UCLA Physics & Astronomy Department  

The UCLA Physics & Astronomy Department is pleased to announce that the department’s Plasma Simulation Group (Warren B. Mori, Philip L. Pritchett, Scott A.Friedman, Tajendra V. Singh, Viktor K. Decyk) with support from the UCLA Institute for Digital Research and Education (IDRE), have been awarded a very competitive National Science Foundation major research instrumentation (MRI) grant for nearly $1.8M, to build a 96-node dual core cluster based on general purpose graphical processing units. It will be called the Dawson 2 computer, and will function as a tool for making major advances in plasma-based accelerator and fusion energy research, as well as providing a resource for campus-wide efforts in implementing advanced architectures for computational based research. This is an essential step in allowing UCLA researchers to be able to compete at the cutting edge of computational based research. This award will have a significant impact on our research infrastructure. It has very positive implications for our programs in computation, student training, and helping guide the direction of frontier experiments in areas of critical scientific importance.

Project Summary
This is a proposal to build a 96-node dual quad core (768 i7 processors) cluster with 48 Nvidia Tesla units (192 GPUs). The network would be based on QDR Infiniband with full cross section bandwidth. Such a system would provide a peak speed in single precision of ~200TFLOPS and it would have the same memory bandwidth between nodes and between the GPU’s to the node. The cluster would be housed within the Institute for Digital Research and Education (IDRE). The cluster would have three main purposes. First, it would serve as a testbed for porting parallel particle-in-cell (PIC) codes used to study a variety of problems to GPU based systems. This experimentation could lead to parallel PIC algorithms that would run efficiently on future heterogeneous multi-core computers housed at large supercomputer centers. Second, if codes are successfully ported then the cluster will be an instrument for enabling breakthroughs in new particle accelerator technology, fusion energy, space plasmas, astrophysics, and basic plasma science. Third, the cluster will help to advance research and education in broad and diverse areas of Computational Science at UCLA.