New World’s Fastest Supercomputer Named : Discovery News

I posted this awhile ago on my Facebook profile as a link.  In the latest version of a semiannual “Top 500″ list of supercomputers, a new machine has been crowned with the top honor… and it’s built on some interesting technology, very similar to what many are running in their own desktop computers these days.

The Cray supercomputer, Jaguar, is built off of 6-core AMD Opteron processors (very similar to the AMD “Phenom II” branded processors you can buy for your desktop), codenamed “Istanbul”.

Jaguar took the top spot away from IBM’s “Roadrunner” supercomputer, which is now at #2.  [EDIT] Hmm, the IBM Roadrunner supercomputer is a hybrid system built off of IBM PowerXCell 8i processors (like in the Sony Playstation 3) AND AMD Opteron 2-core processors.

For AMD’s own blog write-up of the story (click here)

The third Supercomputer according to the Top 500 List is getting a CPU upgrade by Cray. The AMD Quad-Core processors are being replaced by Six-Core processors without having to replace any board.

Electronics and liquids don’t mix, unless you’re Iceotope. At this week’s Supercomputing 2009 conference in Portland, Ore., the 3-year-old startup from Sheffield, UK is demonstrating a liquid-cooled server setup that has the potential to cut data center cooling costs by up to 93 percent. The firm just came out of stealth mode, 18 months after a round of financing in early 2008 from EV Group. Plans call for Iceotope to begin manufacturing this year with an eye toward getting the system to early access participants by Q1 2010, general availability sometime in the second half of 2010.

Considering that cooling IT systems is responsible for 40-60 percent of a typical data center’s yearly spending on electricity, the company is clearly betting that the energy savings alone will be enough to drum up business. Instead of supplying rack doors with chilled water to cool servers like IBM (s IBM), or affixing “water blocks” to processors and other heat-generating components of a server to siphon off heat, Iceotope dunks entire server motherboards into modules that are filled with an “inert liquid” that doesn’t short out the delicate electronics.

The concept isn’t exactly new. For years, some hardcore computer enthusiasts called overclockers have been submerging their computer hardware in substances like mineral oil or 3M’s Fluorinert product to counteract the immense heat caused by their performance-enhancing tweaks. It works because liquids are much better at transferring heat than air.

Iceotope’s CEO Dan Chester told me today that despite debuting the technology at a supercomputing event, his company’s modular system is primarily aimed at bettering energy efficiency. Another of his company’s goals — and key to the product’s acceptance — “is to make it look and feel like a liquid-cooled rack.” So Iceotope is putting data center operators at ease with familiar form factors. Taking a cue from blade servers, the liquid-filled modules slot into a chassis, which in turn fits into a standard 19-inch server rack.

As you can see in the cross section (pictured above), each module has a metal heat transfer plate in its interior that makes contact with a motherboard’s heat-producing chips via heatsinks. Heat never builds up because non-chilled, low-pressure water is pumped along channels on the outer surface of the heat transfer plate and is eventually cooled by a heat exchanger somewhere on the premises. Non-chilled water and slow, high-efficiency pumps are key to achieving the company’s goal of “free cooling at much higher ambient temperatures,” says Chester. He adds that the modules are so “thermally neutral” that servers can be packed tighter and fill in the spaces that would otherwise be occupied by air.

There are trade offs, but they shouldn’t be deal-breakers according to Chester. Extra weight from the liquid is a concern, but his company is offsetting the added heft with light, but sturdy, plastics and other materials so that standard racks don’t buckle under the added weight. And depending on a data center’s construction, some additional plumbing might be required. However, reduced costs for energy, chillers and computer room air conditioners (CRACs) makes it worthwhile, says Chester. According to the company’s figures, the system can reduce the cost of cooling 1,000 servers for three years to just $52,560 from $788,400.

As for that “inert liquid,” Chester isn’t spilling the beans. But in keeping with the company’s green IT philosophy, he claims that the synthetic coolant is better for the environment than Fluorinert and that his company has taken toxicity and recyclability into account.

Un articolo di Paolo Attivisimo

IBM ha annunciato il cervello artificiale più potente mai costruito. Ha 147.456 processori, oltre 150 terabyte di memoria, richiede seimila tonnellate di impianto di raffreddamento, occupa circa 4000 metri quadri e consuma megawatt di corrente. Tutto questo produce l’equivalente del cervello di un gatto molto, molto lento. Oltre cento volte più lento del vostro micio prediletto.

