The Oak Ridge National Laboratory (ORNL), part of the US Department of Energy, has announced that its three-year Titan supercomputer project is complete.
A mixed GPU and CPU design, the new supercomputer is capable of 20 petaflops (floating point operations per second), putting it on a par with the Sequioa IBM BlueGene/Q system launched in June this year by Lawrence Livermore National Laboratory, which is used for nuclear weapons simulations.
The Cray XK7 Titan system is equipped with 18,688 nodes that each have a 16-core AMD Opteron 6274 processor for a total of 299,088 cores, and an Nvidia Tesla K20 graphics accelerator.
A total of 46,645,248 CUDA cores are available through the Nvidia graphics cards. The supercomputer also has 710 terabytes of memory.
Part of the US Department of Energy's open science system, the supercomputer will be available to researchers from universities, government laboratories as well as private industries.
The supercomputer can model physical and biological phenomena faster than through experimentation alone, and is 10 times as fast as its predecessor, the 2.3 petaflops Jaguar system.
Power efficiency is also improved, with Titan having a budget of nine megawatts, just 30 percent more than Jaguar which required seven megawatts of energy.
Whereas the Jaguar supercomputer would carry out a simulation of a nuclear fuel rod being used in a reactor core in 60 hours, the Titan can do so in just 13 hours, ORNL said.
Climate change modelling will be one area of research to be performed on the Titan, as well as air quaility simulations.
"Titan will allow scientists to simulate physical systems more realistically and in far greater detail," said James Hack, director of ORNL's National Center for Computational Sciences.
"The improvements in simulation fidelity will accelerate progress in a wide range of research areas such as alternative energy and energy efficiency, the identification and development of novel and useful materials and the opportunity for more advanced climate projections."