NAS hyperwall Visualization System Upgraded with Ivy Bridge Nodes
Systems and visualization engineers at the NASA Advanced Supercomputing (NAS) Division have updated the hyperwall system with new hardware in order to meet the agency’s data analysis and visualization needs. This improved visualization capability will enable more detailed analysis of simulation results produced on the NAS facility's supercomputers, enhancing the ability of scientists and engineers to gain greater insight from their research.
The upgrade, completed at the beginning of October 2014, replaced 128 nodes of dual quad-core AMD Opteron 2354 (Barcelona) processors with nodes containing newer dual 10-core Intel E5-2680v2 (Ivy Bridge) processors. This increased the hyperwally’s peak processing power from 9 to 57 teraflops, quadrupled the amount of memory, and nearly tripled the interconnect bandwidth, while simultaneously decreasing the system's physical footprint and overall power consumption. Additionally, the systemy’s 128 NVIDIA GeForce GTX 480 graphics processing units (GPUs) were replaced with GeForce GTX 780 Ti GPUs, which provide nearly 400 gigabytes of graphics memory.
This latest expansion to the hyperwall will also help meet the challenges associated with analyzing the growing archive of high-resolution observational data gathered from NASA instruments such as telescopes, Earth-observing satellites, and planetary rovers.
“The additional memory and computing power of the new hardware lets us process larger datasets,” explained NAS visualization expert David Ellsworth. “This is needed because the power of the Pleiades supercomputer is constantly increasing, which means that our users’ datasets are always getting larger.”
The hyperwall’s 128 nodes are connected to a 128-screen tiled LCD wall at the NAS facility, arranged in an 8x16 configuration measuring 23-ft. wide by 10-ft. high. As one of the largest and most powerful visualization systems in the world, it provides a supercomputer-scale environment for NASA researchers to display and analyze high-dimensional datasets in a multitude of ways. Each of the nodes can display, process, and share data, allowing scientists to quickly explore their high-resolution results and pinpoint critical details in NASA's massive simulation and observational datasets.
