With more than 5,000 dual-core processors and a staggering 54 teraflops (54 trillion calculations/sec) of computing power, Oak Ridge National Laboratory's Cray XT3 — dubbed Jaguar — is helping solve some of science's grand challenges.
The DoE's Innovative and Novel Computational Theory and Experiment (Incite) program allocates processor hours to selected researchers based on the scientific merit of proposed work. Last year's call for proposals resulted in 43 submissions requesting more than 95 million processor hours. The proposals spanned 11 scientific disciplines: accelerator physics, astrophysics, chemical sciences, climate re-search, computer science, engineering physics, environmental science, fusion energy, life sciences, materials sciences, and nuclear physics.
Some examples:
Multidimensional simulations of core-collapse supernovae consume 3.55 million processor hours on Jaguar. The goal is to understand how stars more massive than 10 of our suns explode to produce many of the elements in the universe.
A nanomaterials project consumes 3.5 million processor hours on Jaguar and 300,000 hr on another supercomputer named Phoenix, a Cray X1E with 18.5 teraflops. This study aims to better understand complex functional nanostructures, which could lead to significantly faster and energy efficient electronic devices, as well as better materials for energy storage, transmission, and production.
A wave-plasma simulation project awarded 3 million processor hours on Jaguar could ultimately help remove a key obstacle to viable fusion power. Other users tapping into ORNL's supercomputers include Boeing, Dream-works Animation, General Atomics, Harvard Univ., Auburn Univ., and the Univ. of Washington/Howard Hughes Medical Institute.
In all, some 30 million processor hours will run this year on Jaguar, and roughly 6 million on Phoenix. Super-computers at Lawrence Berkeley, Argonne, and Pacific Northwest National Laboratories will absorb the remaining hours. To keep up with ever-growing demand, Jaguar is scheduled for upgrade to 100 teraflops by year's end. ORNL is managed by UT-Battelle for the DoE.
MAKE CONTACT
Incite, hpc.science.doe.gov/allocations/incite