 500 champion, but because of the sanctions, the follow-up can no longer be sold. The hope of Xeon Phi expanding the market has also been dashed, especially in the big market of China, and finally Intel has abandoned the Xeon Phi product line.