AMD's World's First 16-Core Processor Focuses on High-Level Interview with Cloud Computing

On the afternoon of November 14, 2011, AMD held a new generation of Opteron processor conferences with the theme of "core" for cloud technology in Beijing, and launched the world's first bulldozer architecture with the highest 16-core design. Opteron 6200 and 4200 Series processors, with a focus on cloud computing. AMD executives also received an exclusive interview with the media after the meeting.

Deng Yuanyi, senior vice president of AMD and president of Greater China, said at the press conference that the AMD CEO has recently announced several strategic directions for AMD in the future. First, the products should have lower power consumption, and second, they should devote themselves to the development of cloud computing. Multiple investment in emerging markets, and this conference is very much in line with AMD's three strategic directions: 16 nuclear dragon not only created the core of the highest, each core power consumption is also lower than ever; this conference theme The so-called "core" of cloud technology shows that the new generation of Opteron processors is introduced for cloud computing; the launch of the conference in China demonstrates the importance attached to the Chinese market.

Deng Yuanzhang specifically expounded the issue of cloud computing, saying that the advantage of AMD is that it provides a full range of services. AMD not only has servers and good technology at the terminals, but also AMD is the only company with leading CPU, GPU, and APU technologies in the world. The vendor enables users to have a very good experience when dealing with multimedia and provide the best reliability for virtualization.

In addition, cloud computing is not only about hardware, but also the entire industry chain, including various applications of servers, storage, networks, and operators. Only such an industrial chain can really promote the success of cloud computing and truly affect people's lives.

Nigel Dessau, senior vice president and chief marketing officer at AMD, said that 2011 is an interesting year for AMD. It released the next generation of technology APUs at the beginning of the year and made a second important announcement at the end of the year, not only changing The way the client works also changes the way the server works. With more than 2 million AMD Opteron processors supporting global cloud computing, AMD has made three promises: First, commitment to processor technology, second, commitment to cloud technology, and thirdly, The promise of the Chinese market.

Mike Wolfe, AMD’s senior vice president and chief information officer, talks about how AMD processors make cloud computing more efficient, less costly, and gives specific data: AMD’s cloud system has 124,000 cores and more than 500 tools. Supports more than 3,000 engineers and executes more than 40 million tasks each month. He also gave an example: In 2009, the Bobcat architecture test required more computing power, so the AMD cloud system provided them with 5% core to 45% private cloud occupancy within 24 hours. The two-month development and verification work was shortened to 5 days.

Paul Struhsaker, AMD’s global vice president and general manager of the commercial division, first introduced the bulldozer architecture. This is the first new architecture update of Xiaolong in eight years. The 16 cores can be better virtualized, and the architecture is defined in the form of modules, both dedicated resources and independent components. He also highlighted the value-to-value ratio that Nine Dragons attaches most importance to, and claims that the new generation of products can increase speed by 71%, reduce prices by up to 34%, reduce power consumption by 20%, and save up to 50% in power saving mode.

Lorry Zhang, Director of Wyse China, Henry Wang, Director of Microsoft Technology, and Hao Haibin, Senior Vice President of Shanda, then discussed with Nigel Dessau about their collaboration with AMD on cloud computing. For example: Wyse used to focus on high-performance computing. Now it is. Equipped with AMD G and Z series APUs for thin clients and clients, providing high-definition and green experience with 8% of the PC's power consumption; Microsoft has adopted AMD's processors as a hardware platform in the public cloud, providing it to the global, private cloud. There are also computer based on Xiaolong, and the application program has exceeded 200 cases. Shanda opened AMD Opteron-based cloud computing platform in July of this year and will soon conduct open tests.

Other on-site presentations on AMD's cloud computing application cases include: Chao Yun CEO Wei Yiqun, Dell Sun Greater China Marketing Director Sun Xiufang, Dell Customer Representative Dr. Jin Yijing, National Laboratory of Tsinghua University, Zhang Yueping, Vice President of Shuguang, and Alex Hsu, Chairman of AMD China. Jeff Carlat, director of global industrial server and software business at HP.

Senior executives who participated in the interview after the meeting included Nigel Dessau, senior vice president and chief marketing officer of AMD, Mike Wolfe, senior vice president and chief information officer of AMD, Paul Struhsaker, general manager of AMD’s global vice president and commercial division, and John AMD’s server product marketing director. Fruehe et al.

The following is an overview of the interview.

Q: After the bulldozer was released, what measures did AMD have in China?

Nigel Dessau: AMD attaches great importance to the launch of the new generation of Xiaolong in the Chinese market, which also demonstrates AMD's commitment to the server market, to the cloud computing market and to the Chinese market. When it comes to the server market, AMD considers three major applications, databases, virtualization, and cloud computing. There is no doubt that cloud computing is a hot topic today, but remember that most customers today purchase cloud computing servers primarily to solve problems in their data centers. AMD and Chinese partners attach great importance to the above three important topics, whether cloud computing or non-cloud computing.

Paul Struhsaker: Snapdragon 6200, 4200 two product combinations will appear.

John Fruehe: A lot of people think that AMD talks about cloud computing and what it needs is low power, but actually there are many different requirements in cloud computing. For example, the largest social networking site in the United States, Facebook, uses AMD Opteron to support cloud computing, while Microsoft uses AMD Opteron to achieve low power consumption, but the computing density is very high, so different customers will have different ways.

