TMCnet News

SGI Announces Software Support for Intel(R) Xeon Phi(TM)
[January 28, 2013]

SGI Announces Software Support for Intel(R) Xeon Phi(TM)


(GlobeNewswire Via Acquire Media NewsEdge) FREMONT, Calif., Jan. 28, 2013 (GLOBE NEWSWIRE) -- SGI (Nasdaq:SGI), the trusted leader in technical computing, today announced the availability of SGI software tools that enable customers and software developers to get the most value from the Intel® Xeon Phi™ coprocessor. Additionally, SGI is engaged in software development activities with third party software providers to build and test optimized applications that take advantage of the new coprocessor technology. By combining the power of Intel Xeon Phi, which offers over one teraflop double-precision performance per coprocessor1 with SGI's leading server platforms, customers can now take advantage of dramatic performance improvements for parallel workloads in an x86 programming model.



SGI UPC (Unified Parallel C) compiler, the first UPC compiler for Intel Xeon Phi, supports MPSS, the coprocessor software stack. It enables PGAS programming on SGI servers running Intel Xeon Phi. SGI UPC supports applications in native and offload modes. SGI MPInside, an advanced profiling and performance analysis tool that helps developers find bottlenecks in MPI code, now also runs on Intel Xeon Phi. SGI MPInside provides developers key capabilities to improve MPI application performance enabling "what-if" studies to project how code will perform on future architectures.

SGI is working with its large ecosystem of customers and ISV partners to optimize existing software applications to take advantage of the performance improvements provided by Intel Xeon Phi coprocessors. ISVs such as ESI Group (OpenFOAM), Multipath (FMSlib) and key customers such as the Irish Center for High End Computing (ICHEC) and others have joined with SGI and Intel to take full advantage of the new coprocessor capabilities.


"The Intel Xeon Phi platform has proven to be the most productive many-core environment that I have worked with. During our last collaborative project with SGI on this platform, a researcher's application was successfully ported in just a few hours using standard OpenMP directives and achieved over 99.9% parallel efficiency," said Gilles Civario, senior software architect, Irish Centre for High-End Computing (ICHEC). "One of the key features of the Intel Xeon Phi is its ease of use, making many-core processing accessible to users who would otherwise avoid such technology. As a national HPC service provider, this is also a very important aspect of Intel's latest technology." For scale-out, scale-up and hybrid deployments, SGI provides everything needed for a complete Intel Xeon Phi coprocessor solution including factory-integrated hardware and software "starter kits" and worldwide support and services complemented by the subject matter know-how of fifty experts across many application disciplines and fields of science. With SGI Management Center Premium Edition, the Intel Xeon Phi coprocessor can be monitored and managed using an interactive Graphical User Interface (GUI) that also manages all the other system components.

"To maximize the performance benefits of the Intel Xeon Phi coprocessor, customers not only require the best hardware platform but also software that is built and engineered for it," said Christian Tanasescu, vice president of applications and solutions engineering, SGI. "SGI provides the complete solution as well as continues to invest in enabling the ecosystem to provide ongoing software offerings for Intel Xeon Phi." SGI supports Intel Xeon Phi coprocessors across all server products and is a leading provider of performance with Intel Xeon Phi. SGI offers the highest Linpack efficiency for 32GB memory per node1. Also, SGI® UV™ with Intel Xeon Phi is the most scalable coprocessor system available today, scaling up to 4,096 cores and 64TB of memory. SGI UV is the only system on the market that can support 32 Intel Xeon Phi coprocessors in a single system image, which can add 50 teraflops of sustained SGEMM performance in a single node2. This configuration also drives in excess of 190GB/s of data transfer bandwidth across the PCIe interfaces to system memory, an unmatched level of I/O performance for a single system in the industry3.

SGI UV, the Big Brain computer, can accelerate a range of compute and data-intensive applications, in a single system image, operating just like a workstation. This makes SGI UV easy to manage, which is especially important when programming in a hybrid CPU and coprocessor environment.

"HPC customers require technology not only to deliver the best processing and energy efficiency, but also to speed advanced codes and algorithms to deployment," said Raj Hazra, Intel VP and GM of the Technical Computing Group. "SGI's UPC compiler is leveraging the familiar programming model of Intel Xeon Phi coprocessors. This allows customers to instantly take advantage of Intel's new many-core technology when reusing the existing code and to achieve expected increase in performance." Availability SGI UPC and SGI MPInside and supported SGI hardware platforms for Intel Xeon Phi 5110p are available today.

About SGI SGI, the trusted leader in technical computing, is focused on helping customers solve their most demanding business and technology challenges. Visit sgi.com for more information.

The Silicon Graphics, Inc. logo is available at //www.globenewswire.com/newsroom/prs/ pkgid=14848 Connect with SGI on Twitter (@sgi_corp), YouTube (youtube.com/sgicorp), Facebook (facebook.com/sgiglobal) and LinkedIn.

Contact Information: Ogilvy Public Relations Meghan Fintland 415-677-2704 [email protected] © 2013 Silicon Graphics International Corp. SGI, UV and the SGI logo are trademarks or registered trademarks of Silicon Graphics International Corp. or its subsidiaries in the United States and/or other countries. Intel, Xeon and Phi are registered trademarks of Intel Corporation. All other trademarks are property of their respective holders.

1 212.9 TF on 128 nodes with 2x Intel Xeon E5-2670 and 2x Intel Xeon Phi 5110P with 32 GB per compute node, 70.6% overall efficiency (includes processors and coprocessors).

2 50.1 TF on SGI UV 2000 with 32x Intel Xeon E5-2670 and 32x Intel Xeon Phi 5110P, offload mode.

3 >190GB/s PCIe bandwidth achieved on SGI UV 2000 with 32x Intel Xeon E5-2670 and 32x Intel Xeon Phi 5110P, offload mode.

Source: SGI 2013 GlobeNewswire, Inc.

[ Back To TMCnet.com's Homepage ]