Archive for the ‘High Performance Computing’ Category.

Extreme Supercomputing with Zettaflops

Zettaflop/s mainframes will be a million times faster than a petaflop/s supercomputer. Currently the quickest machine in the world stands at 8 petaflops. A zettaflop/s device could perform a sextillion floating point operations per second. That figure is a one followed by 21 zeroes. Talking about this type of calculating horsepower is really on the fringes of respectability. The pace of Moore’s Law is facing serious challenges. It is all too easy to create a bunch of graphs showing the exponential growth of transistors continuing until transcendence. However when these sorts of trends end, it is often abrupt. CPU clock speed swiftly jumped from kilohertz to megahertz and then finally stopped increasing at around 3-5 gigahertz. Now it seems that scientists are struggling to double the amount of transistors while keeping the power budget the same. They are using a new metric, which is the performance per watt. A recent publication put out by EPFL called Zetta covers ways to obtain and take advantage of upgraded HPC devices.

Read More »

European Exascale Software Initiative

Exaflop/s supercomputers will be able to perform a quintillion floating point operations each second. Developing programs that can best take advantage of this muscle is a daunting challenge. The machines will have millions of chips. As Moore’s law begins to stall, the only way to boost performance is by a massively parallel design. Each individual CPU’s clock speed will remain mostly the same, while the amount of cores per processor may increase. Trying to execute a task by breaking it up into numerous parts is not easy. A meeting about the exascale software initiative has recently discussed progress towards utilizing this computing capacity. Europe has several strengths in the arena of HPC (see technological challenges PDF and ESSI vision).  Read More »

Europe Seeks Exascale High Performance Computing

Attaining exascale performance levels is the dream of many nations. It seems easy for people to discount Europe as a player in extreme supercomputing. Over the past several decades, Japan and the United States have battled it out for the top supercomputer status. Recently China has also managed to take the crown for a brief period of several months. Europe has not been able to surpass any of these countries for a long time. Perhaps in the future this will change. The European Union allows a number of different countries to collaborate with one another. The EU has a population that is greater than that of the US and a very high gross domestic product per capita. In 2010, their total GDP was $ 15.17 trillion dollars compared to 14.65 trillion for the US (according to the IMF). Since the Eastern European economies have increasingly favored more capitalistic policies, the aggregate GDP can still go even higher (assuming they can overcome present challenges). France currently has the ninth most powerful (debuting at 6th place) mainframe in the world that has a sustained performance of one petaflop/s and a peak of 1.25 petaflops. Out of the top 500 in June 2011, Germany has 30 supercomputers with a sum speed of 3.242 petaflops. This is enough to put it in fourth place in terms of aggregate performance, behind the US, Japan and China. After Germany comes France with 25 machines and the United Kingdom with 27. Those three countries alone mean the European Union has at least 82 machines. They are still contenders for the race towards one quadrillion operations per second. Read More »

20 Petaflops Titan Supercomputer Summit

In August, the Oak Ridge Leadership Computing Facility hosted a summit to discuss progress towards 20 quadrillion floating point operations per second.  Currently the top mainframe in the world runs at around 8 petaflops, so this newer one would be over twice as quick as that record holder.  The machine is expected to be inaugurated in the year 2012.  This ultra-fast number cruncher appears to be a sure bet, although IBM had to recently cancel plans for constructing a different supercomputer.  Titan will use a Gemini XE interconnect, a Nvidia Tesla X2090 processor running at 665 gigaflops, an AMD Opteron 6200 Interlagos 16 core CPU and a Tesla X090 memory with 6 gigabytes GDDR5 capacity.  Recent developments may enable Titan to become the fastest device in the world.  That’s assuming China or Japan don’t have something new up their sleeves to challenge it.  See the file “introduction and timeline“ (PDF) for an overview on the project.   Read More »

Exascale Supercomputing Challenges

The first 20 petaflop/s supercomputers should be in service by 2012 and after that comes a machine in the 100 petaflop/s range (2015).  Scientists are moderately optimistic that exaflop/s (1000 petaflop/s) mainframes can be constructed by 2018-2020.  However, are some of these expectations just plain irrational?  A workshop discussed the hardware and software upgrades that will be needed to best exploit this kind of computational muscle.   Read More »

Future Exaflop Supercomputer Technology Trends

The race to assemble supercomputers that are over a hundred times faster than the current record holder is heating up. Japan’s K computer is presently the top machine in the world, benchmarked at 8 petaflops. One thousand petaflop/s (1 exaflop/s) systems are expected to become operational by the end of this decade. There are enormous challenges to constructing these behemoths that may be difficult to circumvent. However, the promise of this much raw power has numerous governments working intensely to make it a reality. The Scientific Discovery through Advanced Computing Program (SciDAC) had a meeting to discuss the engineering obstacles to obtaining this level of brute computational strength. Read More »