Penguin has been delivering integrated GPU computing clusters since the version 1.0 of this technology,” said CEO Charles Wuischpard. “The new Relion 2808GT platform in conjunction with the latest GPU and coprocessor technology delivers unprecedented levels of performance. The Relion 2808GT enables our HPC customers to further accelerate their research by shortening the time to result for their simulations.”

In terms of computational density, a fully configured server with eight NVidia K20 GPUs can achieve 28 TFLOPs of single precision floating point performance.

The Relion 2808GT will be displayed at the NVidia GPU Technology Conference from March 18 – 21 in San Jose, California. Read the Full Story.

Related posts:

• Video: Penguin Computing Showcases High Density Relion 2808GT GPU Server
• First Up: Penguin Offers Early Access to Future Xeons, Delivers World’s First AMD APU Cluster at Sandia
• Penguin Computing Adds Kepler GPU Acceleration and More

The post New Penguin Relion 2808GT Delivers Leading Compute Density for HPC Apps appeared first on insideHPC.

The initiative is intended to accelerate the structural analysis and virtual testing of their next-generation aircraft, starting with the A350-900.

Airbus’ Advans program (accelerated development for vulnerability and nonlinear system) consists of innovative simulation methods to apply the advanced nonlinear dynamic analysis capabilities of Simulia Abaqus to predict, with a high level of confidence, the real-world strength and performance of its aircraft structures.

Accurate, nonlinear analysis, done earlier in the design phase and prior to performing costly physical tests, results in a higher quality design, and at a lower cost. Airbus designers and engineers will create very large simulation models, leveraging their aircraft models designed in the CATIA application. Then, by using high-performance computing clusters, they will be able to perform full-scale, nonlinear structural simulations with Simulia Abaqus overnight.

With this project, we are pushing the boundaries of high-performance computing and our physics-based simulation applications to validate full-scale aircraft designs overnight,’ said Monica Menghini, executive vice president, industry and marketing, Dassault Systèmes. The 3DExperience platform allows users to experience the accurate physical performance of their aircraft design much sooner and well before committing to physical tests. This accelerates their ability to make highly-informed, mission-critical design and engineering decisions.

This story appears here as part of a cross-publishing agreement with Scientific Computing World.

Related posts:

• Dassault Systemes Releases 3DVIA Mobile in Time for NVIDIA GPU Tech Conference
• Mercury Marine Rocks the Boat with Windows HPC
• ANSYS to be used by ITER in building first fusion facility

The post Airbus Teams with Dassault for 3D Design appeared first on insideHPC.

The HPC industry is rapidly embracing hybrid computing architectures based on their superior performance and energy efficiency,” said Sumit Gupta, director of Tesla GPU computing at NVIDIA. “Higher-performance Sandy Bridge CPUs unleash the power of NVIDIA GPUs to an even greater degree, enabling them to maximize acceleration for a range of commercial and scientific HPC applications, include AMBER, GROMACS, LAMMPS, Simulia Abaqus, seismic processing, and others.”

Read the Full Story.

Related posts:

• Dell Spikes Xeon E5/R Servers with Nvidia GPUs
• Benchmarks: Sandy Bridge and Tesla a Powerful Combo
• NVIDIA Inside New IBM Servers

The post New HP Servers Designed From Ground Up for GPU Computing appeared first on insideHPC.

Nvidia has posted videos and slides from the recent GPU Technology Conference in Beijing. You can see the full listing here, but we’ve gone ahead an highlighted some of the featured talks in the the HPC track.

