HYBRID is an experimental cluster that explores an emerging revolutionary High Performance Computing technology based on transforming Graphical Processing Units (GPUs) into low-cost (per MFLOP) general-purpose computation units (GPGPUs). It will combine two quad-core processors and NVIDIA TESLA 10 series GPUs that support calculations in double precision.
The project consists of an interdisciplinary consortium that will use the cluster for science research, assess the applicability of GPGPU techniques for scientific applications, provide feedback on practical problems with the cluster utilization and demonstrate the overall proof of concept needed for further investment into the technology.
16 CPUs (8 nodes, 64 cores), 16 GPUs (4 Tesla S1070)
2 CPUs per node (Core 2 Quad Xeon), 2 GPUs per node
8 Gb memory per node