Nov 22, 2010

UNM Named an Official NVIDIA CUDA Research Center



The computer graphics research program at the Advanced Graphics Lab in UNM's Electrical & Computer Engineering Department (ECE), along with the high-performance computing research being performed at UNM's Center for Advanced Research Computing (CARC), have together earned the University of New Mexico selection as an NVIDIA CUDA Research Center.

This selection follows the announcement in September of a National Science Foundation Major Research Instrumentation award of $435,077 to UNM CARC to acquire a GPU-accelerated supercomputer and associated terabyte storage system. The new supercomputer, Xena, was formally announced at the recent Supercomputing '10 conference in New Orleans and will enable UNM students and researchers to perform advanced, computationally intensive research.

CUDA is NVIDIA's parallel computing programming architecture that enables software developers to tap into the parallel processing power of graphics processing units (GPUs) to process numerical data as well as the geometry and pixels of computer-rendered images. CUDA thus enables researchers to perform a variety of processing-intensive scientific simulations, numerical solvers, and other research applications on GPUs.

A GPU is a highly parallelized microprocessor that offloads and accelerates calculations from the central processing unit (CPU). Originally, GPUs were developed to handle the graphics calculations required to render images at real-time rates for video games and other interactive applications. More recently, researchers have begun to port other, nongraphics applications to GPUs. NVIDIA's CUDA architecture makes it easier for software developers to harness the power of GPUs by supporting high-level programming languages such as Fortran, C, and C++.

NVIDIA selects just a few universities worldwide each year to become part of its CUDA Research Center Program. The company based its selection of UNM on "the vision, quality, and impact of [its] research leveraging CUDA technology," according to an announcement letter sent on November 11 to ECE Assistant Professor Pradeep Sen by NVIDIA Corporation's director of research.

"Being named a CUDA Research Center is an exciting opportunity for UNM," Sen said. "We really look forward to working with NVIDIA on interesting research problems that leverage the GPU-based supercomputer that we are in the process of building." Sen, a co-principal investigator (co-PI) on the NSF grant, initiated and led the effort to obtain NVIDIA CUDA Research Center status for UNM.

The selection brings with it NVIDIA Tesla C2070 GPUs based on the Fermi architecture. These are high-end GPUs designed to handle data-intensive computational tasks beyond graphics, especially the fast double-precision calculations required by science and engineering research teams. It also brings user licenses for NVIDIA's Parallel Nsight software, the first development environment for massively parallel computing integrated into Microsoft Visual Studio, thus allowing programmers to develop for both GPUs and CPUs within Visual Studio.

Long-Term Research Plan

The collaboration will enable the UNM team to get the most out of the new GPU-based, 256-processor, high-performance parallel computer that was funded in September by the $435,077 grant from the NSF Office of CyberInfrastructure and $186,462 in matching contributions from UNM. Physics & Astronomy Research Associate Professor Susan Atlas, director of CARC, is principal investigator for that NSF proposal, titled MRI: Acquisition of a GPU-Accelerated Parallel Supercomputer for Computational Science and Engineering Research at the University of New Mexico.

Her four co-PIs are Electrical & Computer Engineering (ECE) professors Sen and Jamesina Simpson, Chemistry Professor Hua Guo, and CARC Deputy Director Timothy Thomas, who is also a research faculty member in Physics & Astronomy. Their team includes 12 additional faculty members: Vince Calhoun of ECE; Walter Gerstle of Civil Engineering; Abhaya Datye, Boris Kiefer (NMSU) and Plamen Atanassov of Chemical & Nuclear Engineering; Zayd Leseman of Mechanical Engineering; Stephen Boyd, Keith Lidke and Sudhakar Prasad of Physics & Astronomy; Evangelos Coutsias of Mathematics & Statistics; Stephanie Ruby of Molecular Genetics & Microbiology at the UNM Health Sciences Center; and Mousumi Roy of Earth & Planetary Sciences. The 17 collaborating faculty span nine departments and three UNM colleges.

