HPC Wire

Subscribe to HPC Wire feed
Since 1987 - Covering the Fastest Computers in the World and the People Who Run Them
Updated: 16 hours 30 min ago

Tsinghua Crowned Eight-Time Student Cluster Champions at ISC

Thu, 06/22/2017 - 13:10

Always a hard-fought competition, the Student Cluster Competition awards were announced Wednesday, June 21, at the ISC High Performance Conference 2017. Amid whoops and hollers from the crowd, Thomas Sterling presented the award to the team with the best overall score, Tsinghua University. Sponsored by Inspur and using Nvidia graphics processors, Tsinghua was also one of the three teams who had a perfect score for the deep learning part of the competition.

The team is a force to be reckoned with. They are the only team to achieve a “triple crown” victory by winning competitions at ISC, SC, and ASC. They’ve won eight championships in total and this is their third win at the ISC competition.

Rounding out the list for the overall winners in this “very close competition” was 2nd place for Centre for High Performance Computing (CHPC) and 3rd place for Beihang University.

The award for Fan Favorite, the second year in a row, went to Universitat Politècnica De Catalunya Barcelona Tech (UPC) who captivated ISC attendees and garnered the most votes. Over 2,100 people voted for their Fan Favorite — a record for this conference attendee participation portion of the competition.

2nd place CHPC team with CHPC Director Dr. Happy Sithole and friends

The award for the Highest High Performance Linpack went to FAU boyzz from Friedrich-Alexander University Erlangen–Nürnberg. They are one of the few teams that has competed worldwide at all three competitions — the Supercomputing Conference (SC), Asia Student Supercomputer Challenge (ASC) in China, and ISC. They used a traditional cluster with 12 GPUs.

Onto the award for running deep learning applications, Vice General Manager XuJun Fu of Baidu Cloud, who brought the deep learning applications to the competition, announced the winners. Tsinghua University, Nanyang Technological University and Beihang University all took home the top prize for solving the Captcha Challenge, achieving the highest degree of model accuracy.

This year, ten teams from around the world came to Frankfurt to build a small cluster of their own design and test their HPC skills by optimizing and running a series of benchmarks and applications. The teams must keep their power consumption below 3000 watts on one power circuit while running the benchmarks and applications.

The teams used a variety of designs. Two teams utilized liquid cooling technology, eight teams used GPUs and one team used Xeon Phi. UPC built an ARM based cluster with 48 core chips, liquid cooled. EPCC University of Edinburgh (EPCC) were described as the Linpack junkies, driving their results with a liquid cooled system.

Gilad Shainer and Thomas Sterling reveal Tsinghua University as the team with the highest overall score

Scot Schultz, director of Educational Outreach, HPC Advisory Council, provided some additional perspective for this article. “We decided to use a slightly more interesting use case of solving for Captcha,” he said, “because it not only highlights the power of deep learning to be a useful tool create models to recognize and classify unwieldy data, such as distorted characters, grainy images and overlapping characters, but it also demonstrates that it is possible for this powerful technology to be used in less positive ways, such as solving for security or privacy.  Realizing that everyone has access to the tools we use to move society forward, we need to be aware of the possible mis-use, especially as it becomes more pervasive across industry, healthcare, financial services, and the like.”

“We’re always amazed and impressed by the quality of work and solutions we see from students during the ISC Student Cluster Competition,” said Doug Miles, director of PGI Compilers & Tools at Nvidia, a key sponsor for the student teams, many of whom rely on Nvidia GPUs to drive their workloads. “We support this event to encourage students from around the world to dive in and learn how to program parallel systems using the same compilers and programming models used by HPC professionals every day on the world’s fastest supercomputers.”

Gilad Shiner, chairman of HPC Advisory Council, said, “The HPC Advisory Council is proud to host, together with ISC, the 6th student cluster competition at the ISC High Performance conference. It enables the next generation of HPC professionals and drives HPC technologies to be used in more areas and applications.

“We want to thank all of the university teams that participated in the competition. Through this competition they have gained knowledge and expertise in HPC, deep learning technologies, and solutions. We hope to see these teams as well as new teams at the ISC 2018 competition.”

The post Tsinghua Crowned Eight-Time Student Cluster Champions at ISC appeared first on HPCwire.

GPUs, Power9, Figure Prominently in IBM’s Bet on Weather Forecasting

Thu, 06/22/2017 - 12:19

IBM jumped into the weather forecasting business roughly a year and a half ago by purchasing The Weather Company. This week at ISC 2017, Big Blue rolled out plans to push deeper into climate science and develop more granular forecasting models and to port those models onto the Power platform including forthcoming Power9-based systems. Power9 is IBM’s next generation Power processor and will be used in the CORAL supercomputer project.

A key aspect of the new effort is leveraging GPUs on the Power platform. “If you know the weather space, the utilization of GPUs has been pretty limited. There’s the COSMO (Consortium for Small Scale Modeling) model used in Switzerland,” said Dave Turek, IBM vice president of exascale system. While not completely new, use of GPUs has been limited in weather forecasting (see HPCwire articles, Swiss Weather Forecasting Achieves 1.1km Resolution on ‘Piz Kesch’ and Today’s Outlook: GPU-accelerated Weather Forecasting).

GPUs have shown promise helping to tame one of weather forecasting’s thorniest problems which is being able to scale the models to both small areas and large areas and integrating those into forecast spanning local, regional, and global weather. Turek says GPUs will become increasingly important for achieving the necessary near real-time, multi-scale simulations required for accurate forecasting.

Dave Turek, IBM

“The early indications are we should see significant performance gains and it is really going to be an enabling technology, if you will, to let us go down to that level of scale,” he said. “You should see some results [from this effort] around the end of the year.”

The collaboration announced at ISC includes IBM, The Weather Company, University Corporation for Atmospheric Research (UCAR) and the National Center for Atmospheric Research (NCAR), a federally funded research and development center for the atmosphere and Earth’s geospace systems.

Specific work will be around further developing MPAS – the Model for Prediction Across Scales (MPAS) – and porting it to the IBM system. MPAS itself is a collaborative project for developing atmosphere, ocean and other earth-system simulation components for use in climate, regional climate and weather studies. To date, the primary development partners working on MPAS have been the climate modeling group at Los Alamos National Laboratory (COSIM) and the NCAR.

As noted, the basic idea is to be able drill down on small scale events. Turek noted today’s operational global forecast models predict weather patterns down to regional-scale weather events, such as snowstorms and hurricanes. The new model could improve weather and climate forecasting by better accounting for the small-scale phenomena, such as thunderstorms, that can impact the weather.

Here’s a bit more detail on MPAS from GitHub: “The defining features of MPAS are unstructured Voronoi meshes and C-grid discretization used as the basis for many of the model components. The unstructuctured Voronoi meshes, formally Spherical Centroidal Voronoi Tesselations (SCVTs), allow for both quasi-uniform discretization of the sphere and local refinement. The C-grid discretization, where the normal component of velocity on cell edges is prognosed, is especially well-suited for higher-resolution, mesoscale atmosphere and oceanic simulations. The land ice model takes advantage of the SCVT-dual mesh, which is a triangular Deluanay tessellation appropriate for use with finite element-based discretization.”

A key goal is to adapt MPAS to run more efficiently on next-generation computers. While regional models have been run at scales that predict thunderstorms for over a decade, the enormity of global models have made this challenging on a global scale. Enabling “convection-allowing models” on a global scale will not only enable short-term thunderstorm forecasts, but also lead to more accurate long-range forecasts days, weeks and months in advance, according to IBM.

The UCAR/NCAR teams involved include the Computational Information System Laboratory and the Mesoscale & Microscale Meteorology Laboratory.

“This is a major public-private collaboration that aims to advance weather prediction and generate significant benefits for businesses making critical decisions based on weather forecasts,” said UCAR President Antonio Busalacchi in the official release. “We are gratified that taxpayer investments in the development of weather models are now helping U.S. industries compete in the global marketplace.”

Currently, Power9 silicon is not available although Turek said IBM is doing some early testing with “partners” not customers. Very likely that includes GPU specialist Nvidia, which along with IBM, is an OpenPOWER founder. The Power9 chip, said Turek, will be the first in the line to have cache coherency such that the GPU will be seen as a coprocessor which should make programming easier. Power9, of course, is intended to support several types of accelerators using different optimized interconnect schemes.

The post GPUs, Power9, Figure Prominently in IBM’s Bet on Weather Forecasting appeared first on HPCwire.

Intersect 360 at ISC: HPC Industry at $44B by 2021

Thu, 06/22/2017 - 08:25

The care, feeding and sustained growth of the HPC industry increasingly is in the hands of the commercial market sector – in particular, it’s the hyperscale companies and their embrace of AI and deep learning – that will drive healthy and stable, if not spectacular, expansion of the market through 2021 and beyond.

Those are some of the top-line findings announced today by Addison Snell, CEO of HPC industry watcher Intersect 360 Research at the ISC conference in Frankfurt. Snell’s other key findings include:

  • Total worldwide HPC market (servers, storage, software, etc.) reached $35.6 billion in 2016, up 3.5 percent from 2015.
  • Servers were the largest component, reaching $11.5 billion.
  • While servers grew by 3.3 percent over 2015, storage systems grew by a healthy 5.0 percent – Snell said this growth was spurred by commercial users.
  • Total market is forecast to grow to $43.9 billion for 2021, for a 4.3 percent CAGR. The industrial sector is expected to deliver a CAGR of 6.5 percent, followed by government at 2 percent and a flat academia sector.
  • Government represents 26 percent of the total market.

That 3.5 percent growth in 2016 is a significant come-down from 2015’s growth of 9 percent (mostly due to expansion in the storage sector) over 2014, according to Snell. His overall take on the coming years: long-term moderate, stable growth through 2021 and beyond.

Among HPC server vendors, HPE continued its leadership in HPC server sales, claiming more than 33 percent market share over second place Dell EMC’s 26.4 percent. HPE’s lead was boosted by “a couple of points” by the incorporation of sales by SGI, which HPE acquired last year. They were followed by Lenovo, IBM and Cray. Snell noted that it was only two years ago that IBM was the top seller of HPC servers until announcing the sale of its x86 server business to Lenovo (currently with 6.9 percent of the market), which he said has underperformed based on the market share it acquired from IBM.

Source: Intersect 360

On the storage side, Dell EMC continued its leadership with a 24.4 percent share, with NetApp slipping slightly to 15.8 percent, followed by HPE, IBM and HDS. DDN was reported to have a 3 percent share and Seagate at 2.9 percent.

Snell emphasized that AI and deep learning are the drivers for HPC in the commercial sector.

“Deep learning and AI drives a lot of HPC-like investments, they’re using a lot of GPUs, Infiniband, and so forth,” he said. “It’s being driven primarily by the hyperscale companies, like Baidu, Google, Amazon, Facebook, Microsoft. Traditional HPC companies might be looking at deep learning and AI but they’re not spending a lot differently than they would have spent otherwise. We’re not seeing big changes in budgets, even among major public sector supercomputer facilities.”

He said an academic institution may have spent money last year on a new supercomputer that it plans “to do some AI things with, it’s the same supercomputer they would have bought otherwise.”

Among non-hyperscale companies, he said HPC growth among commercial companies is led by the financial services sector, followed by manufacturing, then oil and gas.

