New Air Force 3.1 Petaflop Thunder Supercomputer helping develop hypersonic vehicles and railguns

The Air Force Research Laboratory officially presented its latest supercomputer acquisition in a ribbon-cutting ceremony Dec. 14, 2015.

The Thunder supercomputer is part of the DOD High Performance Computing Modernization Program. It joins two other large systems — Spirit and Lightning — already located at the center. The Silicon Graphics Incorporated ICE X is named Thunder after the Air Force’s P-47 Thunderbolt and its subsequent namesake, the A-10 aircraft, both of which have played key roles in significant armed conflict for the U.S. military.

Since beginning operations in October, Thunder has solved complex simulations ranging from hypersonic flight to the limitations of a futuristic electromagnetic rail gun.

Aerospace engineer Susan Cox-Stouffer used computational fluid dynamic simulations on the AFRL supercomputers to test the X-51 Waverider, a hypersonic vehicle that reached more than five times the speed of sound during flight tests over the Pacific Ocean.

“You can’t design these on the back of an envelope,” she said. “It takes a lot of simulations.”

The newest supercomputer is the 21st fastest high-performance computing system in the world, and can calculate about 3.1 petaFLOPS, or 3,126,240,000,000,000 floating point operations per second, according to AFRL.

New SGI ICE XA

To address the increasing size and volume of HPC workloads, SGI has developed a 6th generation supercomputer, SGI ICE XA. Extending SGI’s technology leadership, ICE XA enables scientists, researchers and engineers to achieve and accelerate computational breakthroughs at Petascale and with high return on investment.

  • Fully leverage leading commercial applications as well as in-house developed codes using industry-standard, state-of-the art x86, Linux, and integrated InfiniBand technologies
  • Enjoy optimum flexibility in processor and accelerator mix, blade and storage configuration, and network topology to meet specific workload requirements
  • Reduce operational costs through superior power and cooling efficiency coupled with advanced SGI software for system, health and power management
  • Power up and be production-ready in hours to days, not weeks to months
  • Grow existing ICE clusters by adding next generation nodes and without user interruption
  • Capitalize on SGI’s path to Exascale and continuing innovation to solve Terascale and Petascale problems

SPEED

SGI ICE XA is designed to run complex HPC workloads at petaflop speed. Compute nodes feature the Intel® Xeon® processor E5-2600 v3 series – SGI’s expertise in maximizing Xeon performance is reflected in successive world records to include SPEC MPI2007 benchmarks for the Xeon® E5-2600 v3. Nodes can be further augmented with Intel® Xeon Phi™ Coprocessors or NVIDIA ® GPU Accelerators. In addition, SGI Application Engineers have extensive expertise tuning systems to specific workload environments. What matters most is that SGI’s performance leadership extends from the laboratory to customer sites, which utilizing pure Intel® Xeon®, InfiniBand and Linux technology includes the most powerful commercial supercomputer in the TOP500.

SCALE

SGI ICE XA can provide up to 191 teraflops per rack and grow to tens of thousands of nodes with minimal increase in overhead. Choose from two high-density blade types – a quad node compute blade delivering maximum Intel® Xeon® processor capabilities, or a dual node compute blade that coupled with HDD/SSD drives or PCI-e slots increases processing and I/O capabilities. Blade enclosures provide power, cooling, system control, and network fabric for up to 9 compute blades via an integrated midplane, and up to four enclosures in a single rack. SGI ICE XA utilizes industry-standard InfiniBand networking with complete flexibility in topology. Choose from All-to-All, Fat Tree, Hypercube, or Enhanced Hypercube fabrics with single or dual plane to best meet performance, size, application, and budget needs. ICE XA runs standard SUSE® Linux Enterprise Server, Red Hat® Enterprise Linux®, or CentOS. Different Linux operating systems can be provisioned on different nodes, thereby allowing a broad range of Linux-based HPC applications to run simultaneously.

EFFICIENCY

SGI ICE is the industry leader in Megaflops per Watt for x86 platforms and currently holds 4 of the top 5 Most Efficient Supercomputers in Top500. ICE XA extends SGI’s leadership in power and cooling efficiency with 2nd generation SGI Cell technology. Systems are deployed in sealed E-Cells, with multiple cells connected to form large systems. “Closed-Loop Airflow” ensures no air from within the cell is mixed with data center air. E-Cells also utilize facility-supplied water for cooling and will not add heat to the data center if the water temperature is within 45-90 degrees Fahrenheit (7-32C). Often 30% or more efficient than industry competitors, this high 90F “Room Neutral” water temperature can save millions in cooling costs.

SGI ICE X

SGI ICE X is SGI’s 5th generation ICE system and powers many of the largest supercomputing environments in the world. ICE X delivers the record speed of ICE XA, a variety of high-density blades, and optimum Infiniband flexibilit