Xena will be housed at CARC in support of high-throughput, data-intensive applications for science and engineering research at UNM. Key research areas will be: nano-bio-materials science, including molecular biophysics, chemical and condensed matter physics, materials physics, mathematical biology, molecular biology, catalysis, novel sensor materials, and structural materials; advanced graphics, image processing and visualization, including biophysical imaging using quantum dots, 3D animation and rendering, fMRI image analysis and 3D computed tomography; and geophysics, including computational electromagnetics and geological modeling. The new system will increase CARC's available compute cycles from 5 to 41 TFlops peak and its corresponding online storage capacity from 23 to 55 TB.

GPU-Based Research

GPUs are efficient at manipulating computer graphics, and their highly parallel structure makes them more effective than general-purpose CPUs for a range of complex algorithms. Although GPU processors are typically thought of as performing the floating point operations required for playing 3D games or high-end 3D rendering, their increased capabilities in the past decade have allowed them to extend to fields beyond entertainment. GPU computing, where the GPU performs general-purpose scientific and engineering computing using both a CPU and GPU as heterogeneous co-processors, is an emerging field in disciplines such as computational biology, chemistry, geophysics, and materials physics. The sequential part of an application typically runs on the CPU while the computationally intensive part that can be parallelized is accelerated by the GPU. Applications run faster, and researchers can run simulations with more complex data sets or increase the mathematical detail they use to describe a physical system.

As a CUDA Research Center, UNM and its researchers will be able to leverage these new parallel architectures in order to push the limits of their simulations and run more complex data sets. Selection as a CUDA Research Center also provides training tailored to UNM's needs, technical liaisons, and participation in a network of NVIDIA engineers and researchers. The selection lasts one year and is then renewable.

#   #   #   #   #

UNM's Center for Advanced Research Computing (CARC, pronounced "CAR-see") is a Strategic Research Emphasis Center of the University of New Mexico and is funded by UNM's Office of the Vice President for Research. CARC provides the capabilities needed to continue the growth of advanced-computing-based research at UNM, including support for traditional supercomputing, parallel cluster-based computing, large-scale research databases, and novel architectures. The center fosters new, interdisciplinary collaborations based on computation, and it encourages novel applications of computation in research while continuing to grow traditional applications of computing in the science and engineering disciplines.

With 31 tenured and tenure-track professors, the Electrical & Computer Engineering Department in UNM's School of Engineering is the largest ECE department in the state of New Mexico. The department's two graduate programs, Computer Engineering and Electrical Engineering, are ranked among America's Best Graduate Schools by U.S. News & World Report, the only such programs in New Mexico that are ranked. Both undergraduate programs are accredited by the national Accreditation Board for Engineering and Technology. The department has a range of 14 research concentrations and offers graduate degrees in 14 areas; it confers more than 100 degrees annually; and it has an average enrollment of about 400 students.

The University of New Mexico is the state's flagship university, serving more than 32,000 students. It is North America's only institution designated both by the U.S. Department of Education as Hispanic-Serving and by the Carnegie Foundation as a research doctoral university "with very high research activity," the highest level of research that Carnegie designates.

NVIDIA Research explores challenging topics in visual, parallel, and mobile computing and spans many domains, including computer graphics, physical simulation, scientific computing, computational photography, programming languages, circuit design, and computer architecture. NVIDIA supports advances in these fields by collaborating with academic and industrial research institutions, and it disseminates results in technical conferences, journals, and other academic venues. The company's website states that "Institutions identified as CUDA Research Centers are doing world-changing research by leveraging CUDA and NVIDIA GPUs."

NVIDIA (NASDAQ:NVDA) awakened the world to the power of computer graphics when it invented the GPU in 1999. Since then, it has consistently set new standards in visual computing with breathtaking, interactive graphics available on devices ranging from tablets and portable media players to notebooks and workstations. NVIDIA's expertise in programmable GPUs has led to breakthroughs in parallel processing that make supercomputing inexpensive and widely accessible. The company holds more than 1,600 patents worldwide, including designs and insights that are essential to modern computing.