Contents
China has long been a powerhouse in the supercomputing world, and its technology is only getting better. In this blog post, we take a look at how china built its supercomputer tech and what the future holds for the country’s high-performance computing ambitions.
Checkout this video:
The Birth of Chinese Supercomputing
Once only a laggard in the supercomputing race, China has made astonishing progress over the past decade. In 2010, the country produced just one of the world’s top 500 supercomputers. But by 2019, it had 426 systems on the list, more than any other nation. How did China catch up so quickly?
The Chinese Academy of Sciences
The Chinese Academy of Sciences (CAS) is the primary research and development organization for the Chinese government. Founded in 1949, it is committed to conducting scientific and technological research across a wide range of disciplines. In recent years, CAS has been at the forefront of Chinese supercomputing development.
The Tianhe-1A, China’s first petaflop supercomputer, was developed by a team from the National University of Defense Technology and installed at the CAS Computing Center in 2010. Since then, CAS has continued to invest heavily in supercomputing, with a particular focus on building exascale machines.
In 2015, CAS unveiled the Sunway TaihuLight, the world’s fastest supercomputer at that time. The system is powered by Chinese-designed SW26010 processors and achieves 93.01 petaflops on the LINPACK benchmark. Two other CAS-affiliated machines, the Tianhe-2A and Tianhe-2B, are also ranked in the top 10 of the most powerful supercomputers in the world.
CAS’s commitment to supercomputing is driven by a belief that these machines will be essential for driving economic growth and scientific advancement in China. Supercomputers can be used for a wide range of applications, including weather forecasting, climate modeling, data analytics, oil and gas exploration, drug discovery, and much more. By investing in this technology now, CAS hopes to position China as a global leader in many industries in the years to come.
The Dawning Era of Chinese Supercomputing
The history of Chinese supercomputing is relatively short, but in that time, the country has made incredible strides in developing its own indigenous capabilities. In just over a decade, China has gone from being a non-player in the supercomputing world to being a major force, with three of the world’s top five fastest machines.
How did they do it?
It all started with a little-known Chinese company called Dawning Information Industry Co. Ltd. In the early 2000s, Dawning began working on a new machine that would eventually become China’s first supercomputer, the Dawning 5000A.
Dawning had no experience in building supercomputers, so they partnered with IBM to learn the necessary technology. The 5000A was based on IBM’s POWER4 architecture and used standard off-the-shelf components. It was completed in 2004 and officially unveiled at the International Supercomputing Conference that year.
While the Dawning 5000A was not the fastest machine in the world, it was a major achievement for China and opened the door for further development. In just a few years, Chinese companies and institutions had built several more powerful machines, including the Tianhe-1A, which became the world’s fastest supercomputer in 2010.
Since then, China has continued to invest heavily in supercomputing, with the goal of becoming the global leader in this field. They are well on their way to achieving this goal, with three of the world’s five fastest machines currently located in China.
The Rise of Chinese Supercomputing
China has been making great strides in the world of supercomputing, and is now home to the world’s most powerful supercomputer. The Sunway TaihuLight, which was built entirely with Chinese-made components, is more than twice as fast as the previous record holder, Tianhe-2. In this article, we’ll take a look at how China built its supercomputer tech.
The First Generation of Chinese Supercomputers
The first generation of Chinese supercomputers was based on foreign technology. In the early 2000s, China’s National University of Defense Technology (NUDT) began working on the “Godson” line of processors, which were designed to be compatible with IBM PowerPC chips. The first Godson processor, the Loongson 1A, was released in 2002.
In 2004, China unveiled its first homegrown supercomputer, the Dawning 1000A. The 1000A was based on x86 processors from Intel and AMD, and it was used for weather forecasting and nuclear weapon simulation.
In 2006, China released theLoongson 2F, which was faster than the Loongson 1A. The same year, NUDT unveils Tianhe-1 (TH-1), a supercomputer that uses a hybrid of AMD Opteron CPUs and Nvidia GPUs. TH-1 is later upgraded to TH-1A.
The first generation of Chinese supercomputers relied heavily on foreign technology, but the second generation would be different.
The Second Generation of Chinese Supercomputers
Two years after the Tianhe-1A was retired from the TOP500 list of the world’s most powerful supercomputers, China unveiled its next generation of supercomputing technology with the launch of the Tianhe-2.
The Tianhe-2, which means “Milky Way 2” in Mandarin, was developed by a team of over 1,000 scientists and engineers from over 100 institutions across China. It is powered by 3.12 million processor cores and can perform quadrillions of calculations per second (petaflops).
At the time of its launch, the Tianhe-2 was twice as fast as the previous fastest supercomputer in the world, and it held that title for three years. In 2016, it was overtaken by Summit, a US-built machine.
The Tianhe-2 remains in operation today and is used for a variety of tasks including weather forecasting, earthquake simulation, and trouble-shooting for nuclear power plants.
The Third Generation of Chinese Supercomputers
The first Chinese supercomputer was built in the early 1990s and was based on imported technology. The second generation of Chinese supercomputers, which began around 2001, was based on home-grown tech. But it wasn’t until the third generation of Chinese supercomputers, which began to be developed around 2011, that China really began to catch up to the US in terms of supercomputing power.
The third generation of Chinese supercomputers is based on a new type of microprocessor called a Xeon Phi. These processors are designed by Intel specifically for supercomputing applications, and they offer much higher performance than the previous generation of processors (which were also designed by Intel).
In 2013, the Chinese government announced a major investment in supercomputing, with the goal of building a exascale computer by 2020. An exascale computer is one that can perform one million trillion calculations per second—a million times faster than a petaflop computer (which itself is a million times faster than a standard desktop PC).
As of early 2018, China has the world’s fastest supercomputer, the Sunway TaihuLight, which is capable of 93 petaflops. The US has the second-fastest supercomputer, Summit, which is capable of 200 petaflops. But Summit is still far behind Sunway TaihuLight in terms of raw performance—and it’s also significantly more expensive, costing around $325 million compared to Sunway TaihuLight’s $55 million price tag.
The Future of Chinese Supercomputing
Ten years ago, the idea that China would lead the world in supercomputing would have seemed absurd. The country was known for its cheap knockoffs of Western products, not for cutting-edge technology. But over the past decade, China has quietly built the world’s most powerful supercomputer network, assembling the largest fleet of these machines and using them to power everything from weather prediction to nuclear weapons research.
The Fourth Generation of Chinese Supercomputers
The fourth generation of Chinese supercomputers is expected to be based on the Loongson 3B processor, which is a quad-core 64-bit MIPS64 processor that was unveiled in 2014. The Loongson 3B is expected to have a clock speed of 1.8 GHz and a peak performance of around 8 GFLOPS per core. These computers are expected to be used for research into fields such as weather forecasting, earthquake simulation, and quantum computing.
The Fifth Generation of Chinese Supercomputers
The fifth generation of Chinese supercomputers is set to debut in 2020. These systems will be based on the new SUNWAY SW26010 CPU, which is reportedly 50 times faster than its predecessor. The new supercomputers will also use a new type of memory called HMC (High Bandwidth Memory), which is said to be 10 times faster than traditional DRAM.
In addition to the new hardware, the fifth generation of Chinese supercomputers will also feature a new operating system called SHENWEI OS. This OS is said to be much more efficient than Linux, the current standard for supercomputing.
The Fifth Generation Chinese Supercomputer Initiative is a massive undertaking that includes not only the development of new hardware and software, but also the building of a nationwide computing infrastructure. When completed, this infrastructure will allow Chinese scientists and engineers to share data and resources on a scale never before possible.