Snell said the study’s most surprising finding: lack of market uptake for HPC in public clouds. Starting from a low base last year, Intersect 360 reported growth of 6.4 percent – not the 10 percent growth that had been expected. He said the market remains selective in the jobs it offloads to public cloud platforms, drawing an analogy to transportation decisions we make in everyday life: usually it makes the most sense to drive our own cars, but sometimes taking a taxi, the train or flying is the best choice. Similarly, he said, it usually makes the most sense for organizations to use own on-prem HPC capabilities.

With public clouds holding 2.5 percent of the HPC market, Snell said Amazon has the largest share, adding that he is “reasonably sure” Microsoft Azure overtook Google last year for second place among public cloud services providers, followed by IBM Softlayer.

The post Intersect 360 at ISC: HPC Industry at $44B by 2021 appeared first on HPCwire.

K Computer Takes First Place for the Second Consecutive Time on HPCG Benchmark

Thu, 06/22/2017 - 08:15

TOKYO, June 22, 2017 — On June 19, the K computer took first place for the second consecutive time in the HPCG benchmark, a new index developed to create a more realistic view of supercomputer performance compared to the commonly used LINPACK benchmark. This success, which surpasses the second place achieved in 2014 and 2015, was made possible by subsequent improvements of the performance of the system and applications.

The HPCG (High Performance Conjugate Gradient) benchmark measures how fast a computer can solve symmetric sparse linear system equations using the conjugate gradient method preconditioned with a multi-grid symmetric Gauss-Seidel smoother. Problems of this type are typically encountered in actual engineering and industrial applications, and require a balance between calculation performance, memory performance and communication performance, unlike LINPACK, which looks at calculation speed alone.

For this result, all of the K computer’s 82,944 compute nodes were used, achieving a performance of 602 teraflops. This figure is higher than the supercomputers that placed higher than the K computer in the TOP500 rankings, demonstrating outstanding performance in various science and engineering fields.

According to Mike Heroux of Sandia National Laboratories, who developed the HPCG benchmark, “The HPCG benchmark is very demanding. A good score requires strong and versatile memory system performance, excellent interconnect network performance at scale, and an overall balanced system. The top ranking of the K computer is no surprise.”

The award will be presented at ISC High Performance 2017 in Frankfurt, Germany.


RIKEN is Japan’s flagship research institute devoted to basic and applied research. Over 2500 papers by RIKEN researchers are published every year in reputable scientific and technical journals, covering topics ranging across a broad spectrum of disciplines including physics, chemistry, biology, medical science and engineering. RIKEN’s advanced research environment and strong emphasis on interdisciplinary collaboration has earned itself an unparalleled reputation for scientific excellence in Japan and around the world. For more information, please see: http://www.riken.jp/.

About Fujitsu Ltd

Fujitsu is the leading Japanese information and communication technology (ICT) company, offering a full range of technology products, solutions, and services. Approximately 155,000 Fujitsu people support customers in more than 100 countries. We use our experience and the power of ICT to shape the future of society with our customers. Fujitsu Limited (TSE: 6702) reported consolidated revenues of 4.5 trillion yen (US$40 billion) for the fiscal year ended March 31, 2017. For more information, please see http://www.fujitsu.com.

Source: Fujitsu

The post K Computer Takes First Place for the Second Consecutive Time on HPCG Benchmark appeared first on HPCwire.

ASRock Rack Showcases Frugal Servers at ISC17

Thu, 06/22/2017 - 07:58

FRANKFURT, Germany, June 22, 2017 — ASRock Rack aims to satiate your tech fix with its latest parallel computing solutions. But its high-performance servers are anything but power-hungry. Our 3U10G will be a HPC transformer by different PCI-E usages, and our 2U4N-F/X200 is designed for a maximum 72 cores Xeon Phi CPU on each node. Furthermore, the next Open Compute Project generation has optimized space and power efficiency.
“3U10G” can literally max out the capability and parallel the workload by 10 full-sized GPU cards. However, you are welcome to interpret it as “9+1”, 1 InfiniBand card delivers all the results to the cloud. Furthermore, two Mezzanine slots give this ultimate turbo engine storage and NIC expandability. It represents the next generation of converged structure.

While Knights Landing made an entrance four years ago, ASRock Rack has now given them an imposing 4U height. These 4 Intel Xeon Phi beasts are each equipped with 72 cores, and are a natural choice for machine learning or AI tasks. Each unit reserves extra PCI-E and mezzanine slots, and comes packed with 2 NVMe SSD slots. It serves the needs of high-speed networks in HPC platform, and also accommodates InfiniBand or Intel’s Omni-Path framework.

“Cost-effectiveness” is our main goal for designing the OCP series. To achieve this, ASRock Rack uses Professional Power Usage (PUE), duo gigabit LAN and excellent I/O flexibility to minimize power waste. Twins OCP-1L & 6S are the fruits of our labour — not only capable of handling massive tasks, but also consuming up to 9% less power than existing providers. Duo Intel i210 GLAN controllers design means the customer has no need to purchase additional LAN cards and can also flexibly deploy failover and minimize downtime.

What are you waiting for? Come to visit these beasts at Booth A-1323 in Frankfurt, Germany.

About ASRock Rack

ASRock Rack Inc., established in 2013, specializes in providing high-performance and high-efficiency server technology in the fields of Cloud Computing, Enterprise IT, HPC and Datacenter. We adopted the design concept of “Creativity, Consideration, Cost-effectiveness” from ASRock, and the company devotes passion to think out-of-the-box in the Server Industry. Leveraged by ASRock’s growing momentum and distribution channels, this young and vibrant company targets the booming market of Cloud Computing, and is committed to serve the market with user-friendly, eco-friendly and do-it-yourself Server technology, featuring flexible and reliable products.

Source: ASRock Rack

The post ASRock Rack Showcases Frugal Servers at ISC17 appeared first on HPCwire.

RSC Group Presents RSC Tornado Cluster Solution at ISC17

Thu, 06/22/2017 - 07:45

FRANKFURT, Germany, June 22, 2017 — RSC Group, the leading developer and integrator of innovative solutions for high-performance computing (HPC) and data centers in Russia and CIS has demonstrated its ultra high-dense, scalable and energy-efficient RSC Tornado cluster solution with direct liquid cooling (all cabinet elements including high-speed interconnects are liquid cooled) at ISC’17 international exhibition. This RSC solution based on 72-core Intel Xeon Phi 7290 processor has established the world computing density record for x86 architecture in 1.41 Petaflops per cabinet or over 490 Teraflops/m3.

RSC has showcased a full set of components for modern HPC computing systems of different scale with 100% liquid cooling in ‘hot water’ mode, including high-performance RSC Tornado computing nodes based on 72-cores Intel Xeon Phi 7290 processor and Intel Server Board S7200AP, Intel Xeon E5-2697А v4 and Intel Server Board S2600KPR(F) with Intel SSD DC S3520 Series, Intel SSD DC P3520 Series solid state drives with NVMe interface in high-dense М.2 formats and the latest Intel Optane SSD DC P4800X Series. The next generation of RSC Tornado is ready to support the newest Intel Xeon Processor Scalable Family (code named Skylake-SP) which are expected to launch in second half of the year.

RSC Tornado solution based on Intel server processors has leading footprint and computing density (up to 153 nodes in one standard cabinet 80cmx80cmx42U), high energy efficiency and provides stable operation of computing nodes in ‘hot water’ mode with cooling agent temperature up to +65 °C at inlets of switching nodes and interconnects. Operation in ‘hot water’ mode enables all-year free cooling (24×365) using only dry coolers running at ambient air temperature up to +50 °C, and complete

elimination of freon circuit and chillers. Therefore, average PUE (power usage efficiency) of the system is less than 1.06. Cooling consumes less than 6% of total consumed power, which is an outstanding result for HPC industry.

At ISC’17, RSC specialists have also introduced the world’s first 100% ‘hot water’ liquid cooled 48-port Intel Omni-Path Edge Switch 100 Series for high-speed interconnects (with up to 100 Gbps per port non-locking switching speed). Intel Omni- Path Architecture (Intel OPA) is a complex solution for high-speed switching and data transfer improving application performance in entry-level HPC clusters and large-scale supercomputer projects with minimum expenses. 48-port Intel OPA switch enables connection of 26% more servers than competing solutions with lower budget and power consumption reduced to 60% providing more energy efficient switching and system infrastructure.

RSC BasIS integrated software stack for cluster system management

Innovative management and monitoring system based on RSC BasIS integrated software stack also provides high availability, resistance to failures and ease of use of HPC computing systems based on RSC solutions. This system is an open and easily expandable platform based on open source software and micro-agent architecture. It enables controlling full data centers and their individual elements such as computing nodes, interconnects, infrastructure components, workloads and processes. All system elements (computing nodes, power supplies, hydraulic regulation modules, etc.) consist an integrated management module providing broad capabilities for detailed telemetry and flexible management. Cabinet design supports replacement of computing nodes, power supplies and hydraulic regulation modules (with redundancy) in hot-swap mode without interruption of system operation. Most components of the system (such as computing nodes, power supplies, network and infrastructure components, etc.) are software-defined, and this significantly simplifies and speeds up initial deployment, maintenance and future upgrades of the system. Liquid cooling of all components ensures their longevity.

New functionality of RSC BasIS for monitoring and control of territorially distributed data centers was presented by RSC at ISC’17.

Unique projects at JSCC RAS and SSCC SB RAS

This year RSC Group has completed an upgrade of computing resources of the Joint Supercomputer Center of the Russian Academy of Sciences (JSCC RAS) and the Siberian Supercomputer Center of the Siberian Branch of the Russian Academy of Sciences (SSCC SB RAS) on the basis of the Institute of Computational Mathematics and Mathematical Geophysics. Both projects are unique as they are world’s first deployments that have server nodes with ‘hot water’ liquid cooling based on the most powerful 72-cores Intel Xeon Phi 7290 processors and 16-cores Intel Xeon E5- 2697А v4 processors. For the first time in Russia and CIS region, implementation of these unique projects involved deployment of communication subsystems of two cluster systems based on Intel Omni-Path Architecture.

Joint resources of JSCC RAS (Moscow) and SSCC SB RAS (Novosibirsk) will be used as the basis for territorial distributed computing facility for solving relevant tasks in fundamental and applied sciences, including advanced research in fields of AI (Artificial Intelligence), ML/DL (Machine Learning, Deep Learning), Big Data and others.

Total peak performance of these computing facilities is currently about 1.1 Petaflops. Innovative management and monitoring system based on RSC BasIS integrated software stack also provides high availability, resistance to failures and ease of use of computing resources at JSCC RAS and SSCC SB RAS.

About RSC Group

RSC Group is the leading developer and integrator of ;full cycle’ new generation solutions for high-performance computing (HPC) segment and data centers in Russia and CIS based on Intel architectures, innovative liquid cooling technologies and a number of its own know-hows. RSC has the potential to

create the most energy efficient solutions with record-breaking power usage effectiveness (PUE), the highest computing density in the industry with standard x86-based processors, to use fully “green” design, provide the highest solution reliability, noise-free operation of computing modules, 100% compatibility and guaranteed scalability with unmatched low cost of ownership and low power consumption. RSC specialists also have the experience of developing and implementing an integrated software stack of solutions to improve work efficiency and application of supercomputer systems from system software to vertically oriented platforms based on cloud computing technologies.

RSC participates in Intel Technology Provider Program at Platinum level, is member of Intel Fabric Builders Program, holds of Intel Solutions for Lustre* Reseller Elite and Intel HPC Data Center Specialist statuses. Performance and scalability of solutions based on RSC PetaStream and RSC Tornado architectures are Intel Cluster Ready certified. For more information please visit www.rscgroup.ru.

