updated 06:48 am EDT, Mon October 29, 2012
GPU-based supercomputer to help biological, physical modelling
The world's fastest open-science supercomputer has finished its NVIDIA Kepler-based upgrade. The newly-rebranded Titan at the US Department of Energy's Oak Ridge National Laboratory in Tennessee now contains 18,688 Tesla K20 GPU accelerators that helps it achieve a peak performance of more than 20 petaflops.
Previously known as Jaguar, the 200-cabinet supercomputer now runs 10 times faster and five times more energy efficient than before. Each NVIDIA GPU is accompanied by a 16-core AMD Opteron 5274 processor, and allows the 299,008 CPU cores and 710 terabytes of memory to guide the simulations while the GPU does the bulk of the processing.
The massive amount of processing power that Titan is capable of will help government researchers and academics as well as private companies model physical and biological phenomena faster and more accurately than before. While the GPUs being used in Titan are not the same as those found in off-the-shelf graphics cards, technology developed for supercomputer efforts does transfer over.