Vector Processing Facilities
Most of the group's efforts in the area of vector processing involve Graphic Processing Units (GPUs). The group's "workhorses" are its four NVIDIA Tesla S1070 computing systems. An S1070 consists of 4 Tesla GPUs each of which has 240 cores (for a total of 960 cores).
Each core operates at about 1.4 GHz. Hence, each Tesla S1070 has a peak single precision floating point performance of about 4 TFlops.
Custom software is being developed for these systems using both CUDA and OpenCL. In addition, these systems support several "off the shelf" software products.
Shared Memory Computing Facilities
The group has several multi-core/multi-processor computing systems. The "workhorse" is a server with four quad-core Xeon CPUs and 128GB of fast RAM and eight 150GB hard drives.
Custom software is being developed for these systems using OpenMP and other frameowkrs/architectures.
Cluster/Grid Computing Facilities
The group uses a variety of different cluster-based (e.g., Beowulf) and grid-based (e.g., ) computing systems. Some of these systems are dedicated to high-performance computing (e.g., the high-resolution display wall) and some of which serve multiple purposes (e.g., the cross-campus grid).
