The Minnesota Supercomputing Institute (MSI) has installed a new graphics processing unit (GPU) system that it hopes will enable researchers to take advantage of this new technology for innovation and improved performance for research.
This new system complements MSI's current set of GPU tutorial nodes (four nodes with four Nvidia GTX480 GPUs) and the S2050 GPUs available to NIH researchers on the Koronis system. The new system includes a Dell R710 dual-socket, quad-core head/login node in a C6100 chassis, with 48 GB of memory; eight Dell compute nodes, each with dual X5675 six-core 3.06 GHz processors and 96 GB of memory; and 32 Nvidia M2070 GPUs.
Each compute node is connected to four GPU nodes, and the GPUs, together, have 448 3.13 GHz cores and 3 GB of memory per card. Moreover, each GPU is capable of 1.2 single-precision TFLOPS and 0.5 double-precision TFLOPs. The external chassis allows for easy expansion of the system to a larger node count and also connection to a variable number of GPUs per node. The system is interconnected by QDR infiniband and cooled with a passive, water-cooled door. Installation was completed the week of Dec. 6-10, 2011.