China Owns the Fastest Supercomputer …Now What? (video)

11 6 Loading
chinese-supercomputer

Tianhe-1A is the world's fastest supercomputer, and it's Chinese.

Early in October, China revealed that they have developed the world’s fastest supercomputer at their National Supercomputing Center in Tianjin. Known as the Tianhe-1A, the machine is capable of a remarkable 2.5 petaflops. That’s 2.5 thousand trillion floating point operations per second. Impressive to say the least and more than 40% faster than Jaguar, the current leading supercomputer in the US (according to Top500.org). Tianhe-1A follows a new kind of architecture, relying as much on graphics cards (from Nvidia) as it does on traditional CPUs (from Intel). While the major processing components of the Chinese supercomputer are all from the United States, the software and hardware that allows those chips to talk to each other efficiently was developed in China. What does China’s occupation of the top spot in computing mean for the world? Depends on Congress. The last time that a nation surpassed the US in supercomputing (Japan with Earth Simulator) funding was found to propel the US back to the top. With the economy still in trouble it’s unclear whether China’s supercomputer success will rally US research, or serve as the beginning for a long era of Chinese dominance.

Here’s a brief summary of the situation from the Wall Street Journal:

At least one major official is hoping to use the creation of Tianhe-1A to spur efforts in the US. Steven Koonin is the Undersecretary for Science at the Department of Energy. Among other things he’s their supercomputer guy. He recently helped organize a Simulations Summit that brought together some of the leading minds in the field to discuss future progress. Koonin also wrote a blog post on the DOE site that discussed the importance of the Chinese push in supercomputing. Tianhe-1A is just the beginning. The Undersecretary believes that China will have a supercomputer capable of a petaflop or more in the next 12 to 18 months. Only this new computer won’t rely on US chips – it will be completely home grown. Koonin’s message seems to be clear: China isn’t going to be beholden to the US for computing power, not any more. If the US wants to stay ahead it’s going to have to redouble its efforts, and soon.

I’m all for inspiring new research. In fact, I’d go so far to say that playing up the intimidation from a competitor (China) is fair game if it gets your team to work harder towards a common (peaceful) goal. Still, we need to have a reality check. China may have created the world’s single fastest supercomputer, but the US supremacy in computing is far from challenged yet. Top500.org maintains a list of the world’s fastest supercomputers. Jack Dongarra, a professor at the University of Tennessee and researcher at Oak Ridge National Labs, is one of the forces behind Top500 and the one who visited Tianhe-1A for its testing. Dongarra and Top500 report that roughly half of the world’s fastest 500 supercomputers are located in the US. An overwhelming 90% of all top 500 computers use US hardware for main processing. The US may not be at number 1, but if they occupy all the other spots does it really matter?

I also want to point out that Tianhe-1A isn’t some superior tactical planning device, it isn’t slotted to be used for military purposes. According to the Wall Street Journal, the supercomputer will be used as an “open access system”. In other words, China will be selling off time on Tianhe-1A to people all over the world to help with complex calculations (protein folding, analyzing geological surveys for oil, etc). There’s a very good chance that the creation of Tianhe-1A will lead to amazing discoveries in science that will help everyone all over the world. This supercomputer then doesn’t appear to be a threat to anyone.

So maybe it’s not about threats. Maybe it’s not about China defeating the US in supercomputing. It could just be about foresight. Tianhe-1A is a remarkable achievement in computing, but it’s also a pretty predictable one. China held the number two spot last year and they have spent years working towards becoming a major player in computing, genetics, stem cells, robots, energy…everything. You can almost hear the surge of pride and confidence that is fueling their growth in science and technology. They will not be relegated to the back seat anymore. In the future, we have to expect this nation to take a larger, often leading role, in research and development. You can’t think about the global change in technology without asking, how’s China going to play a role in this? The same is going to be true for many other nations that are transitioning towards a science fueled economy. India is another great example, as are Brazil, Turkey, and many other spots in Latin America and the Middle East.

Tianhe-1A can be a wake up call for the US to redouble its efforts in supercomputing. I’d like that. But Tianhe-1A should also be a big reminder that the future of science and technology is not going to be dominated by the major players of the 20th century, at least not totally. China is going to develop its own computer chip industry to rival those in the US. It already has developed innerconnects (software and hardware to get processing chips to talk to each other) that rival or surpass those here. Expect similar competition for every field and from a wider array of competitors. Part of the amazing effect of accelerating technology is that it allows groups with shorter histories in empirical science to quickly ramp up and partake in global development. Tianhe-1A knocked the US out of the top spot, but the real change will be when the field of supercomputing gets more crowded in general. That day is coming…though perhaps not for many years.

[image credit: NVidia]

[source: Wall Street Journal, Department of Energy, NVidia, NY Times, Top500]

Discussion — 6 Responses

  • Keith Curtis November 7, 2010 on 5:21 pm

    Supercomputers are happy to add 0 + 0 all day long. It is only via good software that we can put them to effective use. Here is an article I wrote about the relationship between free software, the free market, and tech progress http://bit.ly/d6Ctl7

  • IntervalGoat November 8, 2010 on 12:15 am

    Not sure what the deal is with this. Truly demanding calculations (e.g., protein folding) are much better achieved with distributed computing schemes (e.g., Fold@Home, SETI@Home) which are very well adapted to nature of US society (and perhaps less well to Chinese society).

    Also, has China achieved sub-micron CPUs? You can build a supercomputer with almost anything. This is a country that doesn’t build jet engines (that work, anyway), and building a competitive CPU is similarly demanding…

  • ME November 8, 2010 on 3:36 am

    A very well written article, however I should point out one area that could have been better researched: whether the US will try to build a faster supercomputer. In 2009 IBM announced it was constructing a 20 petaflop supercomputer that would be completed in 2012. Articles to that effect can be found here http://www.wired.com/wiredscience/2009/02/supercomputer/, and at the IBM website here http://www-03.ibm.com/press/us/en/pressrelease/26599.wss.

    Granted, this hasn’t gotten much press lately, but as far as I know, no delays to completion have been announced either. I’d say this pretty much answers the question of whether the US will take back the number 1 ranking. Unless China unveils an even faster supercomputer in 2012.

  • HIVE45 November 8, 2010 on 9:40 am

    China is already the hub for AI research too. This should get interesting…

  • Ben Selinger November 10, 2010 on 5:39 pm

    Who happens to know how many MIPS this machine can churn through?

  • Anonymous November 20, 2010 on 2:38 am

    I was just watching CSPAN-Richard Clarke speaking about cyber security threats and ways to protect the nation’s computer infrastructures. He was talking about the Russians exploiting devices as they are made using embedded controls as backdoors. He mentioned China in other contexts as a generalized potential enemy. If anyone has the capability to embed backdoors in computers it is China. I always wondered about the kind of thing that could be put in a NIC. Sure, in the known world of networking there are the seven layers of network protocols that keep our packets flying around cyber-space in an orderly fashion, but what about all of the extraneous chaos of all the other noise on the inter and infranets?