Sina Technologies Zheng Jun from Silicon Valley, USA
“This is the most powerful X86 server chip at present!” AMD CEO Su Zifeng stood on the stage of the Landmark Architectural Palace in San Francisco and officially released the second generation EPYC (Xiaolong) data center processor “Rome” with full confidence, expecting to take more share from the server chip market dominated by Intel.
After the release of the new product, AMD shares rose sharply for two consecutive trading days, with a market capitalization of more than $43 billion. Compared with a few years ago when the stock price was less than $2 and the market value fell below $2 billion, AMD shares are now at their highest level in the past 15 years. And this magical reversal took place in just three years.
Enter Intel Backyard
In June 2017, AMD released its first generation of EPYC chips in Texas, re-launching its impact on the data center processor market and opening up a new revenue source for itself. Although AMD has not withdrawn from this market, due to the long-term non-competitiveness of its products, the previous market share has been very small, so it is not too bad to say that it will enter this field again. So far, AMD not only covers CPU and GPU chip areas, but also PC processor and data center processor markets.
Comparing with the PC market which has been in recession for many years, the data center processor market is still growing, and the advent of cloud computing brings more data center demand. RBC Capital, an investment bank, estimates that the 19 global cloud computing companies it tracks spent $93 billion on capital last year, up 42% from a year earlier. In the second quarter of last year alone, the four Internet giants, Google, Microsoft, Amazon and Facebook, invested $18.3 billion, up 48% from a year earlier. In 2018 alone, the data center processor business brought Intel $23 billion in revenue.
Compared with PC processor market, server chip is more like Intel’s own garden. According to IDC data, 98% of servers sold worldwide in 2016 use X86 architecture chips, while Intel dominates almost all data center X86 processors. It was not until AMD introduced EPYC processors that Intel had a real competitor in the market. (In addition, Google and Amazon have also launched self-developed, self-service, AI-focused cloud computing data center processors.)
Apparently, Intel, which is too comfortable in the data center market, is finally feeling the pressure of competition. In early 2018, Brian Krzanich, a former Intel CEO, told analysts before leaving that AMD would certainly take a share of the data center processor market from Intel, and Intel would do its best to prevent AMD from gaining a 15-20% share of the PC market.
However, this is the goal of AMD. After the Zen 2 architecture was released last year, Su Zifengyi said that EPYC server chip performance was better than Haolong, and she was confident to win the market share similar to Haolong (AMD once had a 25% share of the server chip market in the era of Haolong Opteron). Interestingly, AMD’s share of the PC processor market is also between 20% and 25%.
Strongest X86 Architecture Chip
Interestingly, AMD, nicknamed “Agricultural Enterprise”, likes to name PC processors with bulldozers, road roller, pile driver and other engineering vehicles, giving people a feeling of being in the construction site and flying dust. When it comes to data center processors, AMD begins to change its literary and artistic youth line, naming EPYC processors after famous Italian cities, and “all the way north”. The first two generations of EPYC processors are coded Naples and Rome in southern Italy, while the next two generations will be named Milan and Genoa in northern Italy.
The second generation EPYC processor “Rome” is AMD’s backlash in the data center market. The processor carries a maximum of 64 Zen 2 cores with a 7-nm manufacturing process and is the first 7-nm data center processor in the world. The second generation EPYC 7002 series processors can provide up to 64 Zen 2 cores per SoC. Compared with the previous generation, the performance of the second generation EPYC processor has improved by two cups, the IPC performance of each core server load has improved by 23%, and the L3 cache has increased by up to four times.
Su said that the second generation EPYC (Xiaolong) is the most powerful X86 architecture processor at present, with record performance, and can reduce the total cost of ownership by up to 50% under a variety of workloads. For enterprise data centers, the second generation EPYC processors bring 83% performance improvement of Java applications, 43% performance improvement of SAPS 2 Tier, and a new industry record in real-time analysis performance of Hadoop.
As before, new AMD processors are almost everywhere compared with Intel products. Su made it clear that the competitiveness of AMD EPYC processors is more flexible, affordable and better performance. EPYC 7742 processor performance is 97% higher than Intel Xeon 8280L and cost is 50% lower.