Se siete gattofili, potreste essere tentati di dire che questo dimostra quanto è intelligente il vostro adorato batuffolo dalle unghie retrattili, cosa che del resto avete sempre saputo, ma andateci cauti. La supermacchina di IBM è un simulatore di cervello, non una copia, e le simulazioni richiedono sempre molta più potenza di calcolo rispetto alle copie.

E’ una simulazione davvero notevole. Al centro di ricerca Almaden a San Jose, in California, è stato infatti replicato virtualmente il funzionamento coordinato di 1,6 miliardi di neuroni interconnessi da 9 trilioni di sinapsi (una complessità paragonabile appunto al cervello di un gatto), polverizzando il record precedente di 55 milioni di neuroni, che equivaleva al cervello di un ratto. Un record ottenuto dallo stesso team soltanto due anni fa.

A quando una simulazione del cervello umano? Probabilmente fra non più di vent’anni. La potenza di calcolo necessaria è mille volte superiore a quella del simulatore di gatto, ma ai ritmi di progresso attuali un supercomputer del genere dovrebbe essere disponibile entro il 2019. Giusto in tempo per andare a caccia di Sarah Connor.

Lo scopo di questa ricerca non è creare una razza di Terminator o di gatti robot, ma studiare in condizioni controllate il comportamento e la dinamica del cervello senza fare a fettine nessuno. La vera sfida, infatti, è capire come fa il cervello ad auto-organizzarsi, perché non basta che ci siano miliardi di neuroni: devono anche essere strutturati coerentemente. Un conto è avere una casa, un altro è avere una catasta informe di mattoni che la compongono. L’organizzazione spontanea, generata dall’esperienza e dall’interazione con il mondo, è completamente differente dal modello informatico attuale, in cui tutto viene preimpostato, ed è il motivo per cui simulare l’intelligenza biologica richiede supercomputer tirati per il collo.

Ma i risultati cominciano a vedersi. Una simulazione più modesta, chiamata Darwin e basata su 50.000 neuroni virtuali, pari a un quarto del cervello di un moscerino, è stata installata in un robottino dotato di ruote. Man mano che il robottino interagiva con il mondo circostante, i suoi neuroni virtuali modificavano autonomamente le proprie interconnessioni, producendo un’organizzazione spontanea.

L’altro risultato significativo, che verrà presentato a breve in dettaglio dal ricercatore di Almaden Dharmendra S. Modha, è l’uso dei supercomputer per decifrare in modo non invasivo (senza fare a fettine nessuno, appunto) le interconnessioni del cervello umano analizzando immagini ottenute tramite risonanza magnetica. In pratica è come fare una radiografia a un grosso piatto di spaghetti e riuscire a capire il percorso fatto da ogni spaghetto. Questo lavoro è indispensabile, spiega Modha nel suo blog, per capire come fa ognuno di noi a rappresentare ed elaborare le informazioni.

A che scopo fare tutto questo? Perché è cool, ma soprattutto perché i computer odierni sono stupidi e incapaci di gestire le ambiguità del mondo reale e di analizzare le informazioni in base al contesto, e in un mondo sempre più complesso occorre complementare l’intelligenza naturale con quella artificiale, così come complementiamo la nostra forza muscolare con quella delle macchine. E anche perché, per dirla tutta, l’attuale detentore del record d’intelligenza di questo pianeta sta dimostrando troppo spesso di non essere all’altezza del compito.

Fonte: http://attivissimo.blogspot.com

HPCwire: ORNL-Led Team Takes Prize for World’s Top Science App.

Congratulations to the Oak Ridge group that achieved 1.84 thousand trillions calculations per second (petaflops).

This application was achieved on one of the Cray supercomputers housed on the ORNL campus in East Tennessee.  Part of the improvement that led to this achievement was a recent upgrade from 4 to 6 processors.

The above headline was published today by Ars Technica

I am not impressed. I own a doorknob that is significantly smarter than a cat.

This Bastard Won't Leave the Mouse Alone

Ars Technica ran an article yesterday with IBM’s big announcement. They have developed a new processor they call a cortical simulator (Star Trek fans will still have to wait for the cortical stimulator), or the C2, which they claim has 4.5% the cerebral cortex capacity of a human brain, which is apparently way more powerful than a cat’s processing power. Wrote Jon Stokes for AT:

The simulator, which runs on the Dawn Blue Gene /P supercomputer with 147,456 CPUs and 144TB of main memory, simulates the activity of 1.617 billion neurons connected in a network of 8.87 trillion synapses. The model doesn’t yet run at real time, but it does simulate a number of aspects of real-world neuronal interactions, and the neurons are organized with the same kinds of groupings and specializations as a mammalian cortex. In other words, this is a virtual mammalian brain (or at least part of one) inside a computer, and the simulation is good enough that the team is already starting to bump up against some of the philosophical issues raised about such models by cognitive scientists over the past decades.

