| View previous topic :: View next topic |
| Author |
Supercomputing |
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
 Posted: Sun Apr 15, 2007 6:42 pm Post subject: Supercomputing |
 |
|
Cray's supercomputer at the Department of Energy's Oak Ridge National Laboratory (ORNL) upgraded (doubled) its performance to 119 teraflops - putting it number 2 on the world's Top 500 list of supercomputers:
http://www.hpcwire.com/hpc/1370386.html |
|
| Back to top |
|
 |
| Author |
Supercomputing Replies |
rffrydr
Moderator
Joined: 30 Oct 2005
Posts: 11518
Location: Sunny California
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
| Posted: Fri May 29, 2009 3:27 am Post subject: |
|
|
The latest trends in the supercomputing space, courtesy of Top500.org:
http://www.top500.org/blog/2009/05/20/top_trends_high_performance_computing
| Quote: | Introduction
If we calculate the performance predictions for the TOP500 list, then we will see a 100 Petaflops system most likely in the year 2016. In the past we had a performance increase within 11 years from Gigaflops (Cray2 in 1986), via Teraflops (Intel ASCI Red in 1997) up to the Petaflops ( IBM Roadrunner in 2008 ) by a factor of 1000. Despite these performance improvements the HPC arena will more and more be defined indirectly via the highly compute-intensive applications. They are coming from a variety of areas, involving quantum mechanical physics, weather forecasting, climate research, molecular modeling (computing the structures and properties of chemical compounds, biological macromolecules, polymers, and crystals), physical simulations (such as simulation of airplanes in wind tunnels and research into nuclear fusion), cryptanalysis, and improved seismic processing for oil exploration for continued supply. For most of these applications detailed results may only be achieved with systems in the Petaflops range. And hopefully Exascale Systems will be seen first in 2019.
But what are the trends in the nearest future which will form the basis for all these performance predictions. The major challenges to all processor requirements for HPC systems now and in the future will be: low cost, low power consumption, availability of support for parallel programming, and efficient porting of existing codes. |
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
| Posted: Wed May 27, 2009 12:23 am Post subject: |
|
|
Germany unveils the fastest supercomputer in Europe - and the third fastest in the world:
http://www.google.com/hostednews/afp/article/ALeqM5gR8XY_dducd-C_KFpWxRTeBQh1Cw
| Quote: | The "Jugene," capable of 1,000,000,000,000,000 calculations per second, ranks behind the "Roadrunner" and "Jaguar" computers in the United States, said Kosta Schinarakis from the Juelich research centre, where the computer is located.
Jugene will be used for a wide variety of operations, including research on fuel cells for electric cars, weather forecasting and the origins of the universe, the centre said.
The machine is no ordinary PC, requiring 295,000 processors located in 72 lockers each the size of a telephone box. |
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
| Posted: Thu Apr 30, 2009 7:12 pm Post subject: |
|
|
Aside from modeling our nation's nuclear stockpile and various astrophysics projects, the most powerful of them all is also assisting with other projects such as those that deal with alternative energy, protein folding, etc.
http://www.wired.com/wiredscience/2009/04/supercomputers-2/
| Quote: | Jackie Chen of Sandia National Laboratories is using 30 million processor-hours on Jaguar to simulate the combustion process of alternative fuels, like biofuel and ethanol. Her modeling of flames, ignition and turbulence can influence engine design, allowing for higher-efficiency and lower emissions vehicles.
“To understand the underlying physics of what’s going on in the internal combustion engines with alternative fuels,” said Chen, “we need some of the world’s largest calculations.”
Other ongoing projects seek to understand how proteins misfold in neurodegenerative diseases, develop thermoelectric materials to capture wasted heat from tailpipe emissions and create high-resolution climate models. |
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
| Posted: Sun Apr 26, 2009 11:42 pm Post subject: |
|
|
IBM's Blue Gene to compete on ‘Jeopardy!’:
http://www.nytimes.com/2009/04/27/technology/27jeopardy.html?ref=technology
| Quote: | Eric Nyberg, a computer scientist at Carnegie Mellon University, is collaborating with I.B.M. on research to devise computing systems capable of answering questions that are not limited to specific topics. The real difficulty, Dr. Nyberg said, is not searching a database but getting the computer to understand what it should be searching for.
The system must be able to deal with analogies, puns, double entendres and relationships like size and location, all at lightning speed.