Weather & Climate Modeling

• GPU-based Operational Weather Model with Horizontal 500m Resolution. Takayuki Aoki - Tokyo Institute of Technology. MP4 * MP4 CN * PDF
• Unified Modeling System for Seamless Weather and Climate Predictions of Monsoons. Subodh Kumar - IIT Delhi, New Delhi, India. MP4 * MP4 CN
• GPU Considerations for Next Generation Weather and Climate Simulations. Thomas Schulthess - Swiss National Supercomputing Center, ETH Zurich. MP4 * MP4 CN * PDF
• GPU Computing in Numerical Space Weather Modeling. Xueshang Feng - State Key Lab of Space Weather, Chinese Academy of Sciences. MP4 * MP4 CN

Digital Manufacturing

• Using GPUs to Power Abaqus Finite Element Analysis to Enhance Speed. Jack Huang - Dassault Syst. MP4 * MP4 CN * PDF

• Using GPU Acceleration in SAIC Motor. Yang Weng - SAIC Motor. MP4 * MP4 CN

• From the Dynamics of Sand to the Dynamics of Tanks: Using GPU Computing to Advance the Pace of Innovation and Improve Designs in Mechanical Engineering. Dan Negrut - Simulation-Based Engineering Lab Wisconsin Applied Computing Center University of Wisconsin. MP4 * MP4 CN * PDF

GPUs in HPC

• A Fast, Scalable High Order Compressible Flow Solver for Unstructured Grids. Patrice Castonguay - Stanford University. MP4 * MP4 CN * PDF

• State-of-the-Art Analysis and Perspectives of China HPC Development: A View from 2011 HPC TOP100. Yunquan Zhang - Institute of Software, Chinese Academy of Science. PDF

• Tsubame 2.0: Running a 4,000 GPU Supercomputer in Full Production. Satoshi Matsuoka - Tokyo Institute of Technology, Global Scientific Information and Computing Center. MP4 * MP4 CN * PDF
• The Critical Role of Software Infrastructure in Future Exascale Systems. Jeff Vetter - Oak Ridge National Laboratory & Georgia Institute of Technology. MP4 * MP4 CN
• Why Hybrid Systems Work So Well for Simulation Based Sciences. Thomas Schulthess - Swiss National Supercomputing Center, ETH Zurich.
• GPUs are Revolutionizing Supercomputing. Wen-mei Hwu - University of Illinois at Urbana-Champaign. MP4 * MP4 CN * PDF
• Opportunities and Challenges Posed by Exascale Computing: ORNL’s Plans and Perspectives. Jeff Nichols - Oak Ridge National Laboratory. MP4 * MP4 CN
• Petaflop Biofluidics Simulations on the TSUBAME 2.0 Supercomputer. Simone Melchionna - National Research Council Italy. MP4 * MP4 CN
• Opening Keynote with Jen-Hsun Huang, NVIDIA. Jen-Hsun Huang - NVIDIA. WMV * MP4

Related posts:

• GPU Technology Conference Heads to Beijing, Dec. 14-15, 2011
• GPU Technology Conference Agenda Chocked Full of Supercomputing Sessions
• Travel Preview – GPU Technology Conference Comes to Beijing Dec 14-15

The post Featured Sessions from GPU Technology Conference in Beijing appeared first on insideHPC.

In this video, panelist from the Mathworks GPU Roundtable at SC11 discuss direction and HPC challenges and the gap between software and hardware, cloud, and the future of GPUs. I was only able to catch the last 18 minutes of the discussion, but I think panelists all did a great job.

Panelists (left to right):

• Professor Anne Trefethen is a Director of the Oxford e-Research Centre. Before joining Oxford Anne was Director of the UK e-Science Core Programme, having been the Deputy Director for four years. The Core Programme focussed on the generic issues for e-Science applications and Grid infrastructure through the development of appropriate middleware and infrastructure in collaboration with UK industry.

• Jerry Chen is responsible for strategic alliances for NVIDIA’s HPC business, including responsibility for NVIDIA’s partnership with the MathWorks. Prior to NVIDIA he was a developer for Abaqus (SIMULIA), and holds a Masters of Engineering from Cornell University and an MBA from the University of California at Berkeley.

