> > Holy Chip! > Daniel Lyons, 01.30.06 > IBM's radical Cell processor, to debut in Sony's PlayStation 3, could reshape > entertainment and spark the next high-tech boom. > Later this year millions of homes will get a new supercomputer for the living > room. Or maybe the playroom. Sony's long-awaited PlayStation 3 game console, a > slender yet muscular machine the size of a DVD player, performs a > mind-boggling 2 trillion calculations per second. This kind of power, once > reserved for seismic exploration and nuclear-weapons design, will let > programmers create videogames that look as realistic as film. > > Some techies say PlayStation 3, which may debut by midyear and could end up in > 100 million homes in five years, will usher in the next microchip revolution. > The Sony system owes its prowess to a microprocessor called Cell, which was > cooked up by chip wizards at IBM (with help from Sony and Toshiba) at a cost > of $400 million over five years. The Cell chip, based on a design inspired by > supercomputers, runs at least ten times as fast as Intel's most powerful > Pentium. More important, Cell boasts a staggering fiftyfold advantage in > handling graphics-intensive applications that will define the next generation > of visual entertainment--blindingly fast and seductively immersive games, > virtual-reality romps, wireless downloads, real-time video chat, interactive > TV shows with multiple endings and a panoply of new services yet to be dreamed > up. > > > > > > IBM reckons Cell, potent and versatile, can do a lot more than just play > games. It sees a role for it in mobile phones, handheld video players, > high-definition televisions, car design and more. Scientists at Stanford > University are building a Cell-based supercomputer. Toshiba plans to use the > superchip in TV sets, which one day could let fans watch a football game from > multiple camera angles they control. Raytheon is set to use Cell in missile > systems, artillery shells and radar. Other companies envision new > high-definition medical imaging. "Cell is the next step in the evolution of > the microprocessor. It's a peek into the future," says Craig Lund, chief > technology officer at Mercury Computer Systems, which makes medical and > military systems and is taking orders for Cell servers. > > IBM is already at work on beefier versions of Cell, and it has launched an > allout campaign to woo a new generation of code-crunchers and game boys to > write software for its futuristic chip. In an extraordinary move IBM disclosed > hundreds of Cell's design secrets on the Internet, releasing a developer's > guide that 10,000 programmers have since downloaded. IBM, with annual sales of > $94 billion, says Cell could power hundreds of new apps, create a new > video-processing industry and fuel a multibillion-dollar buildout of tech > hardware over ten years. > > "We think this is going to spawn the next generation of growth in the > industry," says James Kahle, 45, the renowned chip designer and IBM Fellow who > oversaw the creation of Cell. "This chip will give you performance that is not > achievable with any other architecture." Adds H. Peter Hofstee, an IBM > scientist and the chief architect of a key part of the Cell chip: "We're > talking about everything from making TVs to shooting things up into space to > building huge supercomputers." He and Kahle spend much of their time on the > road, running mind-blowing demos and proselytizing prospective licensees and > geek groupies. > > But IBM will have to clear some high hurdles to deliver on Cell's prodigious > promise. Myriad competitors, including hotshot Silicon Valley startups > ClearSpeed and Stream Processors, are in pursuit of next-gen chips. High tech > is littered with the remains of chips that boasted remarkable abilities in the > lab but failed in the marketplace, starved by reluctant programmers and > recalcitrant customers and strangled by their own makers' miscues. A quarter > of a century ago Gene Amdahl, the famed architect of the IBM 360 computer > family, had an ambitious scheme to pack supercomputer power onto a chip but > was too far ahead of his time, and his Trilogy Ltd. went down in flames. In > the early 1980s the chip in the Amiga home computer far outraced those in the > Intel line, but Intel conquered the market anyway. In the early 1990s Digital > Equipment Corp. made the first 64-bit processor. It was an engineering tour de > force and a commercial flop. > > If anything stops Cell's commercial success, it is likely to be the chip's > very power. It is, to put it politely, a challenging platform for software > creators. "The programming model is a nightmare," says Marc Tremblay of Sun, > chief architect of a rival chip called Niagara, which uses a more traditional > approach. He argues Cell's balky design will snag widespread adoption beyond > gaming. > > Even the hard-sell salesmen at IBM are quick to say Cell poses no threat to > Intel, the world's leading chipmaker. Intel's processors do a great job on the > basic business applications for desktops, laptops and servers. In this mature > and mundane market Cell, specially geared to spin out intricate images at very > high speeds, offers no real advantage. But the Intel architecture, 25 years > old and constrained by having to be compatible with predecessor chips, is ill > suited to next-gen imaging. Thus the world must move up to Cell, IBM argues. > "We are going