Source: RSC Group

The post RSC Group Presents RSC Tornado Cluster Solution at ISC17 appeared first on HPCwire.

At ISC – Goh on Go: Humans Can’t Scale, the Data-Centric Learning Machine Can

Thu, 06/22/2017 - 07:00

I’ve seen the future this week at ISC, it’s on display in prototype or Powerpoint form, and it’s going to dumbfound you. The future is an AI neural network designed to emulate and compete with the human brain. In this game, the brain doesn’t stand a chance.

Scoff at such talk as farfetched or far off in a hazy utopic/dystopic future. Roll your eyes and say we’ve heard the hype before (some of us remember a supercomputer company 25 years ago with the inflated name of “Thinking Machines,” long defunct). But it’s neither futuristic nor hype, it’s happening now, the technology pieces are taking shape, and the implications for business, for the work world and for our everyday lives – for good or ill – are as staggering as they are real.

Aside: It’s somewhat unsettling that conference attendees here in Frankfurt don’t seem particularly interested in those implications. For the moment, ISC is at the gathering point computer scientists bringing about massive technological change, but nearly all the talk here is about the “how” of AI systems, not the “what then?” But there’s one anecdote making the rounds that has raised eyebrows: when Google engineers were asked to how its AlphaGo machine the winning move against the world champion of Go (the world’s most complex board game), the answer was: “We don’t know” (more on this below).

Quite consciously, engineers are architecting HPC systems along the lines of our brain. The new architecture is an emerging style of computing called “data intensive” or “data centric.” It replaces processing with memory (i.e., data) at the center of the computing universe. Combined with advanced algorithms, new memory and processor technologies are coming on line to make the new architecture a practical reality. Once the pieces are in place, the next step will be to scale these systems beyond all measure of human brain capacity.

What does data centric computing mean? How does it work? Why does it represent a major shift in advanced scale computing?

Let’s start answering those questions by first looking at how data centric systems are measured. The benchmark for new AI systems isn’t how fast they solve linear algebra problems (i.e., Linpack). That’s how processor-centric systems have been measured for decades, and considering the capabilities of data-centric systems under development, that benchmark seems wholly inadequate.

Rather than throughput, AI-based systems are measured in relation to people: their ability to compete with humans at our most intellectually challenging games of reason – checkers, chess, Go, poker. The standard of success isn’t training the system to become at it, or to “solve” the game (i.e., work out every possible combination of moves). The benchmark is playing the game better than any human.

That’s the objective. Once the system is better than any of us, it’s ready to move into an advisory role, providing guidance and suggestions, augmenting our capabilities. For now. In a decade or so, these systems will take over tasks for us altogether.

Driving is a prime example. If driving were a game, humans would still beat machines – even though statistics show we’re getting worse at it (according to Dr. Pradeep Dubey, Intel Fellow, Intel Labs & Director, Parallel Computing Lab, who presented at ISC on autonomous vehicle technology). Around the world, two people are killed in car accidents each minute. In the U.S., 40,000 people are killed annually and 2 million suffer permanent injuries.

Meanwhile, AI is enabling machines to get better at driving. A convergence point is coming. For now, the car’s intelligence is limited to navigating, warning us about traffic conditions and setting off beepers when we get close to curbs and other cars.

The next step: our roads will have special lanes where we’ll temporarily hand over operation of the car to itself. A few years after that, we won’t drive at all. Driving is a game in which machines will soon be much better than we are.

Dr. Eng Lim Goh, Vice President of HPE and an industry visionary for decades, is a prime driver of new AI system development. At ISC this week, he discussed why AI in all its forms – machine learning, deep learning, strategic reasoning, etc. – is the driving force bringing about “data intensive” computing architectures.

Here’s his schema for the data intensive computer:

The left side of the diagram is old-style, LINPACK-benchmarked, processor-centric computing. That’s where HPC happens. The processor is at the center. Data is sent to the CPU, simulations are run, and new, and more, data comes out. These systems have hit a wall of their own making. The problem occurs when HPC systems run their simulations, generating exponentially more machine-generated data than they started with. They’re producing data beyond the capability of data scientists to analyze. Big data isn’t big enough.

“For 30 years we’ve lived in this world where small amounts of data go in, and we apply supercomputing power onto our partial differential equations, or our models, to generate lots of data,” he said.

Already, Goh pointed out, there aren’t enough data scientists to meet demand for today’s data analytics requirements. For the torrents of machine-generated data to come, there’s an overwhelming need to automate how data is analyzed.

Take for example seismic exploration.

For exploration of energy reserves at sea, ships drag cables with hydrophones, fire shots into the ocean floor and collect the echo on sensors. Goh said for every 10TB of data collected by the sensors, 1PB of simulation data is produced – 100X the original data.

That’s where the right side of the diagram comes in: high performance analytics (HPA), self-learning AI systems that can take voluminous amounts of data produced by HPC, put it in memory, and work up answers to questions.

Dr. Em Ling Goh

The key to the data-centric system of the future is the border area in the middle of the diagram. That’s where memory (i.e., data) resides, like a queen bee. It will be surrounded by a variety of processors (CPUs, GPUs, FPGAs, ASICs, assigned jobs appropriate for their capabilities) operating around the data, like drones.

Looked at this way, in a world where most companies have analyzed only about 3 percent of their data on average, traditional HPC systems seem glaringly incomplete. Combining the left side of the diagram and the right, integrating HPC with HPA – that takes supercomputing somewhere new. That’s a machine with a new soul.

But Goh conceded there are barriers to HPC and HPA joining forces.

“The two worlds are very different,” Goh said. “The HPC world where I lived, I’m guilty of this. All these years we assumed data movement was free. Guess what? When Linpack started 20 years ago we didn’t consider data movement. Yet we’re still ranking our Top500 systems that way. We’re still guilty that way.

“But the data scientists of the world also have something to say about us,” he added. “They assume compute is free. Take Hadoop. Hadoop is a technique where you map your data out onto compute nodes, then do your computation, then you reduce the data you bring back. The data world called this MapReduce. So we have to bring the two worlds together. More and more now, people should be investing in one system of left and right, not just the left.”

Goh pointed to the middle of his diagram and said that’s where the big architectural challenge lies. “If you have to move an exabyte of data between system A and B, if they are two different systems, it will be impractical. The world will come to this (integration of HPC and HPA).”

That’s why the U.S. effort to develop a “capable” exascale computer by the early 2020’s puts as much emphasis on compute power as memory capacity. A mission document issued by the Exascale Computing Project said its intent to build a system not just with an exaflop of processing power but one that also can handle an Exabyte of data in memory.

Goh described HPE’s “Bridges” system at the Pittsburgh Supercomputer Center as a data-centric supercomputer that incorporates HPC and HPA, designed specifically for “scalable deep learning.”

“Essentially, it’s a bandwidth machine,” Goh said. “It’s a supercomputer, but really it’s a data mover. Not only are NVlinks all connected, they’re also GPU-connected, so clumps of four GPUs can talk to other clumps of four GPUs directly. Then we have four OPA’s coming out of each node, giving one OPA per GPU. So this is really a data machine.”

The Bridges supercomputer pulled off one of the most impressive game wins of the emerging AI era when it defeated four of the world’s top poker players earlier this year. Actually, the competition stretched across two years, Goh said, with the AI system losing $700,000 to the players the first year they played. The second year, with 10X more compute from the Bridges computer, the AI system (“Libratus”) took the four humans for $1.7 million, a classic hustle.

While IBM Deep Blue (chess) and Google’s AlphaGo have grabbed most of the machine-defeated-human headlines of late, it’s less well known that machines have beaten humans at checkers, which has 1020 “naïve” (or possible) combinations, since the early 1990s, several years before IBM beat the world’s top chess player. Chess has 1047 naïve combinations. How big is 1047? An exascale machine running for 100 years would complete only 1028 combinations. The point being that without integrated AI techniques, processing only gets you so far.

Go, meanwhile, has 10171 combinations. Poker, with “only” 10160 combinations, offers up the added complexity of “incomplete information.” By contrast with the three board games, in which you can see the pieces held by your opponent, in poker, you don’t know what your opponents have in their hands.

“So we didn’t solve chess, machines didn’t solve chess,” Goh said, “all they did was be good enough to be superhuman – to beat any human. That’s a term were going to hear more and more now.”

After Goh’s presentation, he was asked to response to Google not understanding how AlpaGo won the Go tournament. The issue, he said, is overcoming opacity.

“We’re working very hard to increasing transparency,” he said. “Some people have discussed the idea that there are many stages in a neural network, to intercept it in between those stages, and take its output and see if you can make sense of it.”

Leaving a strong role for human supervision also is important. He pointed out that since the Industrial Revolution, workers get promoted from first operating a machine to supervising machines.

He also discussed the distinction between the “correct” and the “right” answer. An AI-based system may deliver a correct answer, but whether it’s “right” – acceptable within human social mores, the bounds of business ethics, or even an aesthetic judgment – is something only humans can decide.

“Societal values need to be applied, human values need to be applied,” he said.


The post At ISC – Goh on Go: Humans Can’t Scale, the Data-Centric Learning Machine Can appeared first on HPCwire.

Inspur “Champion Server” helps Tsinghua University achieve its 8th championship in international supercomputing competitions

Thu, 06/22/2017 - 01:01

FRANKFURT, Germany , June 21, 2017–The 2017 International Supercomputing Conference (ISC17) came to its end successfully. Tsinghua University defeated 12 teams , and won the championship  of this year’s ISC Student Cluster Competition. It is the third-time that Tsinghua University won the ISC-SCC (Student Cluster Competition ) championship, and its eighth championship at ISC, ASC and SC, aka the three biggest international college student supercomputing competitions. In the mean time, Inspur server, known as “champion server”, Inspur has help China’s college student supercomputing teams to win the world championship for 13 times.

Tsinghua University won the championship of ISC2017

A total of 12 colleges/universities around the world have participated in this year’s ISC-SCC. The Competition requirement is total power of no more than 3000W. The supercomputing system should run for specified 48 hours consecutively, and finish six tests, including the high performance computing international benchmark HPL and HPCG, the partial differential equation solver FEniCS, the application of modeling material MiniDFT, the artificial intelligence learning system TensorFlow and the mysterious application announced at scene.

the Tsinghua University team

Jidong Zhai, advisor of the Tsinghua University team was thrilled when informed of the championship. He believes that the team will continue to achieve excellent results through joint efforts. The team is serious about this competition and is fully prepared, which is the internal effort; while Inspur, as the sponsor, provides efficient and reliable servers, which is served as an important external effort.

Inspur NF5280M4

Tsinghua University applied eight-computer-eight-card system design in the competition, and used Inspur NF5280M4 as host computers of the competition, withstanding all harsh tests of the finals. Its stability, reliability and excellent performance enable the team to focus on the competition itself. NF5280 is the classic server series of Inspur, and has accompanied China’s college student supercomputing teams for 8 years when attending the three biggest supercomputing competitions globally. After several technical iterations, its excellent quality has helped to win the “Champion Server” title. NF5280M4 uses Intel Xeon E5V4 series processors, supports 24 DIMMs, with DDR4 memory as well as the latest disk controller technology. In terms of power consumption, it uses the design of low power consumption components with high efficiency power supply, which is of the industry’s highest conversion efficiency, saving up to 20% energy. Meanwhile, the conversion efficiency of its option of titanium high-energy-efficiency hot-swappable redundant power supply is over 96%.

At the same time, BEIHANG University supported by Inspur won the third place in this year’s ISC Student Cluster Competition.