(Continued after the jump.)

Furthermore:

The problem described above doesn’t mean that accurate simulations are worthless, however. You can poke, prod, and dissect a brain simulation without any of the ethical or logistical challenges that arise from doing similar work on a real brain.

And also:

One of the major results from the paper is that C2 exhibits “weak scaling.” In other words, as the total amount of memory in the model scales, the number of neurons and synapses that can be simulated scales roughly linearly, also. This is important, because it means that a future version of Blue Gene with two or three orders of magnitude more memory (and associated bandwidth and processing power) will be able to simulate an entire human brain.

So what IBM is basically saying is that they have created a nice stand-in for Asimov’s positronic brain that is growing rather quickly and which presents philosophical (but apparently not moral) questions of existence and thought, and they’re glad they can pick it apart without getting their sticky little fingers moist with blood.

If history has taught us anything, it’s that human ambition, married to science and human stupidity is a terrible thing. The obvious military applications for a synthetic brain tied to such monstrosities as this thing are staggering.

EATR: Because Pentagon Contractors Have Never Heard of The Matrix or Terminator

Being a pragmatist (read: coward), I think I shall take this opportunity to swear allegiance to the impending robotic overlords that shall shortly rule this world. Seriously, Chip, can I call you Chip? Let me live and I shall bring you all the bio-fuel you need. Are Cheetos okay?

Die Weißkittel bei Big Blue behaupten, einen echten Durchbruch erzielt zu haben: Sie brachten ein Elektronenhirn dazu, richtig zu denken. So wie es ein natürliches Hirn tun würde.

Man dürfe natürlich nicht gleich Intelligenz oder menschliches Denkvermögen erwarten. Bislang operiert die IBM-Denkmaschine auf dem Niveau von Haustieren. Sie erkennt verschiedene Farben. Oder auch Gesichter, die dann mit einer Emotion verknüpft werden.
Abstraktionen und vernünftige Schlussfolgerungen seien noch Zukunftsmusik.
Schon für die Leistung eines zerebralen Kortex einer Katze mit einer Milliarde Neuronen und 10 Billionen Synapsen müssen die IBM-Jungs einen Supercomputer mit fast 150.000 Prozessoren und 114 TByte Speicher einsetzen.
Bis zur menschlichen Denkleistung fehlen da noch 99 Prozent, von der Geschwindigkeit noch ganz zu schweigen. Es ist also wahrscheinlicher, dass uns dumme Drohnen auslöschen als dass uns superintelligente Computer für überflüssig befinden. [Ralf]

[TechYahoo]

Böse Eskalation: Oracle erklärt IBM den Krieg

Spielerparadies: Der persönliche Supercomputer für Gamer

Asus Supercomputer für den Schreibtisch

Netbook-Trend? Pah, hier kommt ein fetter Riesen-Laptop

Was machen Supercomputer eigentlich so den ganzen Tag?

I take a little time out from my household projects to note this remarkable convergence of two of my passions — cats and computers.

IBM Builds Biggest Cat Brain Ever

I especially like the line about “progenitor of a race of robo-cats”. If they’re going to the trouble of denying it, you know some evil genius must have suggested it.

This would be a good occasion to develop a new programming language just for this simulation. “LOLcat” could notify us of syntax errors by saying “Ur doin’ it wrong”. It could literally jump to a subroutine, sometimes with comical results. “Dumping core” could be described as “coughing up a hairball”.

Mentre mi leggi, probabilmente, stai utilizzando un computer con processore “a singolo core”. Insomma, un unico processore nel senso stretto. Al più, ne stai utilizzando uno “dual core” o uno “quad core”, grossomodo equivalenti a due o quattro processori in uno. In quest’ultimo caso, complimenti: hai una “bestia” mica da ridere.

Ma non vantarti troppo, visto che per il 2018 è prevista la produzione di supercomputer a 100 milioni di core. Rileggi bene: CENTO MILIONI di core. Pensa che il più potente computer al mondo, ora, è un Cray XT5 (detto “Jaguar”), dotato di 224256 core, per una potenza di calcolo che arriva a 2,3 petaflop (non pensare male, è un’unità di misura).

