China Builds World's Fastest Supercomputer Without U.S. Chips (computerworld.com)
Reader dcblogs writes: China on Monday revealed its latest supercomputer, a monolithic system with 10.65 million compute cores built entirely with Chinese microprocessors. This follows a U.S. government decision last year to deny China access to Intel's fastest microprocessors. There is no U.S.-made system that comes close to the performance of China's new system, the Sunway TaihuLight. Its theoretical peak performance is 124.5 petaflops (Linpack is 93 petaflops), according to the latest biannual release today of the world's Top500 supercomputers. It has been long known that China was developing a 100-plus petaflop system, and it was believed that China would turn to U.S. chip technology to reach this performance level. But just over a year ago, in a surprising move, the U.S. banned Intel from supplying Xeon chips to four of China's top supercomputing research centers. The U.S. initiated this ban because China, it claimed, was using its Tianhe-2 system for nuclear explosive testing activities. The U.S. stopped live nuclear testing in 1992 and now relies on computer simulations. Critics in China suspected the U.S. was acting to slow that nation's supercomputing development efforts. There has been nothing secretive about China's intentions. Researchers and analysts have been warning all along that U.S. exascale (an exascale is 1,000 petaflops) development, supercomputing's next big milestone, was lagging.
Other than the singular purpose of doing the rare but interesting exascale problem the real utility of figuring out how to build exascale is to figure out how to build petascale cheaply on the "desktop". DOE's target for exascale is a scalable architecture that will sip mere megawatts of enegy this means that when they get there petawatt will be of the order of magnitude of 10KW. in other words, something easily power by a car engine. it will mean that petascale will be ubiquitous. Every hospital could have one in their basement.
the Titan super computer at oakridge has 299,008 cores and 18petaflops. if they built 5 of these they could hit 91 petaflops with 1.5 million cores. That's ten time fewer than the chinese super computer requires.
Now cores ain't everything. ultimately it's petaflops per watt but I don't have those statistics so I'm using cores as a proxy. I will admit that there is a school of thought in computer that having a lot of slow low power cores may be better than fewer fast high power cores whenever the bottleneck is memory bandwidth. And since I don't know the chinese architecture I don't know if that's what they are doing here.
Nonethe less. for mere factors of 10, brute forcing is always possible and doesn't really advance the state of the art--- that is, you cant scale that to factors of 1000 by brute force. You have to drop the power per petawatt to get anywhere. Scaling up to more cores can even be counter productive for any problem involving long range coupling between cores. Thus while it gets you more embarassingly paralell flops it doesn't get you better calculations on real problems.
Some drink at the fountain of knowledge. Others just gargle.
I used to teach a pretty decent load of Chinese students in my classes in Manhattan (I taught at both NYU and on CUNY). By the '00s, they were significantly more creative, sophisticated, well-rounded, and learned (I make no claims about "intelligence") than my American students, who were really sort of "decadent" in the worst, stereotypical ways—knew only a few things about a few things but a lot about consumer goods and fashion, and didn't seem to think they needed to work, just didn't feel the global pressure from competing workers. Very entitled.
The Chinese students tended to cluster in 'A' territory and always approached me after class to talk about class topics until I had to leave, then followed up with serious questions by email. The American students always had one or two in the 'A' group and the rest clustering around low B and high C, and it was a struggle just to learn their names, as they had nothing at all to say to me unless I called on them in class. Ironically, many of the Chinese students had better formal English as well, though there were always also about half that were clearly 'winging it' and needed ESL—but were killing it in class performance anyway, managing to learn and to get through books by relying on a dictionary, a study group, and sheer determination.
STOP . AMERICA . NOW
It's entirely possible this is an extravagant expense to hit #1 in Top500, but not able to do anything else. Looking at the architecture, about all that system can do is run xhpl really good. They talk about practical applications, but it seems the described system skimps on all sorts of things that would be important for real work, but not needed as much for xhpl benchmark runs.
XML is like violence. If it doesn't solve the problem, use more.
> was it stolen outright?
About 20 years ago, our IT manager at DEC noticed an extremely large download being made to the office in Japan over the weekend. He killed the connection and investigated. They were accessing the Alpha design docs and got away with pretty much everything. They were just starting to download the process docs and didn't get very far. So we at the time knew we'd see something Alphaish popping up in China about a decade later. In 2006 the ShenWei appeared on the scene and it bore striking resemblance to the Alpha 21164A, the processor design docs that happened to be downloaded that weekend long ago.
I'm not sure any nation has put as much work into the development of semiconductor technology as the US. Yes, lots of work was done elsewhere, and yes, lots of people from elsewhere come to the US and work in academia or for private companies working on that technology, but I think the history of the development of computers puts the US in a rather special place; first country to develop a first completely electronic computer, the country that invented the transistor, and the first country to develop an integrated microprocessor was the US. So I'd say the US was at the forefront of a lot of innovations in computing.
The world's burning. Moped Jesus spotted on I50. Details at 11.