How China is Exascaling Out Old Tech

China is making a big push to exascale computing, and they’re doing it with a mix of old and new tech. Here’s how they’re making it happen.

Checkout this video:

Introduction

China isn’t satisfied with leading the world in supercomputing—it wants to dominate. The country has been on a constant upward trajectory in the Top 500 list of the world’s most powerful supercomputers for years, and it now boasts 227 systems on the latest rankings, more than double the number from just five years ago. China also operates the world’s three most powerful supercomputers, according to the list.

But China’s ambitions don’t stop there. The country has set a goal of building an exascale supercomputer—a machine that can perform a billion billion calculations per second—by 2020. That’s just four years from now. And China is on track to achieve its goal, thanks to a massive investment in research and development and a concerted effort to bring together the country’s vast resources.

China’s push for exascale supercomputing is part of a larger national strategic plan known as Made in China 2025, which aims to make the country self-sufficient in 10 key technological areas, including artificial intelligence, robotics, and semiconductors. In pursuing this goal, China is using both carrots and sticks: investing heavily in key technologies while also restricting access to them.

The country’s quest for exascale computing power has troubling implications for the rest of the world. As China builds ever-more powerful supercomputers, it will gain an increasingly large advantage in data-intensive industries such as artificial intelligence and weather forecasting. And as the country hoovers up more data about its citizens and businesses, it will only become more difficult for foreigners to compete.

The Need for Speed

The global race to build the first exascale computer is heating up. The united states Europe, and China have all made significant investments in research and development with the aim of building a machine that can perform one billion billion calculations per second—or one exaflop. But while the U.S. and Europe are focused on building machines that are energy efficient, China is taking a different approach: building the fastest machine possible, regardless of the cost.

The U.S. Department of Energy has invested $258 million in exascale research and development since 2012, with the majority of the funds going to developers of superconducting chips and various software projects. In Europe, the European Commission has invested €1 billion ($1.14 billion) in five research programs aimed at developing exascale computing, including a project to develop a new generation of memory devices and another to design energy-efficient interconnects between different computer components.

In China, meanwhile, the government has invested billions of dollars in building supercomputers as part of its “Strategic Priority Program on National Science & Technology Infrastructure.” The program includes a $73 million investment in an Exascale Computing Research Center, which is being led by Jack Dongarra, one of the world’s leading experts on high-performance computing. The Chinese government has also invested heavily in hardware development, with one Chinese company, Sichuan Yangtze Memory Technologies Co., Ltd., recently announcing that it had developed the world’s first 16GB NAND flash memory chip for use in next-generation supercomputers.

While the U.S. and European approaches are focused on building more energy-efficient machines, China is instead prioritizing raw speed—even if it means sacrificing energy efficiency. This difference in approach highlights the different priorities of each region when it comes to exascale computing.

The Slower Pace of Moore’s Law

Here’s something you don’t hear every day: China is outpacing the United States in technology. Exascale computing, in particular, is an area where China has taken the lead.

Moore’s Law, which states that the number of transistors on a chip will double approximately every two years, has been the guiding principle of the semiconductor industry for decades. But as chipmakers have reached the limit of what they can pack onto a chip, the pace of Moore’s Law has slowed.

While the United States is still working on its first exascale supercomputer, China has already built five of them. Exascale supercomputers are a thousand times more powerful than today’s most powerful computers. They can perform a billion-billion calculations per second and are used for complex tasks such as weather forecasting and nuclear weapon design.

China’s exascale supercomputers are based on a different architecture than those in the United States. They use many small chips instead of a few large chips and are connected by an optical network instead of a copper one. This design makes them more energy efficient than their American counterparts.

The United States is not giving up on exascale computing though. The Department of Energy has invested billions of dollars in its development and is confident that it will achieve its goal by 2021.

The Chinese Approach

The Chinese government is taking a different approach to exascale computing than their American counterparts. Exascale systems in China are being developed with a focus on using homegrown technologies, rather than relying on imported components and software. This “go it alone” strategy is in line with China’s overall goal of becoming self-sufficient in the high-tech industries.

There are several reasons why China is pursuing this path. One is that, due to trade tensions, the country has been increasingly cut off from American technology. The Trump administration has put restrictions on exports of American tech products to China, and Chinese companies have been blacklisted from doing business with American firms. As a result, China has been forced to look elsewhere for the high-tech products it needs.

Another reason for China’s focus on homegrown technologies is that the country wants to become a world leader in the field of artificial intelligence (AI). In order to achieve this goal, China needs access to large amounts of data, which it can then use to train its AI algorithms. Exascale systems will provide Chinese researchers with the massive computing power they need to process all of this data efficiently.

So far, China has made significant progress towards its exascale goals. The country’s first exascale system, called the Sunway TaihuLight, was operational in 2016. Since then, several more exascale systems have come online, including the Tianhe-2A and Tianhe-3 supercomputers. China is currently working on an even more powerful system called the Tianhe-4, which is expected to be operational by 2020.

While China’s focus on homegrown technologies may cause some short-term headaches for foreign tech firms, it is ultimately good for the global tech industry Competition from Chinese firms will spur innovation and drive down prices for consumers around the world.

The U.S. Response

As the PRC government continues to invest in exascale computing, the United States has responded with a mix of public and private initiatives aimed at maintaining its leading position in HPC. In 2015, the U.S. Department of Energy (DOE) unveiled its Exascale Computing Project (ECP) with the goal of delivering a capable exascale computing system by 2023. In addition to DOE, other U.S. agencies such as the National Science Foundation and Department of Defense are also pursuing exascale computing development.

The private sector has also stepped up investment in exascale development. In 2016, IBM announced that it was working with Lawrence Livermore National Laboratory (LLNL) on an exascale-capable supercomputer called “Summit” that is expected to be operational by early 2018. Cray Inc., another major HPC player, is also working on an exascale system called “Shasta” that is expected to be operational by 2021. Graphics chip maker Nvidia has also entered the exascale race with its own “Volta” GPU accelerator chips that it says will be able to handle the extreme computational demands of exascale systems.

These investments come as China ramps up its own efforts to develop an indigenous exascale computing capability. In March 2015, China’s National Development and Reform Commission unveiled its “13th Five-Year Plan” for economic and social development, which included a number of targets related to HPC development. These included a goal of building an indigenous petaflop-level supercomputer by 2020 and an exascale system by 2030.

In order to achieve these ambitious goals, China has made a number of strategic decisions with respect to HPC development, including increasing investments in both hardware and software research and development, establishing new HPC centers and test-beds, and building an indigenous ICT ecosystem

Conclusion

In conclusion, China is exascaling out old tech at an alarming rate. While this may be good for the country in the short term, it could have negative long-term consequences. Other countries should keep a close eye on China’s actions and be prepared to respond accordingly.

Scroll to Top