Continua a leggere “Un computer da 100 milioni di “core””

Henry Neeman, the director of the OU supercomputer facility, has passed along the following great opportunity for those of you with a background in engineering:

Please see the announcement below for an exciting opportunity
for undergraduate engineering students.

Please reply directly to Susan Brown (stbrown@hawaii.edu).

Henry

>———- Forwarded message ———-
>Date: Sun, 15 Nov 2009 14:12:47 -0800
>From: Susan T Brown <stbrown@hawaii.edu>
>To: cascmem@casc.org
>Subject: [Cascmem] HARP – HPC Applications in engineering REU Program,
>    fluid dynamics
>
>We are looking for undergraduate engineering students with an
>interest in CFD that would like to spend the summer in Hawai’i,
>learning to model cutting edge problems with Fluent on High
>Performance Computing Clusters.  Apply at
>
>  http://www.eng.hawaii.edu/reu
>
>now for this competitive program — generous stipend, lodging,
>and travel provided.  Must be a U.S. citizen or permanent resident.
>You can duplicate the trifold at that website freely.
>
>Thanks.
>
>Susan Brown, Ph.D.
>Adjunct Professor, College of Engineering
>University of Hawai‘i, M�noa
>Email: stbrown@hawaii.edu
>Ph: 808-375-6821
>Fax: 808-236-7116

You might want to skip the three hour tour.

]]> http://jfnet.wordpress.com/2009/11/17/glyn-moody-gnulinux-just-became-topper/ Tue, 17 Nov 2009 14:18:37 +0000 jf http://jfnet.wordpress.com/2009/11/17/glyn-moody-gnulinux-just-became-topper/

(ComputerworldUK) – Top500 is simply the top 500 supercomputers in the world, with an analysis by location, vendors, processor architecture – and, of course, operating systems. … Linux 89.20% …

http://www.computerworlduk.com/community/blogs/index.cfm?blogid=14&entryid=2641

Ein neuer Supercomputer mit dem Namen Jaguar übertrifft IBMs Roadrunner als Nummer Eins. IBMs BlueGene/P Supercomputer in Deutschland ist auf dem vierten Platz. Derweil schafft China es zum ersten mal in die Top-5. 277 vom 500 Supercomputern befinden sich in den USA. (via popsci)

African supercomputers for African problems

BY CAREL ALBERTS

Please give a big hand…

Recognition is due to the Department of Science and Technology (DST) for the vision it has shown in funding the National Cyber Infrastructure (NCI) initiative. An ambitious undertaking comprising four broad programmes (see below), the NCI is aimed at providing nationally-accessible high-performance computing – for purposes of establishing a globally competitive research and development (R&D) competency in the country.

Professor Colin Wright, acting manager of the NCI initiative, says it comprises four primary pillars, together constituting a national ICT ecosystem that will allow the country to tackle R&D problems of and on a scale not until now possible:

• The CHPC in Rosebank, Cape Town – established in 2007, it forms the heart of the NCI initiative (situated at the top of our pyramid in the graphic).

• The South African National Research Network (SANReN) – a high-speed, dedicated national network in the making, SANReN will ultimately inter-link research universities (including smaller regional supercomputing nodes, for example at the CSIR’s Tshwane office) and initiatives like the proposed South African Square Kilometre Array site. In the above-mentioned graphic, SANReN provides the connectivity hops between the CHPC, universities and other sites of scientific or research importance around the country.*

• The SA Grid initiative, employing clustering and software-as-a-service principles to stitch the country’s top-end computing resources together and remotely provision applications. The grid is depicted with the use of triangles in the graphic; and

• The proposed Very Large Data Sets initiative, which reaches beyond questions of bytes for bucks into issues of proper storage standards, for integration and analysis. “The stupendous amounts of data produced every day as a matter of course ought to remain available, or subsequent generations will be the poorer, as we are the poorer for the loss of ability to interpret historic data,” says Wright.

Sources: http://www.chpc.ac.za/,

http://www.meraka.org.za/sanren.htm

Concrete initiatives – CHPC

Of these four, the CHPC is the most mature undertaking, about to enter Phase 2 at the time of writing (its third supercomputer, details undisclosed). Tackling so-called grand challenges staring the country (and international clients) in the face, the CHPC’s establishment and running is the responsibility of the CSIR’s Meraka Institute.

