Working Data Center

China, supercomputers’ new homeland

Who is winning the battle of the supercomputers, the huge machines used by the most important institutions in the world?

by Eni Staff
06 February 2020
6 min read
by Eni Staff
06 February 2020
6 min read

At first glance, the answer is granted. The United States stand out in this strategic sector. The top two spots on the official Top 500 ranking are held by American computers made by IBM, namely Summit, with its computational power of 148 petaflops (148 million billion calculations a second), and Sierra (94 petaflops), which also contains Nvidia and Mellanox components.

Historically, it was Japan to first compete with the United States in the field of supercomputers, followed by European countries like Germany, Britain, France and Italy. But in the last few years, the situation has changed dramatically. In 2001, China, was not home to a single one of these supercomputers and now boasts 45% of the 500 most powerful supercomputers in the world, while the Americans have seen their share fall year after year to an all-time low of 23%.

So, the People’s Republic does not just lay claim to third and fourth place in the rankings (with its Sunway TaihuLight and Tianhe-2A) it owns 227 of the top 500 supercomputers, against the United States’ 118, Japan’s 29, France’s 18 and Germany’s 16. The Asian giant is making its mark in the most strategically important field in technology (twinned with Artificial Intelligence). 

The Exaflop of Shuguang

“China has acknowledged the importance of supercomputers,” Jack Dongarra explains to Eni. Dongarra is a professor of computer science at the University of Tennessee and a member of the Oak Ridge National Laboratory. He’s regarded as a world-leading expert on supercomputers. “They have invested a great deal and have long-term plans, but then that’s their classic modus operandi. They had no supercomputer systems in 2001 but today they’ve overtaken the United States. The speed at which they’ve grown is unparalleled in the world.”

That’s not all. The Chinese also claim to have come up with a prototype for a new supercomputer, Shuguang, currently housed in the Chinese Academy of Sciences and able to go at 200 petaflops with no difficulties. This is a far greater speed than that of the systems currently ruling the rankings. At its peak, Shuguang can reach speeds of 1 exaflop, or 1 billion billion calculations a second. This figure, inconceivable to the human mind, represents the biggest achievement in computing in the world.

“These supercomputers are really a new species of computer. They’ll not only be faster, but able to manage big data totally differently to how we do today,” Dongarra goes on. “They will clear new paths for Artificial Intelligence, data science and computer simulations, providing new knowledge. These advanced, so-called ‘exascale’ computers could perform 20 or 40 times better than today’s, and show that in this sector, the sky’s the limit.”

According to reports from the Asian media, China has decided not to officially present its prototype to the world solely because of the current trade war with the United States. Essentially, Beijing does not want to provoke its rivals by revealing its latest advances. Behind this gesture of good will – in name at least – lies a hope of scaling back the sanctions currently blocking American manufacturers from working with China to build supercomputers (although this is increasingly irrelevant, as the Asian giant can now do so in total independence).

A supercomputer to fill every need

All this paints a picture of just how important, strategically and geopolitically, the supercomputer sector is. But what, practically speaking, are they actually good for? The systems are used in almost every sector that requires analysis of huge quantities of data or simulations based on mathematical calculations. When it comes to meteorology, for example, analysing currents in the ocean and in the atmosphere can give useful indicators for predicting cyclones and tsunamis, and forecast the weather with a precision of 80% within an area of 1–3 km (as opposed to 100 km in the past).

From the sky up to the space: supercomputers allow the comprehension of dynamics and behaviour of objects in space, even simulating events like the formation of galaxies and the evolution of stars. They are also employed to more pragmatic ends: the buying and selling of stock market shares is done almost wholly by supercomputers (sometimes with unpredictable consequences, as the flash crash of 2010 showed), and the systems can also simulate atomic bomb explosions (reducing the need for real-life tests). More relevant to the wider public is the use of supercomputers for sequencing the human genome, developing new medicines and better understanding about how certain diseases work.

“In medicine, to develop a treatment, you have to find patterns that have never been identified before,” explains Dongarra. “The challenge is not the lack of useful data, because we have plenty of that now. The challenge is processing it to find the patterns that will tell us what causes cancer and fighting it more effectively. With supercomputers, we can find answers to questions that have stumped mankind until now. In these fields, supercomputers are excellent tools for analysing molecular datasets, the medical registers of patients, their family history and a lot of other complex information. The simulations made possible by supercomputers can speed up innovation.”

This acceleration is about to pass another crucial milestone, when the first quantum computers are developed. They will exploit quantum physics to massively speed up these tools’ calculations. So, are “traditional” supercomputers bound to become obsolete? “The progress of quantum computers in tests is fast and impressive, but we’re still very far off actually using systems like them,” concludes Dongarra. “Researchers are working to understand the problems that could be solved by quantum computers and develop algorithms to prove their usefulness. It’s generally thought that these systems will be of massive help in fields like financial trading and the military industry. But it’s definitely too early to start thinking about a quantum computer for general use.”