In a demonstration match here at the I.B.M. laboratory against two researchers recently, Watson appeared to be both aggressive and competent, but also made the occasional puzzling blunder. |
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
| Posted: Wed Mar 25, 2009 11:45 pm Post subject: |
|
|
A glimpse of the next International Supercomputing Conference in June - from talks to exhibits to technologies:
http://www.hpcwire.com/features/Bigger-and-Better-ISC-Moves-to-Hamburg-41842587.html?viewAll=y
| Quote: | HPCwire: The 33rd TOP500 list will be unveiled in Hamburg. Can you tell yet whether there will be any big surprises?
Prof. Meuer: At last year's conference, the 31st edition of the TOP500 list was topped by the first-ever petaflop/s machine, the IBM supercomputer installed at Los Alamos National Laboratory and named "Roadrunner," which carried out more than 10^15 floating point operations per second running the Linpack benchmark. In the 32nd list that we released at the SC08 conference in Austin last November there appeared a second petaflop/s machine -- the Cray XT5 "Jaguar" at Oak Ridge National Laboratory, which placed just behind Roadrunner at the top of the list. I expect that Jaguar will top the next list, and that Roadrunner will lose its top ranking. I've seen several indicators pointing to this change, so it won't be a big surprise if it happens. But to know for sure, we will just have to wait for the ISC'09 opening session. |
|
|
| Back to top |
|
|
rffrydr
Moderator
Joined: 30 Oct 2005
Posts: 11518
Location: Sunny California
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
| Posted: Sat Mar 07, 2009 12:09 am Post subject: |
|
|
A real-world adoption of the Nvidia Tesla GPUs in place of traditional CPUs:
http://www.finextra.com/fullstory.asp?id=19711
| Quote: | The new platform replaces more than 500 traditional CPU cores consuming 25 kW, allowing a 100 fold increase in the amount of calculation achieved per Watt. The overall reduction of response times is a factor of 15, contributing to a reduction in total cost of operation.
BNP Paribas CIB says overall electricity consumption will be reduced by a factor of 190, significantly cutting the bank's environmental impact.
Stéphane Tyc, head, GECD quantitative research, BNP Paribas CIB, says: "We are extremely pleased with these performances, which significantly exceed our initial expectations. We hope to transfer more calculations to the GPU architecture in the near future." |
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
| Posted: Mon Mar 02, 2009 12:50 am Post subject: |
|
|
The "Earth Simulator" - the Japanese pride for more than two years - now just a smaller version of itself, even after a major upgrade:
http://mdn.mainichi.jp/mdnnews/news/20090228p2a00m0na026000c.html
| Quote: | | The old Earth Simulator was the world fastest computer for 2 1/2 years after it was switched on in 2002. As of last November it was ranked 73rd, and fifth in Japan. The upgrades are expected to make it the fastest in Japan, and about the 18th fastest in the world.The old Earth Simulator was the world fastest computer for 2 1/2 years after it was switched on in 2002. As of last November it was ranked 73rd, and fifth in Japan. The upgrades are expected to make it the fastest in Japan, and about the 18th fastest in the world. |
|
|
| Back to top |
|
|
rffrydr
Moderator
Joined: 30 Oct 2005
Posts: 11518
Location: Sunny California
|
| Posted: Fri Feb 13, 2009 9:28 am Post subject: |
|
|
Computers are still computers, super or no:
_________________
Today is the Tomorrow you worried about Yesterday! |
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
| Posted: Mon Feb 02, 2009 9:28 pm Post subject: |
|
|
A clearer view of the supercomputing roadmap over the next few years, as IBM commits to deploying its next-generation BlueGene systems, "Sequoia" by 2012:
http://www.investors.com/editorial/IBDArticles.asp?artsec=17&issue=20090202
| Quote: | A prototype supercomputer called Dawn is set for delivery in the next few months at Lawrence Livermore National Laboratory in California. Dawn, capable of 500 teraflops, or a trillion calculations per second, will help lay the foundation to build Sequoia's applications.
Sequoia, set to go online in 2011, will run at 20 petaflops, or a quadrillion floating operations per second. That's 15 times faster than today's fastest computers — more powerful than the world's top 500 supercomputers combined.
Twenty petaflops equals 20,000 trillion calculations per second; one petaflop equals 150,000 calculations per second for every person on earth.
Sequoia will calculate in a single day what would take 120 billion people armed with calculators nearly 50 years.
In practical terms, Sequoia will let forecasters predict local weather events that affect areas 100 meters to one kilometer away, down from their current 10-kilometer range. Local forecasters might be able to accurately map the path of notoriously unpredictable tornadoes.
Future government uses could include study of the human genome, climate change and new sources of energy.
Companies will be able to buy time on the full Sequoia system according to their computing needs.
The system should benefit many businesses that deal with massive amounts of data but don't want to buy and manage their own supercomputer, says Srini Chari, a tech industry analyst and consultant with Cabot Partners Group.