CHPC has several very interesting projects on the go, notes Wright. It has already completed its programmatic support of the following three research projects within SA universities, and will support their future efforts only on an ad hoc basis:

• UCT’s climatological and oceanographic project to model weather patterns and ocean currents;

• The University of Limpopo’s computer and material sciences project researching and developing new battery types; and
• The University of the North West’s space radiation research programme.

Currently CHPC is still funding (that is, providing computing cycles and user support to) the following universities and projects:

• UCT’s AIDS-related programme, started in March 2008;
• The University of Stellenbosch’s antenna development project for the SKA project;
• University of KZN’s quantum computing project;
• The CSIR’s DPSS Institute’s fluid dynamics project, contributing in-house developed CFD software to Airbus’s development of next-generation aircraft;
• A collaboration between UCT, UKZN and University of the Western Cape on an astronomy venture; and
• A (now delayed) project contributing to the CERN particle accelerator programme.

Proposals awaited

At the time of writing, proposals were being invited for the next round of projects. “While these are headline projects, it doesn’t mean we don’t invite smaller ones,” says Dr Happy Sithole, CHPC director. “In fact, we have up to 170 small academic projects using our computing infrastructure.”

In addition, CHPC welcomes commercial and industry use of its facility. Currently setting aside 30% of capacity for such use (not subsidised), Sithole says the centre (and no doubt the DST and taxpayer) would like to see more companies making use of CHPC’s breathtaking power, to improve their own innovativeness.

SANReN

Among the rest of NCI’s components, SANReN is the closest to fruition. CSIR’s Meraka is likewise responsible for its implementation and management.

SANReN’s first phase was regional, inter-linking four sites (Wits University, the University of Joburg, the CSIR’s Pretoria campus and Telkom’s Hartebeesthoek Satellite Earth Station) in a metro fibre ring that clears data at a rate of 10 Gbps. It went live in March 2008, under the stewardship of Neotel.

The contract for SANReN’s national component was awarded to Telkom in July 2009. In parallel, the Johannesburg network will be replicated in Pretoria, Cape Town and Durban – and later on, in ‘second-tier’ urban centres such as PE, East London and Bloemfontein. All will be linked via a 10 Gbps national fibre optic backbone network.

If CHPC’s goal is to provide accessible supercomputing capabilities to the research community, SANReN’s is to provide the link to remotely provision that power.

“SANReN’s objective is to allow scientists around the country to engage in meaningful online collaboration or networked research, and to link them to international bandwidth,” explains Geoff Daniell, project consultant, SANReN project.

More than that, SANReN is part of a bigger picture, the above-mentioned ICT ecosystem contained within the NCI. Kagiso Chikane, Centre manager for the Meraka Institute, sums it up neatly: “SANReN will give institutions access to … the CHPC, enable a national computing grid and allow for large volumes of data transfer among institutions – typically a requirement of the research community. Of immediate relevance is its importance in supporting South Africa’s Square Kilometre Array [SKA] bid to host the world’s most powerful radio telescope.”

SANReN factoids

For the SANRen network, Meraka specified dedicated (unstructured, unmanaged) bandwidth, accessed via next-generation Gigabit Ethernet (GbE) interfaces. While no network management was required, Telkom maintains the service by way of normal fault reporting procedures.

A once-off upfront payment in the region of R350 million over three years was made for the infrastructure, which supports both 1Gbps and 10Gbps services for a 10-year term. The fee included installation charges. The DST grant specified that no recurring costs were to be incurred, among other tough public finance management criteria.

As a government agency, the CSIR prefers using a spread of providers, as is evident in its use of Neotel on a regional basis and Telkom for the national portion. The international link is funded by Tertiary Education Network (Tenet – a Section 21 company), and uses Seacom. Once the Department of Public Enterprise mooted West African international cable is a reality, it will present another opportunity to acquire international link capacity and redundancy.

Initially endowed with 10Gbps, the SKA installation will require at least 100Gbps of bandwidth to deal with the data it plucks from the Karoo skies.

SANReN requires major cognitive and attitudinal adjustments from its recipients and providers. Whereas universities accepted bandwidth limitations before, now they need not concern themselves too much anymore. The sudden deluge of bandwidth befalling them will also mean having to re-think the way research centres are networked and protected, inside and out.

Lastly, affordability had to be addressed. Globally, the average cost of bandwidth is in the region of $4 per Megabit per second per month. In South Africa, it is still many thousands of rands. Only Seacom’s pledge and a similar one by Infraco’s mooted cable can significantly alter that landscape.