The post Inspur “Champion Server” helps Tsinghua University achieve its 8th championship in international supercomputing competitions appeared first on HPCwire.

ACM and IEEE Honor the Late Chuck Thacker

Wed, 06/21/2017 - 12:07

NEW YORK, N.Y. & LOS ALAMITOS, Calif., June 21, 2017 – ACM, the Association for Computing Machinery, and IEEE Computer Society have jointly announced that the late Charles P. “Chuck” Thacker is the recipient of the 2017 Eckert-Mauchly Award. Thacker was selected for the award in mid-May, but passed away on June 12 before the public announcement was made. The award committee cited Thacker’s fundamental contributions to networking and distributed computing, including Ethernet, the Xerox Alto, and development of the first tablet computers. Often hailed as an “engineer’s engineer,” Thacker made contributions across the full breadth of computer development, from analog circuit and power supply design to logic design, processor and network architecture, system software, languages, and applications.

In 1970, Xerox opened its Palo Alto Research Center (PARC) and hired several leading computer scientists and engineers, including Thacker. Early on, the staff at Xerox PARC was using a time-sharing approach in which various terminals were connected to a single computer. Because time sharing was a slow and cumbersome process, leaders at Xerox PARC conceived the idea of developing personal computers as part of a network that would be used for communication as well as computation.

Mainframe computers in the early 1970s were so large that they took up whole rooms, and their expense made them relatively scarce. Under the paradigm at the time, computer architecture needed to be either scaled up (more hardware) for better performance, or scaled down (less hardware) for lower cost. Thacker realized that a personal computer would need to be designed differently from a standard computer to address space constraints, maintain strong performance, and be inexpensive if it was to become pervasive.

At the same time, the idea of a “personal” computer that would be geared more toward human-paced activities called for the engineers to prioritize input/output (I/O) functions rather than application functions, as had traditionally been the case.

The new design feature Thacker employed as the Lead Engineer in what would become the Xerox Alto Computer was a central processing unit (CPU) microprocessor that used microcode for most of the computer’s I/O functions, rather than hardware. The microcode controlled various tasks, including executing the normal instruction set, memory refresh, and network and display functions. The Xerox Alto was therefore not simply a mini-version of existing computers, but had a novel architecture that allowed it to deploy new kinds of software.

Today the Alto is recognized as being the first modern personal computer. The initial architecture of the Alto gave rise to other important inventions developed by engineers at Xerox PARC including WYSIWIG (What You See Is What You Get) editing, laser printing, drawing and painting, email, mouse-driven graphical user interfaces, and many other features that are commonplace in personal computers today.

Another critical innovation of Thacker’s that was an outgrowth of his work on the Alto was the development of hardware for Bob Metcalfe’s invention of the Ethernet Local Area Network (LAN), which facilitated communication among computers.

Twenty years after the development of the Xerox Alto, Thacker made another foundational contribution to personal computing with the development of the Lectrice, a laboratory prototype for today’s portable PCs. He went on to develop a prototype upon which Microsoft Tablet PC software was developed, as well as a system for reading electronic books that laid the groundwork for many of today’s e-readers. One of Thacker’s most recent contributions is the design of AN3, a low-cost, efficient circuit-switched data center network.

Most recently, Thacker was a Technical Fellow at Microsoft Research, Palo Alto. He held 29 patents in areas including computer architecture, displays, networks, switches, synchronization, and encryption. During his career, he received the ACM A.M. Turing Award, the IEEE John von Neumann Medal, the ACM Software System Award (together with Butler Lampson and Robert Taylor), and the Charles Stark Draper Prize (together with Alan Kay, Butler Lampson and Robert Taylor), among many other honors. Thacker received a BS in Physics from the University of California, Berkeley.

The Eckert-Mauchly Award is known as the computer architecture community’s most prestigious award. The designation of Thacker as the posthumous recipient will be formally announced at the ACM/IEEE International Symposium on Computer Architecture (ISCA) to be held June 24-28 in Toronto, Canada.

ACM and IEEE Computer Society co-sponsor the Eckert-Mauchly Award, which was initiated in 1979. It recognizes contributions to computer and digital systems architecture and comes with a $5,000 prize. The award was named for John Presper Eckert and John William Mauchly, who collaborated on the design and construction of the Electronic Numerical Integrator and Computer (ENIAC), the pioneering large-scale electronic computing machine, which was completed in 1947.

About ACM

ACM, the Association for Computing Machinery www.acm.org, is the world’s largest educational and scientific computing society, uniting computing educators, researchers and professionals to inspire dialogue, share resources and address the field’s challenges. ACM strengthens the computing profession’s collective voice through strong leadership, promotion of the highest standards, and recognition of technical excellence. ACM supports the professional growth of its members by providing opportunities for life-long learning, career development, and professional networking.

About IEEE Computer Society

IEEE Computer Society, https://www.computer.org/, is one of the world’s leading computing membership organizations and a trusted information and career-development source for a global workforce of technology leaders including: professors, researchers, software engineers, IT professionals, employers, and students. IEEE Computer Society provides high-quality, state-of-the-art information on an on-demand basis. The Computer Society provides a wide range of forums for top minds to come together, including technical conferences, publications, a comprehensive digital library, unique training webinars, and professional training. IEEE is the world’s largest professional association for advancement of technology and the Computer Society is the largest society within IEEE.

Source: ACM

The post ACM and IEEE Honor the Late Chuck Thacker appeared first on HPCwire.

Hewlett Packard Enterprise Announces Promotion of Antonio Neri to President

Wed, 06/21/2017 - 10:59

PALO ALTO, Calif., June 21, 2017 — Hewlett Packard Enterprise (NYSE:HPE) (HPE) today announced that it has promoted Antonio Neri to President effective immediately. Previously, Neri served as Executive Vice President and General Manager, Enterprise Group, responsible for the development and delivery of the IT infrastructure solutions at the core of the world’s largest and fastest growing organizations, which includes servers, storage, networking, technology services, converged data center infrastructure, Telco and cloud solutions.

In addition to leading the company’s four primary lines of business, in this new role Neri will oversee the company’s efforts to streamline and optimize the go-forward company following the completion of the spin-mergers of its Enterprise Services and Software divisions. These efforts, which the company calls HPE Next, are focused on driving growth and profitability, and ensuring the company will be well positioned to win in the markets where it competes.

“Antonio is a veteran technology executive who has led some of the most important businesses and initiatives at HP during his 22 years with the company. From overseeing the reinvention of our Technology Services business and the divestiture of our H3C business in China, to the integration of critical acquisitions like Aruba, SGI, SimpliVity and Nimble, Antonio has been invaluable to me as we have worked to establish the new Hewlett Packard Enterprise,” said Meg Whitman, chief executive officer of Hewlett Packard Enterprise. “His promotion reflects the importance of his contributions to the company as leader of the Enterprise Group, which represents more than 80 percent of the go-forward company’s revenue.”

Prior to HP’s separation into two companies—Hewlett Packard Enterprise and HP Inc.—Antonio served as Senior Vice President and General Manager for HP Servers and HP Networking business units, responsible for setting the R&D agenda, bringing innovations to market, and go-to-market strategy execution. Prior to that role, he led the HP Technology Services business unit providing support and consulting services for HP’s Enterprise products and solutions. Antonio joined HP in 1995, as a customer service engineer in HP’s EMEA call center.

Antonio holds a Computer Engineering degree from Universidad Tecnológica Nacional in Argentina and is a Professor of Arts and Drawing.

About Hewlett Packard Enterprise

HPE is an industry-leading technology company that enables customers to go further, faster. With the industry’s most comprehensive portfolio, spanning the cloud to the data center to workplace applications, our technology and services help customers around the world make IT more efficient, more productive and more secure.

Source: HPE

The post Hewlett Packard Enterprise Announces Promotion of Antonio Neri to President appeared first on HPCwire.

Supermicro Announces Server Portfolio for AMD EPYC Processors

Wed, 06/21/2017 - 08:22

SAN JOSE, Calif., June 20, 2017 — Super Micro Computer, Inc. (NASDAQ: SMCI), a global leader in compute, storage and networking technologies including green computing, announces new dual socket servers and upcoming single-socket solutions supporting AMD EPYC, the latest AMD family of high-performance processors.

Supermicro announces availability of 1U and 2U Ultra dual-socket A+ Servers, architected to deliver unrivaled performance, flexibility, scalability, and serviceability required for demanding Enterprise mission critical environments and workloads. The proven Ultra design supports the highest performance AMD EPYC processors, high capacity memory, 25G/10G/1G Ethernet networking, optimized PCI-E expansion options, diverse storage alternatives including 24/10 All-Flash NVMe in 2U/1U, 2.5″ and 3.5″ hot-swap drive bay selections, GPU support, and highest efficiency Titanium Level power supplies.

Supermicro is positioning a flexible, dual-processor Tower / 4U System supporting a feature-rich EATX motherboard, and single-socket Mainstream rackmount server systems that provide cost optimization and scalable I/O choices. With high performance, lower power consumption, and reduced CAPEX, customers can match these systems to their workloads without compromise.

The Supermicro Twin products include a 2U 4-node dual-socket platform that delivers the highest data center efficiency across diverse applications and workloads, supporting up to 32-cores on AMD EPYC processors, maximum 16 DIMM slots, high I/O capacity, with a broad selection of storage options including 24 All-Flash NVMe drives. This Twin system offers optimized thermal capacity for maximum performance, efficiency and free air cooling.

This full Supermicro product portfolio aims to revolutionize the dual-socket server market while simultaneously reshaping expectations for single-socket servers with its support for AMD EPYC processors providing outstanding core density and performance, superior memory bandwidth, and unparalleled support for high-speed input/output (I/O) channels in a single processor. Supermicro servers supporting AMD EPYC will radically lower data center TCO through an optimized balance of compute, memory, I/O, and storage resources.

“Supermicro’s product line is unrivaled in the industry with the strongest optimized motherboard and server configurations for Data Centers, HPC and Cloud Computing,” said Don Clegg, Vice President of Marketing and Business Development at Supermicro. “Boosting performance per watt and per dollar, our broad selection of AMD EPYC-based solutions provides customers with a new range of powerful and energy efficient server options that can lower their overall TCO.”

“The new AMD EPYC processors were designed to deliver optimized performance per watt, and to address the complex data center and cloud computing challenges customers face today,” said Scott Aylor, corporate vice president, enterprise solutions, AMD. “With Supermicro’s strong suite of server solutions shipping with new AMD EPYC processors, customers can deploy cost-optimized solutions tailored for their specific applications and take advantage of the combination of more cores, more memory bandwidth, and more I/O that EPYC delivers.”

