Brian Guarraci is a software engineer who works on Twitter's search team. "I like my tweets how I like my computers... Fast," says his Twitter bio. Indeed. Guarraci built a supercomputer that manages to hit 208 gigaflop speeds at peak performance. That's more than 17 times faster than commercially available processors, according to the German blog Nerdcore.
You may recall the Parallella Kickstarter project that was supposed to "make parallel computing accessible to everyone" with its $99 boards. Guarraci used eight Parallella boards to perform fast computation. He also added with two Intel NUCs — tiny "next unit of computing" devices — that have the power of a third-generation Intel Core i3 processor along with 16 gigabytes of RAM and 120 gigabytes of SSD each.
The setup cost about $2,500 for the core, Guarraci tells The Verge, including roughly $800 for the Parallella boards, $1,000 for the NUCs, and $120 for the switch, power, and case.
The supercomputer's physical design was inspired by the famous 1976 Cray 1 supercomputer and the Connection Machine supercomputers of the 1980s, as well as the new cylindrical Mac Pro. "Although building a system with a round case is much harder than a rectangular system, it is definitely more satisfying," Guarraci told Parallella. He's calling the machine the Parallac.
Parallel computing, which runs processes simultaneously in order to perform tasks faster, is important because processors aren't getting much faster, Guarraci says. "The best way to get more done is to compute in parallel," he says. "One of the things that’s cool about a Parallella cluster is that you can perform multiple heterogenous parallel-computing tasks all at the same time. Some nodes can be doing image processing while other nodes are doing machine learning. It’s a very flexible setup."