Intel unveils Frontera, the fastest academic supercomputer ever

In August 2018, Dell EMC and Intel announced a joint design of a supercomputer called Frontera, funded by $60 million from the National Science Foundation. This supercomputer will replace the Stampede 2 at the University of Texas at Austin (TACC). The above plan was realized in June this year, and Frontera was officially announced recently.

Intel claims that Frontera can achieve up to 38.7 petaflops, making it the world's fastest computer for academic workloads such as modeling and simulation, big data and machine learning. By comparison, Stampede 2 has a peak performance of 18 petaflops.

"The Frontera system will provide researchers with unprecedented computing and AI capabilities for academic research," said Tracy Dankroger, Intel vice president and general manager of Intel's Extreme Computing Organization. "With Intel technology, this new supercomputer opens up new possibilities in science and engineering to advance research, including understanding the universe, medical treatments and energy needs."

Most of Frontera's nodes are liquid-cooled, and each node is expected to consume about 65 kilowatts of power, most of which will come from solar and wind power.

In terms of storage, Frontera uses four different channels built by DataDirect Networks, totaling more than 50PB, and equipped with 3PB of NAND flash, which means that each node has about 480GB of SSD storage space. Three of the channels are interconnected, and the fourth channel has a very fast connection of 1.5TB per second.

Frontera has been used to simulate neutron star collision fusion and try to explain the origin of the explosive energy in the process. In addition, academic teams have also used it to establish a biophysical model of brain tumor development to more effectively diagnose and treat gliomas. The University of North Carolina is using Frontera to train an artificial intelligence model to describe the force field and potential energy of molecules based on their three-dimensional structure.