Q: With what vendors did AMD bulldozers cooperate with HPC? Is there any plan to cooperate with Chinese scientific research institutions or universities? Will future supercomputer TOP500 have bulldozer-based products?

John Fruehe: There are already several high-performance computing systems using the new Opteron processors, such as the HECToR from the University of Edinburgh in the United Kingdom, the Oak Ridge National Laboratory in the United States, the HLRS from the University of Stuttgart in Germany, and the Swiss National Supercomputer Center (CSCS). AMD's global HPC partners include Hewlett-Packard, Dell, Cray, and Appro in the United States. Dawning in China will be an important partner for AMD.

Q: The number of Interlagos cores has reached 16. For enterprise applications, whether it is cloud-based or non-cloud-based, has AMD and partners made some software optimizations for so many cores?

Paul Struhsaker: Yes, there are indeed such optimization measures. Prior to the launch of the next-generation Opteron products, not only did the software partners perform a series of optimizations on the new architecture, AMD also optimized the compilers and super-large computers accordingly, and continued fine-tuning to adapt to the new architecture. There are also new changes to save energy, especially in encryption and floating-point operations in the instruction set.

John Fruehe: In the process of optimization, AMD cooperated with Microsoft and the entire operating system ecosystem. Microsoft has a lot of servers based on the development of Xiaolong, AMD's ranking in open source contributions is also in the top fifteen.

Mike Wolfe: In AMD's internal IT department, AMD also conducted corresponding tests. AMD will continue to further optimize its software development to take advantage of the high performance of next-generation processors.

Q: AMD's innovative floating point design supports 128 and 256 units of AVX instruction set. In terms of software, the 256-bit platform has not yet been universally applied. AMD is seeing the future development trend. Do you think that there will be such a direction in the future?

Paul Struhsaker: This is a combination of customer requirements and AMD predictors. AMD does see that customers use 256-bit instruction sets. Of course, there will be more customers using 128-bit instruction sets. Many HPC users still have requirements for 256-bit, so when designing, AMD is mainly based on customer requirements. , coupled with AMD's judgment.

John Fruehe: AMD's product is called FLEX FP (Flexible Floating Point Unit). This is not just a technical term. It describes its functions more. It is the solution that can get maximum flexibility in floating-point operations. The most important thing is that in the future it can not only support 256 bits, but in today's 128-bit operation, it can provide you with twice the resources.

Nigel Dessau: When AMD launched Opteron eight years ago, it redefined the server—a balance between performance, memory and I/O, and CPU. Today, AMD continues to redefine the concept of servers in the entire world. It is the combination of floating-point operations and integer units to achieve more integrated and more flexible performance. I believe that today's next-generation Opteron platform is the first such type of server, and all servers in the future will be like this.

Q: AMD's HPC has received support from Dell and other manufacturers this year, but the economic environment will not look good next year. What is AMD's view on market demand and computing market next year?

Nigel Dessau: Now many companies think that the future economic situation is not a threat but an opportunity. Sometimes it is also an opportunity for customers to use this opportunity to re-analyze the performance and performance of data center systems, including scalability, economics, etc., to reshape data centers beyond their competitors. Q: The current server market not only has AMD, Intel, and IBM, and the competition is becoming more and more fierce. Some time ago ARM announced that it would launch server chips. How does AMD view the current market? What kind of methods will be used to consolidate and increase their right to speak in the future?

Paul Struhsaker: Only AMD and Intel offer x86 architecture-based server processors in the market. The IBM-supplied Power processor has always existed and will continue to exist because it is another architecture. We believe that ARM will provide very low-end server processors in the market. All future software is written in 64-bit mode, so the server should support 64-bit computing. We have also discussed with many software developers. No one is willing to return to the 32-bit environment. For the processor provided by ARM, unless we can provide 64-bit computing, we are not interested. However, we will continue to focus on and evaluate, because any technology company must carry out continuous assessment of different technical options.

Nigel Dessau: I have been working in sales and marketing for 25 years. I have always had this experience: To be successful in sales and marketing, your products must be compatible with the past, and can transition to the future. In the past 25 years, I haven't seen any company that has achieved success. When it's time for the future, it's backward compatible to support the past.

John Fruehe: AMD's next-generation processors can achieve power consumption as low as 4.3W per core, while being compatible with all past software, and are based on x86 technology. AMD did not ignore the trillions of capital that had been spent on software development in the past.

Question: AMD has provided the two cores as a module. What prompted AMD to make such an initiative? In addition to lower manufacturing costs and lower power consumption for AMD, what advantages can this bring to the customer's business?

Paul Struhsaker: Driving AMD's innovation is, of course, the customer's voice and needs. Today's AMD's new modular design is based on the voice of customers. Many applications have put forward requirements. AMD's innovation is based on the needs of customers.

John Fruehe: What benefits does the modular architecture bring to customers? The most important thing is that it can make AMD easier to innovate in the future, because customers need to constantly change, AMD needs to quickly adapt to this change, introduce new products to meet their changing needs. Modular architecture can meet this requirement, and after AMD's rapid innovation, customers can also use their business opportunities better and faster. It is as if the modularity of the IT architecture allows AMD to allocate resources to businesses in an optimal manner. Modularity in the processor area also allows AMD to allocate the best resources to where customers need it most.