into a new era," Kahle declares. "The world is changing." > > An IBM demo shows the contrast. A terrain rendering program lets you fly over > Mount Rainier at 1,300mph. Cell crunches through millions of lines of > topographical and photographic data per second to paint topographically > accurate, photo-quality pictures at a movie-quality 30 frames per second. On a > similar program a Pentium takes more than two minutes to sketch a single > frame. > > CELL'S GENESIS FIVE YEARS AGO BEGAN WITH AN AUDAcious challenge. Sony's new > PlayStation 2 had just debuted, and Sony videogame chief Ken Kutaragi was > already looking ahead to the next version. He told IBM he wanted a > thousandfold increase in power. IBMers took up the dare, one so bold that it > challenged them to think beyond current chip designs. "We knew we could never > make the existing stuff go a thousand times faster," says Hofstee. > > In early 2000 Sony, Toshiba and IBM set up the STI Design Center, housing it > at an IBM site in Austin, Tex. James Kahle was put in charge. Armed with only > a bachelor's degree from Rice University, Kahle, born in Venezuela and raised > in New York, had joined IBM out of school in 1983 to write software for > designing next-generation chips. His low-key, nice-guy style masks his > intensity as a chip designer--his work shows up in the Apple G3, G4 and G5 > computers, the Nintendo GameCube and IBM's biggest Unix servers; he calls some > chips "my grandchildren." > > The project employed 450 engineers, mostly from IBM. They worked a lot, > socialized a little (group dinners, a few ski trips) and struggled with the > barriers of technology and physics--and of language and culture. Weekly > English classes were held for Japanese staffers. In brainstorming sessions > some Sony and Toshiba engineers had a penchant for diplomacy and mulling every > option; they found it jarring when their IBM counterparts relied on instinct > and blunt, bare-knuckled debate. > > Cell's creators needed to strike a balance between raw power and the > versatility to do more than just play games. Special graphics chips are > superspeedy, but for only one task. General-purpose chips like those made by > Intel devote a lot of muscle to the ability to handle a wide variety of jobs, > but they aren't superfast at any one of them. For two decades Intel boosted > performance by cramming more transistors onto a chip, but now chips draw so > much power and generate so much heat that they can't be cranked up much more. > Intel and others boost performance by lashing together two or more thinking > elements on a single chip. Intel makes dual-core chips. Sun's Niagara boasts > eight cores. For Microsoft's Xbox 360, IBM linked three Power cores. But even > these multicore chips will not be powerful enough to drive the next wave, > Kahle argues. Cell needed an entirely new design. > > Cell uses a single, central processing core that routes work to eight tiny > (but powerful) offspring called synergistic processing engines, or SPEs. A > year into development the engineers had a design ready but then felt compelled > to revise it because it would be too difficult for developers outside the game > business. "Game developers don't mind working with a difficult chip, but we > wanted to reach a wider audience," says Michael Day, an IBM software engineer. > > Months later they devised a new approach that drives Cell today. But more > hurdles arose. Engineers grappled with a highly complex memory-management > system that controls how bits of data are fed in and out of the SPEs. For 16 > weeks Kahle's staff would meet every morning at 9 a.m. to hash out the > problem. "We came up with one design after another and kept throwing them out > and coming up with new designs. We sat there for three or four hours a day. > Sometimes we never got out," Hofstee says. > > By April 2004 the first working chip came off the line at IBM's silicon > factory in East Fishkill, N.Y. The new Cell didn't deliver the 1,000X gain > that Sony wanted--but it did deliver 50X. Cell cranks out 200 billion > floating-point operations per second (200 gigaflops). That is akin to a > full-fledged supercomputer in the late 1990s. Add an Nvidia-designed graphics > chip and PlayStation 3 runs 2 trillion instructions per second. > > By early last year Sony was sending out Cell prototypes and software tools to > get developers started on writing new games for PlayStation 3. "We're seeing > stuff that goes dramatically beyond what we can do with the current generation > [of games]," says Andrew Goldman, chief executive of Pandemic Studios, a Los > Angeles outfit that wrote a series of popular Star Wars games for PlayStation > 2. "And what you will see over time is going to be even more amazing." He says > it will take years to fully exploit Cell's capabilities. > > Last year IBM began its own evangelizing. Instead of revealing design details > to only a small number of potential partners sworn to secrecy, IBM trumpeted > Cell's secrets on the Internet, releasing 700 pages of documents describing > the new architecture and a 1,100-page development kit, free for Internet > download. "We've opened up the architecture and provided all the details," > Kahle says. "We want to see this architecture proliferate in the marketplace." > > The wooing is necessary, for Cell's "asymmetric" design (its eight > co-processors have a different architecture than the main