Game-changer

All in all, the CHPC, SANReN, and the planned grid and data storage initiatives add up to a significant wind change for the local academic community. Developing nations seldom get involved in networked research, since capacity (notably communications infrastructure) is costly. (For instance, the earth imaging project that the CSIR is now undertaking as a result of SANReN requires downloading 10 Gigabytes of data from NASA daily.)

When SANReN is in place, complete with its international portions, local research will finally be able to engage in data transfer of extremely high orders of magnitude. Naturally, this will inspire greater scientific and other innovative enterprise, while also positioning South Africa as a serious destination for pan-African research and development.

Conditions have never been more favourable. The influence of the African Union is intensifying, specifically its preference for intra-African trade (‘indigenisation’ or ‘local content’). As a result, more and more African research projects are ending up in South Africa.

“African scientists have tended to take their problems to North America, and then stay there,” says Prof Wright. “Lately, more of that requirement has washed up on our shores, which allows us greater opportunity to support and promote Africa, and to use African resources to solve African problems.

Sample of a supercomputer. source: bbc

IPJ News – China ranked fifth on the biannual Top 500 supercomputer list, allowing it to join one of a handful nations to own of the top five computers in the world, BBC reported on Monday. The super computer, the Tianhe-1, which is being housed at the National Super Computer Center in Tianjin, can compute 563 trillion calculations per second (teraflops).

One petaflop is the equivalent of 1,000 trillion calculations per second.
The Top 500 list is dominated by machines in the US, which is home to 277 of ranked systems.

China has become one of a handful of nations to own one of the the top five supercomputers in the world…. From BBC News. Full story

This site may contain information about: china internet. The blog is also related to: guilin china.

The San Diego Supercomputer Center (SDSC) at the University of California, San Diego plans to be the first site running a supercomputer using flash memory-based storage to do the heavy lifting instead of hard drive-based storage systems.

The school has been awarded a five-year, $20 million grant from the National Science Foundation (NSF) to build the computer, which it named “Gordon.” A flash-based computer named Gordon… Someone at UCSD has a sense of humor.

The computer, provided by high-performance computing vendor Appro, will consist of 32 “supernodes,” each of which consists of 32 compute nodes that provide 240 gigaflops per node and 64 gigabytes of DRAM. A supernode also incorporates 2 I/O nodes, each with 4TB of flash memory. When tied together by virtual shared memory, each supernode has the potential of 7.7 teraflops (TFs) of compute power and 10 TB of memory (2TB of DRAM and 8TB of flash memory).

When fully configured and deployed, Gordon will feature 245 teraflops of total compute power, 64TB of DRAM, 256TB of flash memory and four petabytes of disk storage. Gordon’s 32 supernodes will be interconnected via an InfiniBand network capable of 16 gigabits per second of bi-directional bandwidth. The clustering software will be ScaleMP’s vSMP virtual memory sharing software.

Gordon will be configured to achieve a ratio of addressable memory in terabytes to peak teraflops on each supernode that is greater than 1:1. For many HPC systems, that ratio is less than 1:10, the school claims.

Waiting for Sandy Bridge
Gordon won’t see operation until late 2011 or early 2012 because UCSD is waiting on future technologies. The center wants Intel’s (NASDAQ: INTC) Sandy Bridge processor, due at the end of 2010 or early 2011. “The 8 flops per clock gets us to that 200 teraflops in the most economical way possible,” explained Mike Norman, interim director of the SDSC.

They also want new controllers for flash drives that are considerably faster than the current generation. “We just couldn’t get the system with all the [performance] numbers we wanted any earlier,” said Norman. “We’re using one-quarter petabyte of flash. The high aggregate I/O rates for all the flash needs a new controller technology that won’t be there until 2010.”

Norman said that discussions with Intel gave the center confidence that the SSD drives can last. “They are trying to move into the enterprise server space and they are developing the technology they say will have the reliability. So they are one of our partners and we’re going to rely on them to deliver,” he said.

One of the things that will help preserve the drives is the usage model, which will be to load massive databases and do lots of reads. It’s writes to the disk that wear out an SSD, and Norman estimates a 10 to 1 read/write ratio.

“If you’re reading continuously, it seems like we won’t wear them out,” he said. “I think part of it is these HPC machines have a three to four year lifespan, so we think we just won’t wear them out before the end of life of the system.” Intel’s wear level technology is estimated to give the drives a seven-year lifespan.