New Products:

  • 1U Ultra A+ Server (AS -1123US-TR4 / TN10RT) is a dual-socket platform with high capacity DDR4-2666MHz memory, 3 PCI-E 3.0 slots (2 FHHL x16 slots and 1 LP x8 slot), dual 10GbE (-TN10RT) or quad GbE (-TR4) networking options, that supports 10 hot-swappable 2.5″ NVMe/SATA3 drives in a 1U form factor. With a total bandwidth of 40 PCI-E 3.0 lanes from two processors to provide direct PCI-E 3.0 x4 connections to each NVMe SSD, the system provides up to 12x performance and 7x latency improvements over standard SATA SSD based servers and is equipped with redundant 1000W Titanium Level (96% efficiency) power supplies.
  • 1U Ultra A+ Server (AS -1023US-TR4) is a dual-socket platform with high capacity DDR4-2666MHz memory, 3 PCI-E 3.0 slots (2 FHHL x16 slots and 1 LP x8 slot), quad GbE networking, that supports 4 hot-swappable 3.5″ SSD/HDD drives (optional 4 U.2 NVMe ports supported) in a 1U form factor. The system is equipped with redundant 1000W Titanium Level (96% efficiency) power supplies.
  • 2U Ultra A+ Server (AS -2023US-TR4) is a dual-socket platform with high capacity DDR4-2666MHz memory, 7 PCI-E 3.0 slots (2 FHFL x16 slots, 4 FHHL x8 slots, and 1 LP x8 slot), quad GbE networking, that supports 12 hot-swappable 3.5″ SSD/HDD drives (optional 4 U.2 NVMe ports supported) in a 2U form factor. The system is equipped with redundant 1600W Titanium Level (96% efficiency) power supplies.
  • 2U Ultra A+ Server (AS -2123US-TN24R25) is a dual-socket platform with high capacity DDR4-2666MHz memory, 7 PCI-E 3.0 slots (2 FHFL x16 slots, 4 FHHL x8 slots, and 1 LP x8 slot), dual 25 GbE networking, that supports 24 hot-swappable 2.5″ SSD/HDD drives in a 2U form factor. The system is equipped with redundant 1600W Titanium Level (96% efficiency) power supplies.

For more information on AMD EPYC based platforms, download the latest A+ Server Brochure and visit our website at www.supermicro.com/epyc.

For complete information on A+ Server solutions from Supermicro, visit www.supermicro.com.

Follow Supermicro on Facebook and Twitter to receive the latest news and announcements.

About Super Micro Computer, Inc. (NASDAQ: SMCI)

Supermicro (NASDAQ: SMCI), the leading innovator in high-performance, high-efficiency server technology is a premier provider of advanced server Building Block Solutions for Data Center, Cloud Computing, Enterprise IT, Hadoop/Big Data, HPC and Embedded Systems worldwide. Supermicro is committed to protecting the environment through its “We Keep IT Green” initiative and provides customers with the most energy-efficient, environmentally-friendly, solutions available on the market. For more information, please visit www.supermicro.com.

Source: Supermicro

The post Supermicro Announces Server Portfolio for AMD EPYC Processors appeared first on HPCwire.

Mellanox InfiniBand Chosen by AMD for EPYC Datacenter Platforms

Wed, 06/21/2017 - 08:12

SUNNYVALE, Calif. & YOKNEAM, Israel, June 21, 2017 — Mellanox Technologies, Ltd. (NASDAQ:MLNX), a supplier of high-performance, end-to-end interconnect solutions for data center servers and storage systems, announced today that the company’s Ethernet and InfiniBand interconnect solutions have been chosen to accelerate the new AMD EPYC data center platforms. Mellanox 25, 50 and 100G Ethernet and EDR InfiniBand represent the ideal networking solutions to connect AMD EPYC CPUs, delivering the highest return on investment for Cloud, Web2.0, Big Data, Machine Learning, storage and high-performance computing infrastructures.

“We have chosen to partner with Mellanox Ethernet and InfiniBand solutions as they can unleash the full potential of our EPYC 7000-series processors and platforms,” said Scott Aylor, CVP and GM at AMD Enterprise Solutions. “The combination of EPYC, with its leading I/O capability, coupled with Mellanox interconnect technology has enabled our partners and users to maximize the new capabilities of our platform to achieve highest performance and efficiency on modern workloads.”

AMD EPYC data center CPUs deliver unmatched data throughout options and when connected with Mellanox intelligent and high speed interconnect solutions, the combined platform delivers world-leading performance for broad set of applications. As demonstrated by AMD at the EPYC announcement event, this fully integrated solution brings together best in breed technologies and was tested for the industry’s most reliable out of box experience.

“The combination of Mellanox intelligent interconnect solutions and the innovative EPYC processors from AMD will enable our mutual customers to increase their data center performance and optimize their total cost of ownership,” said Michael Kagan, Chief Technology Officer at Mellanox Technologies. “The exponential growth in data, and the need to analyze it faster and in real time, requires higher data throughputs and intelligent data processing, which are the main advantages of the EPYC platforms incorporating the Mellanox interconnect technology. We look forward to continuing to collaborate with AMD and OEM partners to offer industry leading data center compute and storage infrastructure.”

About Mellanox

Mellanox Technologies (NASDAQ:MLNX) is a leading supplier of end-to-end Ethernet and InfiniBand intelligent interconnect solutions and services for servers, storage, and hyper-converged infrastructure. Mellanox’s intelligent interconnect solutions increase data center efficiency by providing the highest throughput and lowest latency, delivering data faster to applications and unlocking system performance. Mellanox offers a choice of high performance solutions: network and multicore processors, network adapters, switches, cables, software and silicon, that accelerate application runtime and maximize business results for a wide range of markets including high performance computing, enterprise data centers, Web 2.0, cloud, storage, network security, telecom and financial services. More information is available at www.mellanox.com.

Source: Mellanox

The post Mellanox InfiniBand Chosen by AMD for EPYC Datacenter Platforms appeared first on HPCwire.

D.A.V.I.D.E. Supercomputer Named to TOP500, Green500 Lists

Wed, 06/21/2017 - 08:03

FRANKFURT, Germany, June 21, 2017 — At the International Supercomputing Conference 2017, IBM Business Partner and OpenPOWER Foundation member, E4 Computer Engineering, the Italian technology provider of leading-edge solutions for HPC, data analytics and AI, announced that D.A.V.I.D.E. (Development for an Added Value Infrastructure Designed in Europe), a multi-node cluster powered by IBM POWER8 processor technology with NVIDIA Tesla P100 GPU accelerators and NVIDIA NVLink interconnect technology, entered the prestigious TOP500 list.

Twice a year, Top500.org publishes the TOP500 and Green500 lists. The TOP500 ranks supercomputing environments by performance capabilities, as determined by the Linpack benchmark, and recognizes the vendors and technologies that power the most powerful data intensive environments in the world. The Green500 list ranks the top 500 supercomputers in the world by energy efficiency.

D.A.V.I.D.E., developed within the Partnership for Advanced Computing in Europe (PRACE), provides a compelling solution for workloads with highly parallelized code and demanding memory bandwidth requirements such as weather forecasting, QCD, machine learning, computational fluid dynamics and genomic sequencing.

The supercomputer represents the third generation of the Pre-Commercial Procurement project for the development of a Whole-System Design for Energy Efficient HPC, and its innovative design uses the most advanced technologies to create a leading edge HPC cluster that provides powerful performance, low power consumption and ease of use.

D.A.V.I.D.E. was built with best-in-class components. The machine has a total of 45 nodes connected via Infiniband, with a total peak performance of 990 TFlops and an estimated power consumption of less than 2kW per node. Each node is a 2U form factor and hosts two IBM POWER8 Processors with NVIDIA NVLink and four Tesla P100 data center GPUs, with the intra-node communication layout optimized for best performance. Nodes are connected with an efficient EDR 100 Gb/s networking.

The multi-node cluster was fully configured in April 2017 at the E4’s facility in order to perform initial testing, running baseline performance, power and energy benchmarks using standard codes in an aircooled configuration. D.A.V.I.D.E. is currently available for a select number of users for porting applications and profiling energy consumption.

A key feature of the multi-node cluster is an innovative technology for measuring, monitoring and capping the power consumption of the node and of the whole system, through the collection of data from the relevant components (processors, memory, GPUs, fans) to further improve energy efficiency. The technology has been developed in collaboration with the University of Bologna.

“We are delighted to have reached this prestigious result to be included in the TOP500 list. The team worked very hard to design and develop this prototype and is very proud to see the system up and running; we look forward to seeing it fully available to the scientific community,” said Cosimo Gianfreda, CTO, E4 Computer Engineering. “With our work we have demonstrated that it is possible to integrate cost effective technologies to achieve high performance and significantly improve energy efficiency. We thank all our partners for the close collaboration that contributed to this great achievement.”

“HPC and AI are converging and the D.A.V.I.D.E. supercomputer will help the scientific community to run both kinds of workloads on an accelerated system,” said Stefan Kraemer, Director of HPC for EMEA at NVIDIA: “Engery-efficient accelerated computing is the only way to reach the ambitious goals Europe has set for its HPC future.”

About E4 Computer Engineering

Since 2002, E4 Computer Engineering has been innovating and actively encouraging the adoption of new computing and storage technologies. Because new ideas are so important, we invest heavily in research and hence in our future. Thanks to our comprehensive range of hardware, software and services, we are able to offer our customers complete solutions for their most demanding workloads on: HPC, Big-Data, AI, Deep Learning, Data Analytics, Cognitive Computing and for any challenging Storage and Computing requirements. E4. When Performance Matters.

Source: E4 Computer Engineering

The post D.A.V.I.D.E. Supercomputer Named to TOP500, Green500 Lists appeared first on HPCwire.

ARM CEO Simon Segars Joins SoftBank Group Board of Directors

Wed, 06/21/2017 - 07:58

CAMBRIDGE, United Kingdom, June 21, 2017 — ARM Holdings plc today announced that CEO Simon Segars’ appointment to the SoftBank Group Corp. (SBG) Board of Directors was formally approved at SBG’s 37th Annual General Meeting of Shareholders held today in Tokyo.

SBG earlier announced new board candidates to strengthen corporate governance and accelerate growth. Mr. Segars’ appointment comes following SoftBank’s historic 2016 acquisition of ARM for US$32 billion.

“I’m honored to join the SoftBank Board at this exciting next stage of innovation,” said Segars. “In the same way it transformed mobile computing from chip to cloud and all points in between, ARM shapes the smart, connected world of IOT across a common platform. This is at the heart of the SoftBank vision and I look forward to working with the team to make it a reality.”

More information on SBG’s 37th Annual General Meeting of Shareholders can be found here.

About Simon Segars

Simon Segars, CEO, has been at the heart of transformational change in the technology industry for 25 years. He led the development of early ARM processors for the world’s first digital mobile phones and now spearheads ARM’s vision for smart and connected technologies that create better social, economic, education and health prospects for all. Before being appointed CEO in July 2013, Simon held several leadership positions within ARM’s engineering and business development teams. He is based in California’s Silicon Valley, and spends much of his time meeting leading influencers in the world’s technology hubs in the US, Europe and China.

In addition to joining the Board of SoftBank Group, Simon serves on the Boards of the Global Semiconductor Alliance (GSA), the Electric System Design Alliance (ESD Alliance), and as a non-executive director at Dolby Laboratories, Inc.

About ARM

ARM technology is at the heart of the computing and connectivity revolution that is transforming the way people live and businesses operate. Our advanced, energy-efficient processor designs are enabling the intelligence in 100 billion silicon chips and securely powering products from the sensor to the smartphone to the supercomputer. With more than 1,000 technology partners including the world’s largest business and consumer brands, we are driving ARM innovation into all areas compute is happening inside the chip, the network and the cloud.

About SoftBank Group

The SoftBank Group is a global technology player that aspires to drive the Information Revolution. The SoftBank Group is comprised of the holding company SoftBank Group Corp. (TOKYO: 9984) and its global portfolio of companies, which includes advanced telecommunications, Internet services, AI, smart robotics, IoT and clean energy technology providers. In September 2016, ARM Holdings plc, the world’s leading semiconductor IP company, joined the SoftBank Group. To learn more, please visit www.softbank.com.

Source: ARM

The post ARM CEO Simon Segars Joins SoftBank Group Board of Directors appeared first on HPCwire.

