Am I the only one that thinks that the faster supercomputers in the world are never published due to the fact that they are probably used for military applications?
But the Chinese comp is undoubtedly for military application as well, since it at the national DEFENSE university.
I guess the governments could be hiding many more supercomps, but I don't really see the point. These are not that difficult to make, and single computers like this seem almost as much for bragging rights as for anything else. Also, I would think the amount of total military research would depend more on the combined computing speed of all your computers rather than one or two really fast ones. So if we had 20 "regular" supercomputers, I'm sure all the work that needs to be done will get finished efficiently enough.
What this shows however is the progress China is making, coming from out of nowhere basically. They interviewed some guy in those articles about the "milky way" computer in Tianjin, who said one of its key components is a Chinese proprietary interconnect which is twice as fast as the ones typically used in US supercomputers. That's a fairly innovative bit of technology as the power of the supercomputer not only depends on how many processors you use, it also depends on the speed of the interconnects which link up those processors as well as the software controlling the whole shebang.
So it's a three pronged beast, and while China has already mastered two of the three (inerconnect, software). The same guy interviewed (some supercomputer expert from tenessee where the american supercomp center is) also said that China is developing its own CHIPS for their next supercomputers, due out within the next 2 years, which he expects to be world class.
If they are coming up with chips like that, I think Intel and NVidia needs to be worried about more than AMD ...