core), though key to > the chip's superior performance, is also what makes writing software for it so > difficult. In the mainstream chip world designers use an array of tool kits > and high-level programming languages (such as C++) to easily convert > instructions into a form the chip can comprehend. Such tools exist for Cell, > but the chip's design is so complex and so utterly different from anything > before it that code-crunchers do some of the work "down on the metal," > cranking out basic assembly code, which can take five times as long. > > The good news: Some designers say creating games for Cell is far less > complicated than writing for PlayStation 2. "Anyone who worked on the > PlayStation 2 is jumping for joy," says Jeremy Gordon, chief executive of > Secret Level, a gamemaker in San Francisco that is remaking a classic 1980s > Sega videogame for the new Sony box. > > Selling Cell, Hofstee last year gave eight speeches at technical conferences. > He and Kahle have visited more than 50 companies, enduring abundant skepticism > from jaded industry veterans--until they ran their speedy Cell demos. "It's > just amazing to go meet with people who have been in the industry for 25 years > and just see their jaws drop," Kahle says. When a famous chip designer, a > veteran of Motorola and Apple, visited Austin for a demo in 2004, Kahle showed > him images from the Mount Rainier flyover, eliciting stunned silence. "He just > got really quiet," as he realized "what this is going to do to the industry," > Kahle says. > > Toshiba demos a Cell-based "Magic Mirror" that turns an LCD screen into a > virtual mirror by combining feeds from several cameras. Look left, look right > and your "reflection" mimics you on screen, thanks to that tiny Cell chip > zipping away. In the next generation of TV sets Toshiba hopes to lay the > foundation for interactive viewing. One day you might watch a football game > from the quarterback's perspective and shift to a seat up high on the 50-yard > line, then zoom up to watch from the blimp overhead, backflips that videogame > players take in stride. > > Masakazu Suzuoki, Sony's lead designer on Cell, says Sony aims to use this > power to create movies that are interactive and changeable, with multiple > story lines, so people will watch the same flick more than once. Another idea > Sony is kicking around: placing ads in the background of movies and TV shows > and customizing them to suit the viewer, with Cell processors keeping track of > who sees what. > > Breakthrough chips easily inspire such big ideas, but Cell enjoys a running > start that previous chips didn't have. It is likely to end up in millions upon > millions of homes around the world as the PlayStation 3 rolls out. Once these > Cells throb away in game consoles, TV sets and set-top boxes, they can be fed > digital fare by new networks of Cell-based servers. "As the clients become > very powerful, then the servers will have to become very powerful, too," Kahle > says. > > The PlayStation hook inspires confidence at Raytheon, the Waltham, Mass. > defense contractor, which has studied Cell for 15 months and plans to use it > in scores of next-generation systems. "Sonar, infrared sensors--there are > hundreds of products that Raytheon designs that could use this type of > technology," says Peter Pao, chief technology officer. "Current chips are > going to run out of steam. We always look to the future." > > At Mercury Computer Systems in Chelmsford, Mass. engineers are working on a > Cell system called Turismo, which is due later this year and will pack up to > 128 Cell chips into a 6-foot-high rack, producing up to 25 trillion > calculations per second. Mercury, which sells modules for medical gear made by > General Electric, Philips and Siemens, says Turismo could make a CT scanner so > fast that it will be able to paint a 3-D image in four seconds versus five > minutes on an Intel Pentium. Mercury is even pushing Cell to firms that create > computer-generated special effects for movies. "This chip is opening doors for > us," says Joel Radford, a Mercury vice president. > > Back in Austin Kahle talks about "immersive interaction," 3-D virtual worlds > modeled with such detail that you can see, from your screen, exactly what it > looks like when you're standing on the corner of Fifth Avenue and 23rd Street > in Manhattan. "This is going to open a whole new realm of how people interact > with computers and each other, where we'll mix reality with virtual reality," > he says. For his entire career Kahle has been driven to create a chip that > could change the world; now, he says, it has arrived. "This project is the > culmination of that desire." > > > > Sean Ness > Business Development Manager > > Institute for the Future > Office: 650.233.9517 > Cell: 408.406.7597 > Main: 650.854.6322 > Fax: 650.233.9417 > [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]> > > > IFTF's Future Now Blog - <http://future.iftf.org/ <http://future.iftf.org/> > > > We moved!!! Here's our new address: > 124 University Avenue, 2nd Floor > Palo Alto CA 94301 > >
------ End of Forwarded Message --- You are currently subscribed to telecom-cities as: [EMAIL PROTECTED] To unsubscribe send a blank email to [EMAIL PROTECTED] Manage your mail settings at http://forums.nyu.edu/cgi-bin/nyu.pl?enter=telecom-cities RSS feed of list traffic: http://www.mail-archive.com/telecom-cities@forums.nyu.edu/maillist.xml