Inspur and Intel Announce Enhanced KEEP Program Users Can Experience Intel KNM in Advance

Wed, 06/21/2017 - 01:01

Frankfurt, Germany, June 20, 2017_ Inspur and Intel made a joint announcement at ISC 2017 that they would comprehensively enhance the Knights Landing Evaluation and Escalation Program (KEEP Program) to build the Knight’s Mill (KNM) Infrastructure Experience Platform by the end of the year 2017. Both companies also revealed that the application channels towards worldwide deep-learning and HPC users would open at the first quarter of next year.

Launch view

KNM is the new Xeon Phi processor family that Intel plan to release at the fourth quarter of this year. KNM will use the same infrastructure as KNL, but adopts new instruction set and carries out a number of node optimization for the widely-used deep learning structure. It will achieve both single and double precision variation and efficient expansion, saving a lot of training time for the deep learning model.

Presented by Leijun Hu, VP of Inspur Group

Leijun Hu, Vice President of Inspur Group, said “Inspur and Intel had released the KEEP Program one year ago. We worked together to build the world’s first high-performance computing system based on KNL and provided application testing, optimization and migration platform for the HPC and Deep Learning users. After the program enhancement, the KEEP platform will simultaneously support both KNL and KNM, where the KNL specializes in HPC, and the KNM primarily focuses on AI deep learning. With the support of the leading 100G OPA network, the enhanced KEEP Program will keep up with the trend of HPC and AI integration, and provide satisfying experience for HPC and AI users across the world.”

Barry Davis, Intel’s General Manager of Accelerated Workload Group, stated that AI has already become an exceedingly important computing, and Intel’s KNM has been greatly optimized for deep learning and delivered better performance on the training for the deep learning model. The KNM platform of the enhanced KEEP Program will become an important part in Intel’s global AI computing strategy, and will bring more efficiency to the worldwide users in AI computing.

Caffe Performance on KNL

Apart from the KEEP program enhancement, the two companies also published the Caffe testing report for the world’s first 32 node KNL system. The report shows that a 30X speedup has been achieved on 32 KNL nodes compared with 1 node, and the real performance is 16,516 images per second.

As a leading total solution provider in cloud computing, big data and AI products and solutions, Inspur continues to develop intelligent AI computing. Intel is a world leading processor design and manufacture, and has been leading the way in the entire computing industry. The collaboration between the two companies not only focuses on AI computing processor and HPC acceleration equipment, but also on large-scale AI application research with the establishment of Inspur-Intel calculation collaborative laboratories in China. Intel and Inspur are committed to the application of AI technology in the energy, manufacturing, financial, medical, and insurance industries, and so forth. This will remarkably accelerate the establishment and improvement of global AI computing.

The post Inspur and Intel Announce Enhanced KEEP Program Users Can Experience Intel KNM in Advance appeared first on HPCwire.

AMD Charges Back into the Datacenter and HPC Workflows with EPYC Processor

Tue, 06/20/2017 - 16:23

AMD is charging back into the enterprise datacenter and select HPC workflows with its new EPYC 7000 processor line, code-named Naples, announced today at a “global” launch event in Austin TX. In many ways it was a full frontal assault on Intel’s dominance in the x86 datacenter landscape. Claiming performance and cost advantages and supported by statements from key OEMs, ODMs, and hyperscalers – HPE, Dell, and Microsoft Azure for example – AMD is hoping to convince HPC and datacenter customers it is back for the long haul.

Aware there may be market reluctance after its absence from the datacenter, Scott Aylor, AMD corporate VP and GM of enterprise solutions business, said “It’s not enough to come back with one product, you’ve got to come back with a product cadence that moves as the market moves. So not only are we coming back with EPYC, we’re also [discussing follow-on products] so when customers move with us today on EPYC they know they have a safe home and a migration path with Rome.” AMD has committed to socket compatibility between EPYC 7000 line and Rome, code name of the next scheduled generation AMD processor aimed at the datacenter.

AMD showcased some gaudy performance and price-performance benchmarks comparing EPYC to Broadwell line. In a pre-launch briefing with HPCwire, Aylor said, “These numbers are very big, so they show very measurable separation from what is available with Broadwell. Part of that is quite frankly because we didn’t design EPYC to compete with Broadwell. We designed it to compete with what’s coming. When [Intel’s] Skylake comes later this summer, we think these comparisons will still be very strong against the platinum, gold silver and bronze of Skylake.”

Based on the Zen core, EPYC is a line of system on a chip (SoC) devices designed with enhanced memory bandwidth and fast interconnect in mind. AMD also introduced a one-socket device, optimized for many workloads, which AMD says will invigorate a viable one-socket server market. With EPYC, “we can build a no compromise one-socket offering that will allow us to cover up to 50 percent of the two-socket market that is today held by the [Intel Broadwell] E5-2650 and below,” said Aylor.

AMD clearly has big ambitions. Earlier this spring it introduced Ryzen7 processor line, also based on the Zen core, and targeting high performance gaming. EPYC is aimed squarely at the datacenter. Aylor briefed HPCwire on EPYC before the launch and some of the technical details were still not available. It is an SoC product stack with a range of offerings roughly mimicking the Broadwell product stack. EPYC has up to 32 cores and 8 DDR4 channels per CPU allowing it to address 2TB of memory. The I/O is 128 PCIe lanes.

“The SoC approach we have taken allows all of the IO that has historically lived on an external bridge or IO hub to be fully integrated that into the device,” said Aylor. One result is low latency high performance connections. The PCIe lanes are configurable, “so you can use them to connect to SATA links, directly connect to NVMe links. It also facilitates a strong connection to high performance GPUs.” AMD plans to show an EPYC plus Radeon Instinct GPU machine learning platform at its conference this week.”

AMD presented both SPECint (integer) and SPECfp (floating point) performance comparisons with the Broadwell as well as price point comparisons (how much performance the same number of dollars will be of each processor) some of which are shown below.

“We’re tiering products in 32, 24, and 16-core ranges,” said AYLOR. The idea, of course, is satisfy widely varying needs. The top end aimed at scale out and HPC workloads, he said. The bottom tier allows users to closely manage per core licensing costs. “We have tried to cover the vast majority of the market that exists today in the Broadwell family,” says Aylor. Every product will have a dedicated security processor.

“Sometime people will say benchmarks are interesting but how do you do in the real world. Well we will showcase a fluid dynamics HPC workload, Apache/Spark, and software defined storage reference architecture [at the launch]. We will also have an open stack cloud based implementation,” said Aylor. AMD was expecting on the order 600 attendees for the EPYC launch.

Moving back into the datacenter is a huge bet by AMD that’s required a very substantial investment in the Zen core and EPYC. Seeking to buttress the gamble, AMD has seemingly got buy-in from several market makers and many ecosystem partners. Here are four endorsements included in the official release; while the statements are on the over enthusiastic side they nonetheless suggest AMD has done productive groundwork with partners:

  • HPE. “The EPYC processor represents a paradigm shift in computing and will usher in a new era for the IT ecosystem,” said Antonio Neri, EVP and general manager Enterprise Group, HPE. “Starting with the Cloudline CL3150 and expanding into other product lines later this year, the arrival of EPYC in HPE systems will be welcomed by customers who are eager to deploy the performance and innovation EPYC delivers.”
  • Dell EMC. “As an industry leader, we are committed to driving IT Transformation for our customers,” said Ashley Gorakhpurwalla, president, server solutions division at Dell EMC, “Our next generation of PowerEdge servers are the bedrock of the modern datacenter that are designed to maximize business scalability and intelligent automation with integrated security. The combination of PowerEdge and the AMD EPYC performance and security capabilities will create unique compute solutions for our customers to accelerate workloads and protect their business.”
  • Baidu. “As the world’s largest Chinese language search engine and leading AI-Tech company, Baidu prides itself on simplifying a complex world through technology,” said By Dr. Zhang Ya Qin, president of Baidu. “The AMD EPYC processor powered one-socket server can significantly increase our datacenter computing efficiency, reduce TCO and lower energy consumption. We will start deploying with the launch of AMD EPYC and I look forward to our cooperation leading to scaled EPYC adoption this year, and ongoing innovations.”
  • Microsoft. “We’ve worked to make Microsoft Azure a powerful enterprise grade cloud platform, that helps guide the success of our customers, no matter their size or geography,” said Girish Bablani, corporate vice president, Azure Compute, Microsoft Corp. “To power Azure, we require the most cutting-edge infrastructure and the latest advances in silicon which is why we intend to be the first global cloud provider to deliver AMD EPYC, and its combination of high performance and value, to customers.”

The single socket gambit is another interesting aspect to AMD’s initiative. Currently two socket servers rule the roost.

Here’s the AMD pitch: “In our one socket offering we have come up with a clever way to maintain all of the I/O capabilities that you would get in a two socket as well as the full complement of eight memory channels. Today people buy two socket, sometimes because they need to, but more often than not because they have to. There are many examples in which I/O rich [workloads] like storage, like GPU compute, and some vertical workloads where people don’t necessarily need two sockets from a CPU performance perspective,” said Aylor.

AMD’s single socket offering cuts costs substantially, according to Aylor. “We’ve selectively optimized a couple of skews for one socket only. So these are skews that are one socket capable only,” said Aylor. As an example of how the one socket and two socket offerings are distinguished, he cited on package interconnect, “The infinity fabric that would normally connect the two sockets in a two socket system, we repurpose that interconnect into more I/O lanes and that’s how you have in a two socket solution 128 lanes of PCIe and in a one socket solution you still keep the same level of connectivity.”

AMD has singled out a number of vertical as good fits for one socket EPYC servers. Perhaps not surprisingly, storage is one. “Not only base line storage but software defined storage with EPYC’s ability to attach a massive number of SATA drives to a one socket. We also see a strong opportunity in certain areas of high performance computing, especially those that tend to focus on memory bound application. And we have an oil and gas reservoir simulation demo,” said Aylor.

Link to AMD press release: https://www.hpcwire.com/off-the-wire/amd-unveils-record-setting-epyc-datacenter-processor/

The post AMD Charges Back into the Datacenter and HPC Workflows with EPYC Processor appeared first on HPCwire.

TYAN Launches AMD EPYC Processor-Based Server Platform

Tue, 06/20/2017 - 15:30

AUSTIN, Texas, June 20, 2017 — TYAN, an industry-leading server platform design manufacturer and subsidiary of MITAC Computing Technology Corporation, today introduced new AMD EPYC processor-based server solutions featuring large memory capacity and NVMe connectivity for high performance storage applications.   

The AMD EPYC processor-based, 2U single-socket TN70A-B8026 server and single-socketS8026 server motherboard are designed for all-flash, high performance storage applications. The TN70A-B8026 supports 16 DDR4 DIMM slots, 24 hot-swap NVMe U.2 drives and 2 NVMe M.2 drives; the S8026 supports 8 OCuLink 8x (for NVMe U.2 drive connections), 2 NVMe M.2 and 2 SATA 6G connectors in EATX form factor. TYAN’s EPYC storage server platform and motherboard leverages the features of EPYC processor to provide a large number of direct-connected NVMe SSDs in a dense form factor without the use of PCIe switches.

“AMD EPYC processors are based on the 14nm “Zen” x86 core architecture with 32 cores and 64 threads, featuring 8 memory channels and up to16 DIMMs per socket. The new powerful processor allows TYAN to offer the optimal balance of computing, memory and I/O for our customers,” said Danny Hsu, Vice President of MiTAC Computing Technology Corporation’s TYAN Business Unit. “By adopting the latest AMD EPYC processor technology, TYAN brings high memory channel bandwidth and PCI Express high speed I/O connectivity features to enterprises and datacenters.”

