In a way I have a supercomputer in my pocket. But in another more important way, the pocket computer is a joke to true supercomputers – and Intel and Cray put together one of the biggest contracts with a billion dollar deal from the Department of Energy. It's going to do exaflops!
The "Aurora" program aims to assemble an "exascale" computing system for Argonne National Laboratory by 2021. "exa" is the prefix indicating bigness, in this case 1 quintillion floating point operations, or flops. They are the kind of horsepower of supercomputers.
By comparison, your average modern CPU may have a hundred or more giga flops. Thousand gigaflops make a teraflop, a thousand teraflops make a petaflop, and a thousand petaflops make an exaflop. So despite the huge advances in computing efficiency going into making super powerful smartphones and desktops, we are talking several size differences. (Let's not get into the GPUs, it's complicated.)
And even when compared to the biggest supercomputers and clusters out there, you still see a maximum of 200 petaflops (that would be IBM's summit, over at Oak Ridge National Lab) or so.
Just what do you need that kind of computing power for? Petaflops didn't want to do that? Well, no, actually. A very recent example of computational constraints in real-world surveys was this study of how climate change could affect cloud formation in certain regions, reinforce the trend and lead to a vicious cycle.
This kind of thing can only be estimated with much coarser models before; Computing resources were too tight to allow the type of extremely large number of variables involved here (or here ̵
The more computing resources we have, the more dedication to which the Intel press release offers as examples, "developing extreme-scale cosmological simulations, discovering new approaches to drug-prediction and discovery materials to create more efficient organic solar cells." 19659002] Intel says Aurora will be the first exaflop system in the United States – an important warning since China's mission is to perform the task one year earlier. There is no reason to believe that they will not achieve that either, since Chinese supercomputers have been reliable among the fastest in the world.
If you are curious about what ANL can put their soon-to-build computers to work for, please browse the research index. The short answer is "just about everything."