• Chris Willard is Chief Research Officer at Intersect360 Research, where he leads the demand-side and supplier-driven data analysis practice for Intersect360 Research’s forward-looking market intelligence subscription service. In addition, he supports a range of client-specific services including custom research studies and strategic consulting.

• Jos Martin is currently the principal architect for parallel computing tools at MathWorks, leading the team that develops infrastructure and tools for enabling parallel computing within the MATLAB environment.

• David A. Lifka is the Director of the Cornell University Center for Advanced Computing and the Director of Research Computing at Weill Cornell Medical College. Lifka was recently named Coordinator of Architecture and Design for the NSF’s XSEDE program.

Related posts:

• MathWorks expands support for parallelism, announces TeraGrid tie-in
• MathWorks extends Parallel Computing Toolbox integration
• Video: SC10 Analyst Crossfire Program is Must-See TV

The post Video: Mathworks GPU Roundtable at SC11 appeared first on insideHPC.

I talked to Platform’s VP of Product Management Ken Hertzler recently and he told me how the design of Platform HPC centered around two design points when it was launched 18 months ago:

• Can we make it easier to set up a cluster?
• Can we make it easier to integrate your applications to use the cluster?

We are addressing three major areas with this release. The first is that Platform HPC 2.1 is a complete cluster. In a cluster solution, you need various tools like a cluster manager, a workload scheduler, portal support and application portal interfaces, and reporting and management. Plus it supports Windows and Linux. The second thing is we are supporting and integrating some of the most popular ISV applications such as ANSYS, LS-DYNA, ABAQUS, and others. The third point is that we have certified vendors that this solution is sold through. That way all the right device drivers are there and the customer gets one point of contact.

Platform will be demonstrating the new product at SC10 booth number 2739.

Related posts:

• Platform launches Application Center
• QLogic and Platform Computing Collaborate on Integrating HPC Cluster Environments
• Platform Gets Graphic with HPC Cluster Manager

The post Platform HPC Update Provides a Complete Cluster Solution appeared first on insideHPC.

When company executives asked analysts to accelerate an upcoming project, they migrated from a loosely coupled cluster of Linux systems to an in InfiniBand cluster of IBM servers running Windows HPC Server 2008.

We’re able to achieve more realism in our simulations and more refinement in our designs,” says Arden Anderson, CAE Engineer for Mercury Marine. “As a result, we’re increasing durability, reliability, and weight optimization in our products.”

So far, the migration from Linux seems to be going well and the company’s analysts appreciate the cluster’s ease of use. “It’s so much faster and more intuitive for analysts to submit and manage their jobs, and everyone’s pleased with how efficiently Abaqus runs on the cluster,” says Paul Radavich, Manager of Mercury Marine Design Analysis Group.

Related posts:

• Mercury releases PS3 SDK
• Mercury launches new visualization subsidiary
• Mercury puts 100 GFLOPS in your hand

The post Mercury Marine Rocks the Boat with Windows HPC appeared first on insideHPC.

This week Platform Computing announced that it was turning some of its focus back to HPC, after spending the past year or so aiming at a more traditional large-IT audience. On Monday the company announced the launch of HPC Enterprise Edition, a package of tools that Platform believes will help “enterprise” (I read that as either non-traditional or inexperienced) users in the commercial market ease into HPC.

Platform HPC Enterprise Edition brings together several pieces of Platform’s HPC toolkit into one offering that will help customers manage their cluster from deployment and management to monitoring and job submission. It includes Cluster Manager, LSF, Platform MPI (recall that Platform bought MPI stacks from both HP and Scali), Application Center, RTM monitoring dashboard, and ISF Adaptive Cluster into a single, web-based interface for the entire system lifecycle.

Enterprise Edition is the evolution of what Platform used to call Platform HPC Workgroup Manager — now called HPC Workgroup Edition — but without the 32-node limitation of that solution. Although it could theoretically be used to manage systems of arbitrary size, it really is designed for smaller clusters and COTS applications.

