HPC

US Air Force boots up not one but two AMD-powered supercomputers after five years of Intel Haswell CPUs

Fawbush and Miller supers will crank away at weather-forecasting models


HPE has built two supercomputers focused on weather forecasting for the US Air Force that, when combined, are said to be six times more powerful than their predecessor.

Both of the Cray EX supers are up and running at Oak Ridge National Laboratory in Tennessee. They have 800 compute nodes each, with each node made up of two 64-core second-generation AMD Epyc processors clocked at 2.2GHz, we’re told.

“The new system enables significantly more complex physics in the weather models that will allow the Air Force to more accurately model clouds, for example, and to dramatically increase the resolution of the regional and global weather forecasts with a significantly improved ‘time to solution’,” Jim Rogers, Computing and Facilities Director for the National Center for Computational Science at the Oak Ridge National Laboratory, told The Register.

A graphic of what the El Capitan supercomputer will look like

Look at me. Look at me. I'm the El Capitan now: Cray to build US govt's $600m cray-cray exascale nuke app super

READ MORE

When combined, the peak performance of the supercomputers is 7.2 petaFLOPs, making it 6.5 times faster than the Air Force’s previous machine Thor, which was built in 2016. After nearly five years of using Intel Xeon v3 Haswell processors, the Air Force has tapped up Intel’s competitor AMD.

The two HPE systems are named “Fawbush" and “Miller” after famous Air Force meteorologists Major Ernest Fawbush and Captain Robert Miller, who correctly predicted a tornado was about to hit the Tinker Air Force base in Oklahoma in 1948, hours before the storm tore through. It was the first time a tornado had been forecast; all servicemen evacuated the base in time, and no one was hurt.

The Fawbush and Miller systems will allow meteorologists to build more accurate weather forecasting models capable of mapping out the atmospheric changes with resolutions of 10km, a big improvement on the current 17km.

The simulations are computationally intensive and require a steady pipeline of live weather data. The results are packaged as weather alerts that are issued to pilots deployed on defense missions around the world. ®

Fascinated by supercomputers? Check out our sister site The Next Platform's analysis of China’s Sunway exascale colossus right here.

Similar topics


Other stories you might like

Biting the hand that feeds IT © 1998–2021