“The EPYC CPU is optimized to offer unprecedented levels of performance, bandwidth and connectivity for server platforms. TYAN maximizes the EPYC processor’s value in a 2U single-socket system to deliver high memory bandwidth, high I/O throughput, and high I/O & storage connectivity, and the performance normally seen in a dual-socket platform,” said Scott Aylor, CVP and GM, AMD Enterprise Solutions. “Our collaboration with TYAN not only enables customers to enjoy dual-socket capability with single-socket economics, but also drives the innovation to better address the needs of existing and emerging datacenter workloads.” 

TYAN’s EPYC-based Solutions:

  • TN70A-B8026: 2U single-socket EPYC processor-based platform supports 16 DDR4 DIMM slots with 8 memory channels, 24 2.5” hot-swap NVMe U.2 and 2 NVMe M.2 drives. In addition to onboard 2 GbE connections, the OCP v2.0 LAN mezzanine slot supports a wide range of high-speed networking technology options including EDR InfiniBand and 100Gb Ethernet.
  • S8026: Single-socket EPYC processor-based server motherboard in EATX (12” x 13”) form factor supports 16 DDR4 DIMM slot with 8 memory channels, 8 OCuLink 8x (for NVMe U.2 drive connections), 2 NVMe M.2 and 2 SATA 6G connectors. Through proper BIOS setting, the specific pair of OCuLink 8x connectors can be re-configured to support up to 16 SATA 6G drives, which makes the motherboard provides up to 18 native SATA 6G connections.

About TYAN

TYAN, as a leading server brand of Mitac Computing Technology Corporation under the MiTAC Group (TSE:3706),designs, manufactures and markets advanced x86 and x86-64 server/workstation board technology, platforms and server solution products. Its products are sold to OEMs, VARs, System Integrators and Resellers worldwide for a wide range of applications. TYAN enables its customers to be technology leaders by providing scalable, highly-integrated, and reliable products for a wide range of applications such as server appliances and solutions for high-performance computing and server/workstation used in markets such as CAD, DCC, E&P and HPC. For more information, visit MiTAC’s website at http://www.mitac.com or TYAN’s website at http://www.tyan.com

Source: TYAN

The post TYAN Launches AMD EPYC Processor-Based Server Platform appeared first on HPCwire.

AMD Unveils Record-Setting EPYC Datacenter Processor

Tue, 06/20/2017 - 15:23

AUSTIN, Texas, June 20, 2017 — ​AMD, and a global ecosystem of server partners, today marked a new era in the datacenter with the launch of AMD EPYC 7000 series high-performance datacenter processors. AMD was joined by multiple customers and partners at the global launch event in presenting a wide array of systems, performance demonstrations, and customer testimonials. The innovative, record-setting AMD EPYC design, with up to 32 high-performance “Zen” cores and an unparalleled feature set, delivers greater performance than the competition across a full range of integer, floating point, memory bandwidth, and I/O benchmarks and workloads.

“With our EPYC family of processors, AMD is delivering industry-leading performance on critical enterprise, cloud, and machine intelligence workloads,” said Lisa Su, president and CEO, AMD. “EPYC processors offer uncompromising performance for single-socket systems while scaling dual-socket server performance to new heights, outperforming the competition at every price point. We are proud to bring choice and innovation back to the datacenter with the strong support of our global ecosystem partners.”

The world’s largest server manufacturers introduced products based on AMD EPYC 7000-series processors at today’s launch, including HPE, Dell, Asus, Gigabyte, Inventec, Lenovo, Sugon, Supermicro, Tyan, and Wistron. Primary hypervisor and server operating system providers Microsoft, Red Hat, and VMware showcased optimized support for EPYC, while key server hardware ecosystem partners Mellanox, Samsung Electronics, and Xilinx were also featured in EPYC-optimized platforms.

Leading Server OEM Platforms

“The EPYC processor represents a paradigm shift in computing and will usher in a new era for the IT ecosystem,” said Antonio Neri, EVP and general manager Enterprise Group, HPE. “Starting with the Cloudline CL3150 and expanding into other product lines later this year, the arrival of EPYC in HPE systems will be welcomed by customers who are eager to deploy the performance and innovation EPYC delivers.”

“As an industry leader, we are committed to driving IT transformation for our customers,” said Ashley Gorakhpurwalla, president, server solutions division at Dell EMC, “Our next generation of PowerEdge servers are the bedrock of the modern data center that are designed to maximize business scalability and intelligent automation with integrated security. The combination of PowerEdge and the AMD EPYC performance and security capabilities will create unique compute solutions for our customers to accelerate workloads and protect their business.”

Cloud Datacenter and Enterprise Customers

Datacenter and cloud service providers also welcomed EPYC to the market today. Members of the “Super 7” datacenter services providers, including Baidu and Microsoft Azure, as well as 1&1, Bloomberg, Dropbox and LexisNexis, all voiced their support at launch.

“As the world’s largest Chinese language search engine and leading AI-Tech company, Baidu prides itself on simplifying a complex world through technology,” said By Dr. Zhang Ya Qin, president of Baidu. “The AMD EPYC processor powered one-socket server can significantly increase our datacenter computing efficiency, reduce TCO and lower energy consumption. We will start deploying with the launch of AMD EPYC and I look forward to our cooperation leading to scaled EPYC adoption this year, and ongoing innovations.”

“We’ve worked to make Microsoft Azure a powerful enterprise grade cloud platform, that helps guide the success of our customers, no matter their size or geography,” said Girish Bablani, corporate vice president, Azure Compute, Microsoft Corp. “To power Azure, we require the most cutting-edge infrastructure and the latest advances in silicon which is why we intend to be the first global cloud provider to deliver AMD EPYC, and its combination of high performance and value, to customers.

Record-Setting EPYC Performance

The excitement around EPYC is driven by multiple record-setting server benchmarks achieved by EPYC-powered one-socket and two-socket systems.

AMD EPYC processors set several performance records, including:

  • Two-Socket Server
    • AMD EPYC 7601-based system scored 2360 on SPECint_rate2006, higher than any other two-socket system score1
  • One-Socket Server
    • AMD EPYC 7601-based system scored 1200 on SPECint_rate2006, higher than any other mainstream one-socket x86-based system score2
    • AMD EPYC 7601-based system scored 943 on SPECfp_rate2006, higher than any other one-socket system score3

All EPYC processors combine innovative security features, enterprise class reliability, and support a full feature-set. An AMD EPYC™ 7601 CPU-based one-socket system shifts expectations for single socket server performance, helping lower total-cost-of-ownership (TCO), providing up to 20% CapEx savings compared to the Intel Xeon E5-2660 v4-based two-socket system4. At every targeted price point for two-socket processors, EPYC outperforms the competition, with up to 70% more performance in the eight hundred dollar price band and up to 47% more performance at the high-end of the market of four thousand dollars or more5.

EPYC Product Overview

  • A highly scalable System on Chip (SoC) design ranging from 8-core to 32-core, supporting two high-performance threads per core.
  • Industry-leading memory bandwidth across the line-up, with 8 channels of memory on every EPYC device. In a two-socket server, support for up to 32 DIMMS of DDR4 on 16 memory channels, delivering up to 4 terabytes of total memory capacity.
  • Unprecedented support for integrated, high-speed I/O with 128 lanes of PCIe® 3 on every product
  • A highly-optimized cache structure for high-performance, energy efficient compute
  • AMD Infinity Fabric coherent interconnect linking EPYC CPUs in a two-socket system
  • Dedicated security hardware

EPYC Product Lineup

​MODEL​​ ​CORE / THREAD ​BASE FREQ. ​MAX BOOST ​TDP ​EPYC™ 7601 ​32 / 64 ​2.2 GHz ​3.2 GHz ​180W​​ ​EPYC™ 7551P ​32 / 64 ​2.0 GHz ​3.0 GHz ​180W ​EPYC™ 7501 ​32 / 64 ​2.0 GHz ​3.0 GHz ​155/170W ​EPYC™ 7451 ​24 / 48 ​2.3 GHz ​3.2 GHz ​180W ​EPYC™ 7401P ​24 / 48 ​2.0 GHz ​3.0 GHz ​155/170W ​EPYC™ 7351P ​16 / 32 ​2.4 GHz ​2.9 GHz ​155/170W ​EPYC™ 7301 ​16 / 32 ​2.2 GHz ​2.7 GHz ​155/170W ​EPYC™ 7281 ​16 / 32 ​2.1 GHz ​2.7 GHz ​155/170W ​EPYC™ 7251 ​8 / 16 ​2.1 GHz ​2.9 GHz ​120W


Additional Resources

Cloud Service Provider Support


“At Bloomberg, we handle the flow of information for professionals in the capital markets. We look forward to AMD’s leadership in the open standards communities, such as OCP, NVMe, GenZ, and CCIX, to help accelerate the industry-wide adoption of these innovative data center, storage, and interconnect solutions,” said Justin Erenkrantz, head of compute architecture for Bloomberg, the global finance, media and tech company based in New York City. “With higher density and lower latency NVMe storage a primary feature delivered by the AMD EPYC processor, we fully expect to realize the next level of performance and cost efficiency.”


“Dropbox customers expect fast, reliable access to the content they ask us to manage for them and EPYC delivers on those requirements. We have worked closely with AMD during our evaluation of EPYC in our environment and see significant potential in lowering total-cost-of-ownership while improving performance in single-socket designs,” said Akhil Gupta, vice president of infrastructure at Dropbox. “Our evaluation systems take advantage of the industry-leading 128 lanes of PCIe on EPYC for storage performance and capacity. Dropbox is exploring deployment options for EPYC later this year, and I believe the future looks bright for the relationship with AMD and EPYC.”


At LexisNexis Risk Solutions, we believe in the power of data and advanced analytics for better risk management. As a trusted data analytics provider for organizations seeking actionable insights to manage risks and improve results while upholding the highest standards for security and privacy,” said Flavio Villanustre, vice president, Technology, LexisNexis Risk Solutions, RELX Group. “LexisNexis processes more than 90 million transactions per hour with HPCC Systems®, a proven, open source solution for Big Data. LexisNexis Risk Solutions and AMD have teamed up to optimize the HPCC Systems platform to take advantage of the benefits of thread density, core performance, memory bandwidth and the industry leading 128 lanes of PCIe per socket of the AMD EPYC processor.”

OEM / ODM Support


“We recognized early on the significant performance and scalability potential of EPYC for a number of ASUS platforms spanning HPC and virtualization workloads,” said Robert Chin, head of ASUS Server business unit. “As a longtime AMD technology partner, we are thrilled to design EPYC-based products that will transform the datacenter experience and TCO for our customers.”

Gigabyte Technology

“GIGABYTE Technology is committed to pushing design boundaries in the enterprise server market, and with the AMD EPYC CPU we can take the customer experience to the next level with a single socket solution,” said Daniel Hou, vice president, Research & Development, GIGABYTE Technology. “By packing increased performance, memory bandwidth and I/O into a smaller footprint, EPYC will enable us to expand our portfolio and address channel partners and system integrators in new target markets.”


“Inventec is dedicated to driving innovation in a number of industries by combining AMD CPU and GPU technologies on our platforms, and AMD is a key technology partner for Inventec in the cloud datacenter market,” said Jack Tsai, general manager of Inventec EBG. “By integrating the new EPYC CPU into our server portfolio, we can provide a more competitive, scalable and higher performance solution to our worldwide datacenter customers.”