Platform hopes that it can ease the cluster integration and training worries of the IT admin group by providing its pre-integrated tools for installation, management, and monitoring. HPC EE will probably run on your favorite x86-64 Linux, with support for popular distros such as Red Hat, SUSE, CentOS, and Scientific Linux. If you are a mixed Windows/Linux shop then Adaptive Cluster, which integrates with LSF to reconfigure a pool of nodes for either Linux- or Windows-based jobs on the fly, will also address the real requirement that companies not have to reinvent their application infrastructure to move to HPC.

Getting users over the hurdle

But it is the inclusion of Platform Application Center and the pre-integrated COTS applications templates that will help get the users of these systems over some of their entry hurdles. Launched during SC09, Application Center is a shrink-wrapped role-based portal framework. Users or system managers can create templates for various applications and types of jobs (using XML) that allow users to launch applications and manage files from within a web portal. HPC Enterprise Edition comes with seven templates that integrate applications popular with Platform customers: Abaqus, Ansys, Blast, Eclipse, Fluent, LS-Dyna, and Nastran. Visualization is not supported out of the box with HPC Enterprise Edition, but users can upgrade their installation to enable vis in the portal.

The pre-integration of those seven application templates is part of what William Lu, Platform’s director of HPC marketing, emphasized when I spoke with him as the shrink-wrapped advantage. “HPC Enterprise Edition groups our existing commercial-grade technologies together into a single solution,” Lu says. “It is not just a collection of open source technologies.” Whether this line of marketing continues to hold up over time is entirely up to the open source community.

If you’re going to have an HPC hardware vendor as your launch partner…

When asked about pricing Lu said that HPC Enterprise Edition will run customers “a few hundred dollars per node,” and that Platform expects the majority of licenses will ship out through hardware partners, although they will also sell to users directly through their existing sales channels (you can’t buy this online at the website — not yet, anyway). HPC Enterprise Edition is already available, and Platform is launching with a significant HPC hardware partner: Cray. Cray is rebranding the software with its own Cray Cluster Manager nameplate and shipping it to CX1 and CX1000 customers. Although Lu wasn’t ready to talk about other partners yet, he did mention that they are in discussions with Dell and HP.

Interestingly, Cray’s Ian Miller (senior vice president of the productivity solutions group and marketing) told me that Cray Cluster Manager is now the default clusterware on both the CX1 and the CX1000. The Cray CX1-iWS sold by Dell won’t ship with CCM, which makes sense since it is entirely a Windows platform. “However, some customers have their own preferred cluster management software,” says Miller, “and some of our resellers are also aligned with other partners in this space, giving us the flexibility to be able to handle these situations too.”

Related posts:

• Platform Computing Announces New Cluster Management Tool
• Platform aims at cluster management market
• Platform teams with NVIDIA for GPU management goodness

The post Platform turns its HPC management software up a notch appeared first on insideHPC.

Scalability is a key component of designing Cray supercomputers and this latest version of our Cray Linux Environment continues to be one of the most scalable Linux environments for high performance computing,” said Barry Bolding, Cray’s vice president of scalable systems. “The introduction of Cluster Compatibility Mode significantly enhances the compatibility features of our operating system. Now the most powerful supercomputers in the world can run our customers’ key ISV applications right out of the box.”

The Cray Linux Environment was designed explicitly to scale to 500,000+ cores in an environment made just for HPC.

SIMULIA and Cray have worked closely for several years to ensure that SIMULIA’s Abaqus FEA software operates efficiently on Cray computing systems,” said Matt Dunbar, Chief Architect, SIMULIA. “With the launch of the Cluster Compatibility Mode, Cray has made significant improvements which expand the computing options for Abaqus users. It is now possible to run Abaqus on Cray XT supercomputers without the need for porting or modifications.”

