Visit additional Tabor Communication Publications
November 19, 2009
When first deployed in 2005, the Jaguar supercomputer at Oak Ridge National Lab booted up with a peak speed of only 26 teraflops. Since then it has been continuously enhanced with additional cabinets and new AMD Opteron processors. The latest upgrade involved replacing the quad-core Opteron chips with AMD's latest six-core version, which propelled it to the number one spot on the newly announced TOP500 list.
With a Linpack mark of 1.759 petaflops, it outran the number two Roadrunner system by a good 750 teraflops. Jaguar also managed to come out on top in the HPC Challenge STREAM benchmark, with a sustainable memory bandwidth of 398 terabytes per second.
We asked John Fruehe, AMD's director of Opteron product marketing, and Buddy Bland, the project director for the ORNL Leadership Computing Facility, about the significance of this accomplishment and what it means for the most demanding supercomputing applications.
HPCwire: What do you think is the significance of the first multi-petaflop machine powered purely by x86 CPUs?
John Fruehe: The number one position is really significant because it shows a culmination of supercomputing's shift to industry-standard systems. These mammoth, world-class machines are no longer out of reach for the more average academic or enterprise HPC user. The market for a Cray XT5 or an Appro HyperCluster goes beyond the U.S. national labs and you see HPC customers monitoring regional weather patterns or searching out oil and gas reserves on the very same systems as what you have in the Top 10.
Let's face it, most folks can't go out and buy an Earth Simulator or a Blue Gene. x86 has made world-class supercomputing a lot more democratic. And of course, this number one win shows the raw processing capability of x86 and specifically AMD's brand of x86 in the form of the game-changing Opteron processor. In the past, larger, more expensive and proprietary systems ruled the top of the chart. Today, more economical and scalable x86 platforms are rapidly becoming the norm for supercomputing and that gives customers more flexibility and choice.
HPCwire: In performance-per-watt, Jaguar still lags other more exotic supercomputing architectures. Given the escalating concern of energy efficiency in these large-scale machines, what does that say about the role of the x86 in future supercomputing systems?
Fruehe: Certainly an architecture like Cell is quite the power miser. But as we've seen with "Roadrunner," frankly, the Cell architecture needs Opteron to get the job done. A system like "Jaguar" or any of the other systems that are near the top of the list are deployed to do a specific job. Many times, power is certainly a concern, but not the overriding concern.
For more mainstream HPC, I think we will continue to see x86 dominate because of economics and because it delivers the performance and is what the industry knows best. We'll continue to see the additional low-power improvements that AMD is implementing -- above and beyond what we've already done -- take hold and bring HPC overall more in line with an acceptable level of power draw.
As AMD moves to its future generations, you will continue to see an emphasis on power efficiency in the data center as we continue to drive greater amounts of performance and scalability while staying within approximately the same power/thermal ranges, resulting in increasingly greater performance per watt with each generation.
HPCwire: Jaguar's Linpack performance is certainly impressive. But what types of applications are going to be able to fully utilize the scale of this machine?
Buddy Bland: While Linpack is a test of the computational performance of computer systems, Jaguar was designed to run applications that are demanding on all of the aspects of the system. Within just a few weeks of completing the upgrade of Jaguar, we have several applications that are scaling to use the full performance of the system. Three of this year's Gordon Bell award finalists are running on Jaguar using the full scope of the machine. We also have many of our key applications in areas such as materials science, computational chemistry, fusion energy, superconductivity, and bioenergy using all of Jaguar today. We expect that as the remainder of our users get access to the upgraded system, we will see most of our applications taking full advantage of the size of the system.
HPCwire: What other types of applications are slated to get time on Jaguar?
Bland: The DOE INCITE program allocates time on the leadership systems: Jaguar at ORNL and Intrepid at ANL. A small number of scientifically important, time critical applications from government laboratories, academia and industry are awarded large blocks of time. In 2009, 38 projects received allocations of time on Jaguar as part of the INCITE program.
Jaguar is supporting some of the most important projects of our time such as:
HPCwire: Is there an upgrade path for Jaguar beyond its current configuration?
Bland: The socket replaceable line of processors from AMD and the board compatible line of systems from Cray have been a key part of the success of Jaguar. We have been able to upgrade cabinets from single-core to dual-core to quad-core and now to 6-core processors while preserving much of our investment. This allowed ORNL and Cray to upgrade Jaguar, stepping up from 26 teraflops to 54 TF, 119 TF, 263 TF, and now 2.3 petaflops. Without this series of increasingly powerful systems, we would not have been able to continuously move the users and their applications to higher and higher performance, resulting in the scientific success we have seen from the system. Cray's line of systems may yet provide another upgrade path for Jaguar.
Jun 19, 2013 |
Supercomputer architectures have evolved considerably over the last 20 years, particularly in the number of processors that are linked together. One aspect of HPC architecture that hasn't changed is the MPI programming model.
Jun 18, 2013 |
The world's largest supercomputers, like Tianhe-2, are great at traditional, compute-intensive HPC workloads, such as simulating atomic decay or modeling tornados. But data-intensive applications--such as mining big data sets for connections--is a different sort of workload, and runs best on a different sort of computer.
Jun 18, 2013 |
Researchers are finding innovative uses for Gordon, the 285 teraflop supercomputer housed at the San Diego Supercomputer Center (SDSC) that has a unique Flash-based storage system. Since going online, researchers have put the incredibly fast I/O to use on a wide variety of workloads, ranging from chemistry to political science.
Jun 17, 2013 |
The advent of low-power mobile processors and cloud delivery models is changing the economics of computing. But just as an economy car is good at different things than a full size truck, an HPC workload still has certain computing demands that neither the fastest smartphone nor the most elastic cloud cluster can fulfill.
Jun 14, 2013 |
For all the progress we've made in IT over the last 50 years, there's one area of life that has steadfastly eluded the grasp of computers: understanding human language. Now, researchers at the Texas Advanced Computing Center (TACC) are utilizing a Hadoop cluster on its Longhorn supercomputer to move the state of the art of language processing a little bit further.
05/10/2013 | Cleversafe, Cray, DDN, NetApp, & Panasas | From Wall Street to Hollywood, drug discovery to homeland security, companies and organizations of all sizes and stripes are coming face to face with the challenges – and opportunities – afforded by Big Data. Before anyone can utilize these extraordinary data repositories, however, they must first harness and manage their data stores, and do so utilizing technologies that underscore affordability, security, and scalability.
04/15/2013 | Bull | “50% of HPC users say their largest jobs scale to 120 cores or less.” How about yours? Are your codes ready to take advantage of today’s and tomorrow’s ultra-parallel HPC systems? Download this White Paper by Analysts Intersect360 Research to see what Bull and Intel’s Center for Excellence in Parallel Programming can do for your codes.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?
Join our webinar to learn how IT managers can migrate to a more resilient, flexible and scalable solution that grows with the data center. Mellanox VMS is future-proof, efficient and brings significant CAPEX and OPEX savings. The VMS is available today.