That might include auto and aerospace firms, oil and gas exploration companies, and drug developers.
The new supercomputer will pack a much bigger punch into a smaller space than current versions of BlueGene, using far less space and energy , Chari says.
"The problem until now has been the enormous physical space needed to house supercomputers," Chari said. "That has resulted in an enormous cost for power and cooling, which Sequoia will address."
Sequoia will run on 1.6 million processors housed in 96 racks, each the size of a refrigerator. The entire system will take up just 3,422 square feet, much less space than older supercomputers.
Just one Sequoia rack, filling just 30 square feet, could equal the speed of the five or 10 fastest supercomputers in the world today, says Dave Turek, head of IBM's supercomputing unit.
"This advance makes really powerful computing eligible to fall into the hands of many industrial companies, as well as academic and government users who are the traditional early adopters of supercomputers," Turek said. "The energy consumption component of Sequoia is extraordinarily innovative."
Sequoia will complete 3,000 calculations per watt of energy used, compared with just 200 calculations per watt today, Turek says. |
|
|
| Back to top |
|
|
HenryTo
Site Admin
Joined: 06 Aug 2004
Posts: 9333
Location: Houston, Texas & Los Angeles, California
|
| Posted: Sat Jan 24, 2009 3:19 am Post subject: |
|
|
More on the Nvidia Tesla desktop. The thing to watch out for is (the GPU's) compatibility with today's software, such as Mathematica and Matlab. To that end, both Nvidia and Microsoft are working to solve these problems - literally bringing supercomputing capability to the masses:
http://www.designnews.com/article/162620-Supercomputing_Hits_the_Desktop.php
| Quote: | Snell says one of the longstanding obstacles for a hybrid approach is the issue of programmability since software has to be rewritten to take advantage of the new parallel architecture. "At some level, software needs to contain instructions that tell the system to execute a given calculation on a GPU rather than a CPU," he says. "The approach can be made more straightforward, but it doesn't eliminate the step that software has to be rewritten."
NVIDIA tackles that problem with CUDA, its C++-like programming environment used to tune applications to the GPU's parallel computing architecture. CUDA has been widely embraced among leading software vendors, including Mathematica and National Instruments, giving NVIDIA's GPU hybrid a jump on competing accelerator models, Snell says. However, CUDA has limitations in that it ties developers to the NVIDIA platform, according to Microsoft's Ryan Waite, product unit manager for Windows HPC Server. What the market requires, Waite says, is a general-purpose language for GPU programming that works across several platforms. To that end, he says Microsoft is planning a set of extensions to its Direct X language to enable developers to tap into the parallel processing engines in GPUs.
In addition to Microsoft's efforts, The Khronos Group has formed an industry working group to create royalty-free, open standards for programming parallel computing across GPUs and CPUs. The so-called Compute Working Group includes participation from such leaders as IBM, Apple, ARM, AMD, NVIDIA, Freescale and others. Apple has proposed the Open Computing Language (OpenCL) specification as a basis for the standard and the group is soliciting companies to provide input.
There are other software-related advances helping to propel HPC's march onto the desktop, according to Microsoft. Windows HPC Server, particularly the latest 2008 version, has done its part to help make these very complex systems easier to use and manage, especially for engineers and scientists who aren't necessarily versed in computer administration tasks, according to Waite. "If you're an engineer or scientist using Mathlab, you want to stay in that application," he says. "A personal supercomputer supports this idea instead of having centralized resources that only a limited number of people have access to and which require a lot of care and feeding."
John Stone, a senior research programmer at the University of Illinois at Urbana-Champaign, has long been a user of HPC clusters, but admits a personal supercomputer brings a level of convenience not seen with traditional systems. Stone, part of the university's Theoretical and Computational Biophysics Group, taps HPC technology to create a mathematical model that simulates the behavior of biological molecules at an atomic level - something that's not visible even through highly sophisticated microscopes. A couple years ago, running these highly complex calculations meant tapping some far-off supercomputer for days. Last year, Stone's group began working with CUDA and a custom-built hybrid CPU/GPU system and was able to perform the same work on a single GPU machine in their own lab in only 27 minutes. Stone says early tests show the new Tesla hybrid platform drastically improves even that level of performance.
"Being able to do this work locally without having to transfer files to a remote machine managed by someone else eliminates hassles and is a big time saver," he says. "The same reason why you'd want a personal computer in the first place now applies to a problem that previously required a small cluster or supercomputer." |
|
|
| Back to top |
|
|
Please log in to view without the ad banners |
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB
|
FAQ
Search
Memberlist
Usergroups
Statistics
Register
Profile
Log in to check your private messages
Log in