Accelrys scientific modeling software can now run on the Cray XT line of supercomputing systems without porting or modifications,” said Paul Burrin, Senior Vice President and Chief Marketing Officer, Accelrys. “We are thrilled that Cray’s Cluster Compatibility Mode now gives us the ability to make our applications available on these high-end platforms. Some of our customers need higher levels of performance and scalability to address demanding problems, and Cray’s latest OS release makes the powerful Cray XT line available to our customers.”

For more info on the new Linux digs from Cray, read their full release here.

Related posts:

• Cray poised to grow market through the magic of ISV software
• Fixstars Releases Yellow Dog Linux 6.2
• PGI Compilers Now Available on Cray CX Line

The post Cray Releases New Version of Cray Linux Environment appeared first on insideHPC.

Groundbreaking advancements in both science and industry come about through the collaboration of scientific expertise and the most innovative and highly scalable technology,” said Peter Ungaro, president and CEO of Cray. “ASCS is bringing together industry experts and top scientists with the aim of developing products and energy-efficient technologies as well as enhancing product development and optimization practices for the automobile industry throughout the world. We are thrilled to be part of this project and look forward to seeing and benefiting from the innovations that result.”

ASCS members include the High Performance Computing Center Stuttgart at the University of Stuttgart, Porsche, Daimler, Opel, ABAQUS, Altair Engineering, DYNAmore, Engineous Software and INTES as well as several others.

Read the full article here.

Related posts:

• Day 3 Keynote from GTC: Behind the Science in Automotive Design
• Cray launches new XT5m midrange line
• Cray Appoints New Senior VP of Sales and Marketing

The post Cray Joins European Automotive Simulation Center appeared first on insideHPC.

Despite the theoretical advantage of IB’s lower latency and higher throughput, Woven’s HPL test shows nearly identical performance for the 10 GbE solution with their switch and IB in both performance and efficiency out to 36 nodes (288 processors). The company is cautious about being too optimistic based on these results, though

Clearly not every application will perform as well on 10 GE as on InfiniBand, but it should be pointed out that the majority of applications in the HPC space do indeed have a payload profile of moderate packet sizes that is not unfavorable to 10 GE, including industry standard applications such as LSDyna, Fluent, Abaqus, and others. A few applications, such as Amber and Gamess, operate in a communications regime dominated by small packets, and for these, InfiniBand may be a better choice.

Woven goes on to assert that the ubiquity of Ethernet and broad experience with the technology makes it the technology of choice as long as there is no performance penalty

When the multiuse capability of 10 GE for storage and system backbone functions are considered, along with the inherent ease of design, construction and maintenance of a cluster, it should be clear to any system architect that they should be using 10 GE as the standard scenario. 10 GE is a safe and natural starting point for new cluster designs, and moving away from it is indicated only when there is a specific and compelling need.

I don’t actually design systems, but I’m willing to give them the benefit of the doubt on that argument. I suppose what they needed to do with this paper is to show that there isn’t a performance disadvantage to using 10 GbE, which is the first step in getting into the market.

Sadly, “hey, we’re not worse than the other guys” isn’t a great marketing strategy, so they can’t stop here.

They actually have what is to my mind a much better story to tell here about their adaptive routing capabilities which, as demonstrated in the earlier Sandia tests, show that their Ethernet switch technology can offer real advantages over statically routed fabrics.

The test in this most recent report were too small to stress the network enough for the adaptive component to matter. I’d really like to see both tests run at real scale; thousands of processors would be ideal. If the results they’ve demonstrated in the small translate to large-scale systems, Woven’s technology should find its way into a whole bunch of systems.

Related posts:

• Chelsio and Woven demonstrate 10GbE RDMA solution at Sandia
• Good news for Woven
• New Whitepaper Compares HPC App Performance on 10GbE vs IB

The post Woven compares 10 GbE and DDR Infiniband using HPL appeared first on insideHPC.