Intel and Dell EMC have unveiled Frontera. Frontera is an academic supercomputer to replace the Stampede2 in the University of Texas. Intel now claims to be the fastest supercomputer in the world for academic workloads.

According to Intel, the Frontera can achieve performance of 38.7 quadrillion floating points operations per second – or petaflops, writes Venturebeat. This should make it the fastest computer in the world designed for academic workloads such as modeling, simulation, big data and machine learning.

Dell EMC and Intel announced in August 2018 that they would be working together on the Frontera. For that, they received a $60 million grant from the National Science Foundation. The resulting device must replace Stampede 2, which achieved peak performance of 18 petaflops.

The Frontera was placed at the university in June this year. The device must offer researchers computational and artificial intelligence (AI) possibilities that have never existed before for academic research, says Intel vice president and general manager of the extreme computing organization of the chip manufacturer Trish Damkroger.

Specifications

The Frontera contains hundreds of 28-core 2nd Gen Xeon Scalable (Cascade Lake) processors in Dell EMC PowerEdge servers. It also contains Nvidia nodes for single-precision computing.

The architecture of the chips is based on Intel Advanced Vector Extensions 512 (AVX-512). This is a set of instructions that allows twice as many FLOPS per clock as the previous generation. For memory, the Intel Optane DC device uses persistent memory, which together with the Xeon Scalable Processors delivers performance of 287,000 operations per second.

The storage in the device is provided by four different environments created by DataDirect Networks. Together, it delivers more than 50 petabytes with an additional 3 petabytes of NAND flash capable. That amounts to 480 GB of SSD storage per node.

Already in use

As mentioned before, the Frontera has already been placed at the university. The device has already been used, for example by Astrophysics professor Manuela Campanelli of the Rochester Institute of Technology. Campanelli used the device to make a simulation that could possibly explain the origin of the energy explosions that occur when a neutron star is combined.

Assistant professor Olexandr Isayev of the University of North Carolina is now using the system to train an AI model that describes the force fields and potential energy of molecules based on their 3D structure.