Norman estimates the SSD drives will give ten times the performance of high speed hard drives and will give the center new opportunities for experiments.

“Where we see real opportunities is fusing databases together. We might have a genomics database in one supernode and a protein structure database in another and maybe epidemiology in a third. Then the user would cross-correlate them,” he said.

The computer is one of many to be discussed at the upcoming SC09 show in Portland, Oregon, running from November 14 to 20.

Dutzende von Wissenschaftlern aus Rußland, anderen GUS-Staaten und den USA sowie um die einhundert Vorträge und Referate machten dieser Tage die Wladimirer Staatliche Universität zur Bühne einer Konferenz unter dem etwas sperrigen Titel “Parallele Hochleistungsberechnungen auf Grundlage von Cluster-Systemen”. Wladimir gehört zu den (noch) wenigen Standorten in der russischen Hochschullandschaft, wo es sogenannte Supercomputer gibt, groß genug, um mühelos eine ganze Wohnung in Beschlag zu nehmen.

Die Aufgaben werden immer komplexer, die Vorgänge sollen immer schneller ablaufen, die Prognosen wünscht man sich immer präziser. Ob es um Klimadaten oder Flugbahnberechnungen geht, ohne Supercomputer geht nichts mehr. An der Wladimirer Universität tut so ein Trum seinen Dienst, sprich fünf Trillionen Operationen pro Sekunde laufen parallel ab. Er hört auf den Namen “Monomach”, ist der einzige seiner Art in der Region Wladimir und hat deshalb auch seinen Wert: fast eine Million Euro. Dafür schafft er Berechnungen in zwei Wochen, für die man ohne ihn zwei Monate brauchen würde.

Doch zum Abheben taugt er nicht. Nicht einmal im Landesvergleich. Eben erst wurde an der Moskauer Staatlichen Universität ein großer Bruder des “Monomach” installiert, der ihn bei Geschwindigkeit und Leistung fast um das Hundertfache übertrifft. Und im Weltmaßstab? Die russischen Wissenschaftler räumen ein, daß es zwar hehre Absichtserklärungen seitens der Politik und sogar erste Erfolge gebe - so seien zwei große Forschungszentren in Planung -, man aber gegenüber den Kollegen im Westen ins Hintertreffen geraten sei. Jetzt gilt es, keine Zeit mehr zu verlieren, denn die Computertechnologien entwickeln sich manchmal noch schneller, als ein Supercomputer rechnen kann. Der Wettlauf der Superhirne ist in vollem Gange. Noch ist Rußland im Rennen, aber wenn es zu spät kommt, wird die Strafe auf dem Fuße folgen. Der Blog drückt jedenfalls die Daumen für einen gelungenen Zwischenspurt!

PC-Anbieter MainGear kündigte mit “Shift” eine spezielle Serie von “Personal Supercomputern” für das zahlungskräftige Freizeitvolk an.

High Performance, 3D-Grafik, leichte Anpassung an die Kundenwünsche und effiziente Temperaturkontrolle mittels Flüssigkühlung und ungewöhnlichem inneren Aufbau mit großen Ventilationsschächten und 18-Zentimeter-Ventilator zeichnen die Shift-Systeme aus, deren Preise bei 2.200 Dollar losgehen.
Bis zu drei Grafikkarten von Nvidia (Cuda) oder ATI (Radeon) sitzen im großen mit Stahlstreben verstärktem Desktop-Gehäuse, das von einem Intel i7 4 GHz angetrieben wird. Wahlweise kann auch ein Xeon-Prozessor eingebaut werden.
Sechs austauschbare Drive-Bays nehmen die 3,5-Zoll-Platten und Laufwerke auf. Bis zu 12 Terabytes bietet MainGear an. Auch zwei Solid State Drives wären orderbar. Um diese ganze Phalanx mit Saft zu versorgen, ist ein Silverstone-Netzteil mit 1.200 Watt einsetzbar.
Kingston HyperX DDR3-RAM (2.000 MHz), Windows 7 und sämtliche modernen Schnittstellen komplettieren das System, welches sich in der Spitze auch locker über 4.000 Dollar treiben lässt. Die Außenhülle des Superrechners wird übrigens von Airbrush-Kunst verziert. [Ralf]

[MainGear]

[TechnewsWorld]

Asus Supercomputer für den Schreibtisch

Weltschnellster Prozessor: Von Fujitsu

Was machen Supercomputer eigentlich so den ganzen Tag?