7-nm process technology is the biggest advantage of the second generation EPYC processor. In contrast, Intel, an old rival, is facing the dilemma of 10-nm technology jump. It will not launch 10-nm data center processor until next year at least. This means that AMD is at least one year ahead of the track and can seize the opportunity to grab market share.
Constructing Ecology is the Key
The success or failure of the server market is determined not only by the powerful processors, but also by the complete ecological platform. While returning to the data center market, AMD has joined dozens of partners in the industry. Microsoft, Baidu, Dropbox and Bloomberg, the seven giants in the field of data center services, have also clearly expressed their support for AMD EPYC chips. They prefer to see healthy competition in the server chip industry, rather than Intel’s monopoly.
Over the past two years, AMD EPYC ecosystem has more than 60 partners, has launched more than 50 mass-produced server platforms, and has more than 15 partners in various sub-areas. These include original design vendors such as Gigabyte, independent hardware vendors such as Broadcom Meguiar, Microsoft and many Linux operating system vendors, as well as server vendors such as HP, Dell and Lenovo. Amazon AWS and AMD’s Amazon EC2 cloud service based on EPYC processor chips was also launched three months ago.
At this San Francisco conference, AMD invited many partners in the industry to serve as its platform. Google announced that it has deployed the second generation EPYC processor in the internal architecture production data center environment, and will support a new general purpose computer with the second generation EPYC processor on the Google cloud computing engine at the time of writing this year. Twitter announced the deployment of the second generation EPYC processor in the data center architecture this year. Hewlett-Packard and Lenovo announced that new server platforms based on new processors would go on sale today.
Thanks to the growth of CPU and GPU business, AMD revenue grew by 23% last year compared with the same period last year. In the first quarter of this year, the market share of AMD EPYC processors in global server chips was about 2%. Although there is still a long way to go to Su Zifeng’s long-term target of 20% to 25%, this is a new revenue source for AMD. In the server chip market, even if AMD gets 10% of the market share, it means more than $2 billion in new revenue, equivalent to nearly a third of AMD’s annual revenue growth.
More importantly, demand for game mainframe chips, which once helped AMD out of its predicament, has slowed down, and AMD expects revenue to fall 20% in the third quarter of this year from a year earlier. The expansion of server chip business is an important way for AMD to offset the decline of game host chip revenue. However, Su said that AMD will launch Navi GPU with 7-nm technology by the end of this year, and that if nothing happens, Sony’s next-generation Playstation host will adopt the GPU.
Fifty Years AMD of Agricultural Enterprises
It is worth mentioning that this year marks the fiftieth anniversary of AMD. Like Intel, AMD was founded by several engineers at the legendary company Fairchild. On May 1, 1969, Jerry Sanders officially launched AMD in Sunnyvale, Silicon Valley, with eight engineers and a $1 million loan from Bank of America. Above is the founding ceremony of AMD’s original founding team.
In the past half century, AMD has been the only processor manufacturer to compete with Intel and the only competitor to have X86 architecture authorization. In the shadow of Intel’s giant, AMD has always maintained its tenacious vitality. Although AMD has fallen into a depression many times, even seemingly gloomy prospects, it can always rely on technological innovation, with the help of star products to get out of the predicament again. Leading AMD out of its latest downturn is Su Zifeng, the first Chinese-American CEO in the semiconductor industry.
The immigrant from Taiwan, China, took over the position of AMD CEO in 2014, and began a magical retrograde journey. When she took over, AMD was at its lowest level in decades: its market share had fallen to 10%, its performance had been losing money for years, and its market value had shrunk to only billions of dollars. But under the leadership of Su Zifeng, AMD has re-demonstrated its competitiveness in both CPU and GPU areas through business adjustment, not only turning losses into profits, but also creating a miracle of fifteen-fold stock price growth.
Thanks to the Rzen CPUs launched in 2016 and EPYC processors launched in 2017, AMD’s share price has soared from its lowest level of less than $1.8 in 2016 to below the current level of $30, with a market capitalization of $31.25 billion. AMD shares have recovered 50% this year. AMD shares are at their best in the past decade.
Just before the EPYC announcement, American media had spread the news that Su Zifeng might leave AMD, saying that she might return to IBM as No. 2 executive and even succeed IBM CEO Ginni Rometty in the future. However, Su immediately relayed the report on Twitter, positively denying the news that she was leaving. “This rumor is groundless. I love AMD, the future will be better! “