HPC News, SGI, Blue Waters, Dell
Here’s our review of today’s noteworthy links for the High Performance Computing (HPC) industry:
Cray delivers first Blue Waters Cabinet. On December 1 Cray delivered the first full cabinet for the NCSA Blue Waters system. A photo gallery of the installation day can be found on the NCSA Facebook album, where in the comments it is confirmed that the cabinets will be water-cooled. The National Science Foundation’s Blue Waters project was awarded to Cray last month after NCSA and IBM terminated the original contract last summer.
Dell’s HPC Strategy. The Register reports on how Dell is going to engage the market to grow its HPC strategy. The primary focus for Dell’s HPC strategy is to concentrate on smaller HPC systems where projects are well-bounded with known workloads and customers they know and understand. Dell is putting together recipes for popular HPC apps in small, medium and large configurations. Each of these is fully tested and guaranteed to hit the promised level of performance. Dell’s High Performance Computing web site lists strategic insights, recommended configurations and a High Performance Cluster Advisory tool.
SGI selected for Poznan HPC System. SGI announced that Poland’s Poznan Supercomputing and Networking Center (PSNC) has purchased an SGI HPC solution. The purchased system is comprised of an SGI Rackable C1103-G15 cluster utilizing AMD Opteron 6200 series processors, and includes 120 servers each with one NVIDIA M2050 GPU per node and another 107 servers with two M2050 GPUs per server. The full configuration is expected to deliver 224 peak teraflops and will feature 12.6 TB of memory, 5448 CPU cores and 149632 GPU CUDA cores, and is intended to satisfy the increasing performance requirements of Poznan Supercomputing Center users. ”We decided to go for the combination of the most modern CPU technology together with proven GPU technology,” said Norbert Meyer, director of the Supercomputing facility at Poznan. “Such a combination allows us to ensure optimal parameters for our users. This system, which was financed from structural funds coming from POWIEW and PL-GRID national projects and will be also used in the PRACE (EU project), allowed us to reach 63 teraflops, enough for a place on the TOP500 list today, with less than half of nodes installed for the test. The full configuration will feature more than twice as many more nodes than the measured system, which ensures enough performance for our users as well as a good position on June 2012 TOP500 list.”