China’s University of Defense Technology has released the fastest supercomputer in the world named Tianhe-2, in which one would be able to find a whopping 32,000 Xeon processors located 48,00 Xeon Phi accelerator processors. In theory, the Tianhe-2 supercomputer is capable of managing a quadrillion mathematical calculations per second (that would be 33.85 petaflops when one needs to say it out quickly), which happens to be double the amount of its main rival, the Titan, whom it obviously knocked off its perch. At this point in time, the US can still salvage some pride as it holds slightly more than 50% of the Top 500 supercomputers in the world.
There are vast differences between Tianhe-2 and its closest rival, the Titan (17.6 petaflops) which we covered earlier. First of all, the Tianhe-2 is nearly twice as fast. However, in terms of efficiency, Top500.org still calls the Titan the most power-efficient supercomputer and that’s mainly due to the fact that Titan uses a mix of general purpose processors from Intel, along with graphics processors that are massively parallel and which offer a much higher compute density. Is is believe that China has for goal to build an “all-Chinese” supercomputer one day that only uses components designed and built in China.
The sheer numbers are impressive, but this does not mean that it is simply technologically more “advanced” when compared to the others in the Top500 list. Of course, being the last one to be released, it does benefit from newer technologies, but the main reason for its #1 place is simply due to how much hardware (and money) has been thrown at it.
Since they are often used for deep scientific applications that tend to be computationally unbounded, those super-computers can increase the scientists productivity by a huge amount. However, they are also built for “bragging rights”.