“The AMD EPYC processors present unique opportunities for our customers to lower Total Cost of Ownership via an unprecedented balance of cores, memory bandwidth, and I/O. We are excited to collaborate with AMD and several global Hyperscale customers to develop and deploy single socket and dual socket EPYC-based servers,” said Paul Ju, vice president and general manager, Lenovo Global Hyperscale Business.


“Supermicro’s new generation of server solutions supporting AMD EPYC processors will unlock many opportunities for datacenter customers to leverage these innovative platforms,” said Don Clegg, vice president of Marketing and Business Development at Supermicro. “AMD EPYC processors with more cores, more memory bandwidth, and more I/O integrate effortlessly with our Supermicro application-optimized platforms to provide leadership performance per-watt and per-dollar to optimize data center TCO. We are delighted to embrace this opportunity to strengthen and grow our mutually productive technology partnership with AMD.”


“Tyan has a long history of supporting AMD, and we are excited about the impact the company’s return to the datacenter market will have on our customers,” said Danny Hsu, vice president of MiTAC Computing Technology Corporation’s TYAN Business Unit. “We see major promise for storage and HPC platforms with the EPYC CPU, as well as new possibilities for combined GPU solutions, enabling a wide array of new applications for customers around the globe.”


“Wistron is glad to work with AMD to provide the most cost-effective all-flash array storage solutions for businesses competing in a constantly changing technology landscape,” said Peter Tung, chief operation officer of Enterprise Business Group, Wistron. “The value proposition for AMD’s EPYC CPU, particularly in non-volatile memory storage applications, is a major differentiator, and enables datacenter and enterprise customers to achieve high IOPS and throughput storage solutions.”

Hardware Partner Support


“In a data-centric world, we need to be able to analyze growing amounts of data and to be able to find data insights in real time. The combination of the unmatched data throughput and capacity of EPYC and the intelligent interconnect solutions from Mellanox will enable our customers and partners to maximize their application performance and overall data center return on investment.” said Michael Kagan, chief technology officer at Mellanox Technologies. “We are happy to be the preferred interconnect solution provider for the AMD EPYC platforms and look forward to continuing to collaborate with AMD and the OEM partners to enable world leading data center platforms.”

Samsung Electronics

“Samsung, in continuing our ongoing collaboration with AMD, has been working to deliver new leading-edge memory and flash storage solutions that will be enhanced by EPYC,” said Jim Elliott, corporate vice president, Samsung Semiconductor, Inc. “The industry-leading memory bandwidth and the outstanding I/O capacity delivered by EPYC enable Samsung and AMD to collectively offer an exceptionally high level of performance and efficiency to data center customers. Samsung solutions combined with the power of EPYC are planned for later this year, offering an advanced platform for in-memory database and analytics, High Performance Computing workloads, and more.”


“The launch of the AMD EPYC processor signifies an important milestone in the industry,” said Victor Peng, chief operating officer at Xilinx. “Together with Xilinx’s All Programmable devices, the EPYC platform provides outstanding performance when accelerating data center applications. We are also delighted to be working with AMD in furthering open data center standards, such as the CCIX interconnect, to provide the necessary heterogeneous computing solutions for next generation workloads.”

About AMD

For more than 45 years AMD has driven innovation in high-performance computing, graphics and visualization technologies ― the building blocks for gaming, immersive platforms, and the datacenter. Hundreds of millions of consumers, leading Fortune 500 businesses and cutting-edge scientific research facilities around the world rely on AMD technology daily to improve how they live, work and play. AMD employees around the world are focused on building great products that push the boundaries of what is possible. For more information about how AMD is enabling today and inspiring tomorrow, visit the AMD (NASDAQ: AMD) website, blog, and Facebook and Twitter pages.

Source: AMD

The post AMD Unveils Record-Setting EPYC Datacenter Processor appeared first on HPCwire.

Univa Contributes Universal Resource Broker to the Open Source Community

Tue, 06/20/2017 - 11:12

FRANKFURT, Germany, June 20, 2017 – Univa, the Data Center Workload Optimization Company, today announced the contribution of its Universal Resource Broker (URB) technology to the open-source community.

The Universal Resource Broker is a software solution that allows distributed application frameworks written for Apache Mesos to run seamlessly on Univa Grid Engine. Making URB available as an open-source project opens the door to continued innovation, enabling community contributors to build adapters to additional workload managers and application frameworks. In addition to open-sourcing the project, Univa is extending URB to support Kubernetes clusters as well.

By using URB, users have the option of deploying any Mesos compatible framework (including Spark, Hadoop, Storm, Jenkins, Marathon and Chronos) along with any other type of workload using Univa Grid Engine as an underlying workload management substrate for high performance and high throughput environments. Users can also choose to run URB on a variety of Kubernetes based cluster solutions in cases where containerized microservices are key. Enterprises requiring more powerful scheduling features and policy-based controls will want to select the combination of Navops Command, URB and Kubernetes with the additional option of Navops Command’s Mixed Workload support for non-containerized applications run inside the Kubernetes environment.

According to Fritz Ferstl, CTO at Univa Corporation, “This is an important development for our large installed base of Grid Engine customers and for the burgeoning Kubernetes ecosystem. With the Universal Resource Broker, customers can easily deploy a single cluster supporting batch, interactive, containerized and now Mesos API driven workloads without contention. We’re continuing to deliver on our promise to help customers achieve better business results on a more cost-efficient, shared infrastructure.”

URB provides Kubernetes users with a seamless way to run application frameworks written for Mesos while protecting existing investments. The technology complements Univa’s Navops Command, an advanced, enterprise-proven policy management solution for Kubernetes that supports mixed workloads. With URB and Navops Command, Univa supports the widest variety of container and non-container based application workloads on a shared Kubernetes environment.

“Open source software is critical in the modern application container ecosystem and market, where availability, flexibility and integration can be enhanced by open source software components and frameworks.“ said Jay Lyman, principal analyst for 451 Research. “Univa’s URB also supports the variety of software – including big data, continuous integration and container management and orchestration technology – that is critical to success in enterprise IT today. “

“We are pleased to contribute this key technology to the open source community” said Rob Lalonde, General Manager of Univa’s Navops business unit. “This contribution demonstrates our ongoing commitment to open source software and to helping organizations get the most from their infrastructure investments. We are very excited to extend the capabilities of URB to Kubernetes.”

Univa supports a wide range of technologies including Mesos frameworks, Batch workloads and containerized microservices native to Kubernetes.

The open-source Universal Resource Broker for Grid Engine is expected to be available in July of 2017 and will be released under an Apache open-source license. URB for Kubernetes is planned for August of 2017 availability.

For more information visit www.univa.com, contact Univa at sales@univa.com or stop by booth #C-1214 at the ISC High Performance 2017, June 19-21 in Frankfurt, Germany.

About Univa Grid Engine
Univa Grid Engine is the leading workload management system. The solution maximizes the use of shared resources in a datacenter and applies advanced policy management tools to deliver products and results faster, more efficiently, and with lower overall costs. The product can be deployed in any technology environment, including containers: on-premises, hybrid or in the cloud. A variety of add-ons can be utilized to extend workload management capabilities and create a customized solution for any enterprise infrastructure. For more information, please visit www.univa.com or follow on Twitter @Grid_Engine

About Navops

Navops is a suite of products that enables enterprises to take full advantage of Kubernetes and provides the ability to quickly and efficiently run containers at scale. Navops utilizes workload placement and advanced policy management across on-premises, cloud, and hybrid infrastructures. With Navops, companies can automate microservices applications and efficiently respond to end user demand. For more information, please visit www.navops.io or follow on Twitter @Navops

About Univa Corporation

Univa, the Data Center Automation Company, is the leading provider of automation and management software for computational and big data infrastructures. Our products and global enterprise support give our customers the power to manage all of their compute resources, no matter how big or where deployed. Many of the leading brands in the world depend on Univa’s unsurpassed expertise, and premier services and support. Univa is headquartered in Hoffman Estates, Illinois, USA, with offices in Markham, ON, Canada, Munich and Regensburg, Germany.

Source: Univa

The post Univa Contributes Universal Resource Broker to the Open Source Community appeared first on HPCwire.

Supermicro Announces X11 SuperBlade

Tue, 06/20/2017 - 09:24

SAN JOSE, Calif., June 20, 2017 — Super Micro Computer, Inc. (NASDAQ: SMCI) delivers the Intel Omni-Path Architecture (Intel OPA) in the new generation, high-performance, X11 8U/4U SuperBlade systems supporting the upcoming Intel Xeon Processor Scalable Family (codenamed Skylake).

The X11 SuperBlade is a density and performance optimized solution for high performance and Artificial Intelligence applications. The SuperBlade system supports up to 20x 2-socket servers,10x 4-socket servers or 20x Xeon® Phi based servers, as well as 1x 100G Intel OPA or 100G EDR InfiniBand switch and 2x 10G/25G or 4x 25G Ethernet switches, with open industry standard remote management software for both servers, storage and networking. The integrated 100G Intel OPA switch is optimized for applications that require lowest latency and highest throughput. It can scale to thousands of nodes for high-performance workloads and provides adaptive routing to discover the least congested path, dispersive routing for multiple routes for redundancy and load balancing, packet integrity protection to allow the recovery of transient errors and lane scaling to deal with lane failure. The enclosure has optional Battery Backup Power (BBP) modules replacing high cost datacenter UPS systems for reliability and data protection.

“Our SuperBlade platform integrated with the Intel Omni Path Architecture based switch provides a dense HPC solution optimized with a low 100ns latency and maximized 100 Gb/s throughput for enhanced reliability and quality of service,” said Charles Liang, President and CEO of Supermicro. “Larger HPC deployments will benefit from best in class density, maximum processing performance and integrated high performance fabric provided by the new SuperBlade.”

“Intel Omni-Path Architecture delivers the high performance, reliable and cost effective interconnect demanded by HPC,” said Scott Misage, General Manager for High-Performance Fabric at Intel. “Supermicro with their innovative X11 SuperBlade builds on this foundation to offer a compelling, high density solution for HPC users.”

A 4U/8U X11 based SuperBlade is a scalable, modular solution that includes:

  • 20x 2-socket X11 based blade servers or
  • 10x 4-socket X11 based blade servers or
  • 20x next-generation Intel Xeon Phi processor 72×5 product family based servers (codenamed Knights Mill) Shipping in Q4’17
  • 100G Intel Omni-Path Architecture switching and extensive additional networking options
  • 96% efficient redundant Titanium Level power supplies
  • Integrated Battery Backup Power (BBP) modules replacing high cost datacenter UPS systems for reliability and data protection (Optional)

Supermicro will demonstrate the SuperBlade with the Intel Omni-Path Architecture based switch at the ISC High Performance 2017 event in Messe Frankfurt, Germany, June 18-22, 2017.

About Super Micro Computer, Inc. (NASDAQ: SMCI)

Supermicro (NASDAQ: SMCI), the leading innovator in high-performance, high-efficiency server technology is a premier provider of advanced server Building Block Solutions for Data Center, Cloud Computing, Enterprise IT, Hadoop/Big Data, HPC and Embedded Systems worldwide. Supermicro is committed to protecting the environment through its “We Keep IT Green” initiative and provides customers with the most energy-efficient, environmentally-friendly, solutions available on the market. For more information, please visit, http://www.supermicro.com.

Source: Supermicro

The post Supermicro Announces X11 SuperBlade appeared first on HPCwire.