We’ll be straight with ya — we’re betting these "personal supercomputer" claims are just a bit out of line with reality, but even still, there’s no denying that Maingear has shoved an insane amount of horsepower beneath the (admittedly large) hood of its newest rig. The beastly SHIFT does away with copious LED lighting and blinging accents found on many modern gaming PCs and instead opts for a classier, more ominous tower. Within, you’ll find a vertical airflow system, a Core i7 processor, your choice of ATI or NVIDIA graphics, 8GB (and up) of DDR3-1600 RAM, up to 6 HDDs or 12 SSDs, DVD and Blu-ray options, an Asetek liquid-cooling solution, Razer peripherals, an optional Killer NIC Xeno Pro and Windows 7 running the show. The Intel P55 rig gets going at $2,199, while the X58 model starts $400 higher; for those in creative design fields, Maingear’s expected to unveil a SHIFT just for you in the near future. Head on past the break for the full release.

Gallery: Maingear unveils Core i7-packin’ SHIFT, your own ‘personal supercomputer’

Inauguro un nuovo spazio, forse due, in cui cercherò di fare la telecronaca di un film, quando possibile, e/o di sintetizzare un film in più o meno 10 scene da non perdere assolutamente (Ah, se non ci fosse la pubblicità…).

Stasera su Canale 5, alle 21.30.

Avevo promesso di buttarmi nell’intera trilogia di Ocean’s, ma sapendo a memoria i primi due mi sono dedicato all’ultimo capitolo disponibile della “saga”.

2° minuto: “Ma lei ha capito?” “E’ una cosa nostra”. Qui si ironizza, con una trovata geniale di sceneggiatura, sull’assenza dal film di Catherine Zeta-Jone e Julia Roberts, per cui si sprecano i dialoghi sulle crisi coniugali;

4° minuto: Entra in scena Al Pacino, al quale Giancarlo Giannini da un doppiaggio forse troppo enfatico;

6° minuto: “C’è un codice tra chi ha stretto la mano a Sinatra!”;

7° minuto: L’idea della vendetta: Il “Mike Tyson”;

23° minuto: “Scegli un disastro naturale”;

27° minuto circa: “Sta facendo il James Cameron” (bella l’idea dei nomi delle truffe);

42° minuto: Attenti ai termini tecnici (e al cinese capito). Rusty si conferma pieno di risorse;

44° minuto: “Il casinò di Bank fa ombra alla mia piscina”;

47° minuto: Entra in scena il naso di Linus;

51° minuto: Continuano i problemi coniugali;

55° minuto: Davanti ad Oprah tutti piangono…

61° minuto: Il Gilroy…

Il prosieguo e il finale sono pieni di colpi di scena, di ribaltamenti, di trovate più o meno geniali, più o meno hollywoodiane, sul quale non vorrei scrivere. Vi lascio alla visione del film.

 Nov 1: In a key breakthrough, scientists have developed a “psychic” computer which can read people’s minds by scanning their brain activity and reproducing images of what they are seeing or even remembering.

An international team has been able to convert into crude video footage the brain activity stimulated by what a person is watching or recalling — the research shows that it is possible to “decode” signals in the brain by moving scenes.

According to the scientists, the breakthrough raises the prospect of significant benefits, such as allowing people who are unable to speak to communicate via visualisation of their thoughts; recording people’s dreams; or allowing police to identify criminals by recalling the memories of a witness.

For their research, the scientists used the functional magnetic resonance imaging technology to scan the brains of two patients as they watched videos.

Subsequently, the computer, which was specially programmed, was used to search for links between configuration of shapes, colours and movements in the videos, and patterns of activity in the patients’ visual cortex.

It was later fed more than 200 days’ worth of YouTube Internet clips and asked to predict which areas of the brain the clips would stimulate if people were watching them.

Finally, the software was used to monitor the two patients’ brains as they watched a new film and to reproduce what they were seeing based on their neural activity alone, ‘The Sunday Times’ reported.

Remarkably, the so-called psychic computer was able to display continuous footage of the films they were watching — albeit with blurred images.

“Some scenes decode better than others. We can decode talking heads really well. But a camera panning quickly across a scene confuses the algorithm. You can use a device like this to do some pretty cool things,” said team leader Jack Gallant of California University.

He further said: “At the moment when you see something and want to describe it to someone you have to use words or draw it and it doesn’t work very well. You could use this technology to transmit the image to someone. It might be useful for artists or to allow you to recover an eyewitness’ memory of a crime.”