Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
×
IBM Technology

IBM Makes a Super Memory Breakthrough 164

adeelarshad82 writes "IBM says they have made a significant leap forward in the viability of 'Racetrack memory,' a new technology design which has the potential to exponentially increase computing power. This new tech could give devices the ability to store as much as 100 times more information than they do now, which would be accessed at far greater speeds while utilizing 'much less' energy than today's designs. In the future, a single portable device might be able to hold as much memory as today's business-class servers and run on a single battery charge for weeks at a time. Racetrack memory works by storing data as magnetic regions (also called domains), which would be transported along nanowire 'racetracks.' Instead of forcing a computer to seek out the data it needs, as traditional computing systems do, the information would automatically slide along the racetrack to where it could be used."
This discussion has been archived. No new comments can be posted.

IBM Makes a Super Memory Breakthrough

Comments Filter:
  • by hey ( 83763 )

    That's super-doubleplus good.

  • by eldavojohn ( 898314 ) * <eldavojohn@@@gmail...com> on Wednesday December 29, 2010 @12:43PM (#34700644) Journal
    Really? The summary doesn't even get around to explaining what the alleged "breakthrough" was. It's just trumpeting the awesomeness of race-track memory. From the article:

    "We discovered that domain walls don't hit peak acceleration as soon as the current is turned on, and that it takes them exactly the same time and distance to hit peak acceleration as it does to decelerate and eventually come to a stop," commented Dr. Stuart Parkin, an IBM Fellow at IBM Research. "This was previously undiscovered in part because it was not clear whether the domain walls actually had mass, and how the effects of acceleration and deceleration could exactly compensate one another. Now we know domain walls can be positioned precisely along the racetracks simply by varying the length of the current pulses even though the walls have mass."

    Don't get me wrong, race track memory is some pretty exciting stuff but I think we're dealing with an observation that means they can now proceed along a certain strategy for storing and retrieving bits. I don't think I would call this a breakthrough, it sounds like they set out to investigate domain walls and learned something. How is that a breakthrough? We're still in the ten to fifteen years away period which is that magic flying car period that, in many instances of exciting new technology, never seems to shrink.

    "Breakthrough" no longer means anything to me. I don't know what you would have to put in the title to get me genuinely excited about a real breakthrough ... probably something like "Researchers Shitting Themselves Over New Discovery."

    • I hear you.

      I'm only on board with scientists are "baffled" or experts are "shocked."

    • "Breakthrough" no longer means anything to me.

      I, for one, look forward to 15 years of news articles proclaiming new breakthroughs mean we'll have racetrack memory "within ten years"

    • by wealthychef ( 584778 ) on Wednesday December 29, 2010 @01:05PM (#34700980)
      Actually, the word "breakthrough" is pretty applicable here. There was this undiscovered property that acted as a barrier and prevented moving forward with the technology, but now that it is discovered, the barrier has been broken through and progress can continue. You might not be satisfied unless it's an announced product, and I'm with you there, but it's still a breakthrough in the technical sense of the word.
      • Yeah, it's interesting how every story about a new technology ends up full of comments about how it's not a big deal, how it only works in a lab, how any real applications are decades away, etc... yet there's new faster, better, cooler, more efficient, etc... products coming out all the time.

        That's not to say that if someone starts crowing about their exciting new discovery that you should automatically rush in and invest all your money in it, but technology does actually move forward, and not everything is

    • by LWATCDR ( 28044 )

      I have to wonder if there is going to be any market for these advances. The high end is shrinking very quickly so the market for really super high end cutting edge stuff is also shrinking.
      Even super computers are using a large number of COTS technology these days. In the future will their be any customers for the first very expensive race track memory systems?

      • Id imagine researchers would want to use it for more powerful computer clusters or supercomputers. There is always a demand for that, mostly DoD contracts to universities.
        • by LWATCDR ( 28044 )

          It will be interesting to see. Even the DOD is going to more COTS. Will the DOD and DOE contracts be enough to get this memory from the lab into production and then finally to mass market?

      • by mcgrew ( 92797 ) *

        I have to wonder if there is going to be any market for these advances.

        Did you even read the summary? "In the future, a single portable device might be able to hold as much memory as today's business-class servers and run on a single battery charge for weeks at a time."

        Fifteen years ago nobody thought we'd be watching YouTube.

        • by LWATCDR ( 28044 )

          I did. but there may be the need between the consumer and the lab for this to mature.
          A price is no object performance market. As I said even super computers are now often made up of COTS parts. This is more a question of economics and not technology. Will any company be willing to make the long term investment to bring this to the consumer market without a nice cash injection from a high margin market?

    • by SageMusings ( 463344 ) on Wednesday December 29, 2010 @01:20PM (#34701186) Journal

      It's "Bubble Memory" all over again.

    • Well I'd have to say that since you don't know what the breakthrough is, nor do you really appear to understand the technical issues involved (other than quoting words from the article)... I'm not sure you are really saying anything. Much like my post :)

      Breakthrough's in science can be very simple things that move projects forward significantly. A breakthrough doesn't require a nuke going off, or a plane breaking a new mach record.... it could be as simple as what they state as to resolve an issue that

    • Perfect XKCD for this:
      http://xkcd.com/678/ [xkcd.com]
    • by Twinbee ( 767046 )
      How about we all use a qualifier, like a number out of 10. A breakthrough-0 would be a small tiny one, and a breakthrough-10 would be a "oh wow, we just discovered free energy, and 5 top universities agree and have published it on their university home page - go look now!". hmmm...
  • Exponentially (Score:5, Insightful)

    by serviscope_minor ( 664417 ) on Wednesday December 29, 2010 @12:44PM (#34700654) Journal

    I hate it when people misuse the word exponentially to mean big.

    At best, it will allow the current exponential growth to increase.

    • by mlts ( 1038732 ) *

      Here what it boils down to, IMHO: Will the racetrack memory provide enough addressable space, at a decent price, to allow Adobe and other large applications to run decently?

      • by mlts ( 1038732 ) *

        Correction, Adobe applications. It would be nice to be able to feed stuff large amounts of RAM addressable in nanoseconds, because it is a *lot* easier to throw more hardware at something than to get most vendors to tighten up their products.

    • At best, it will allow the current exponential growth to increase, exponentially.

    • Yep, it's like "exponentially" is the ultimate word to use.

      • But how about even faster growing functions like factorial and the up arrows? When are we going to see an advert like this..

        "Whazzarop is a super technological breakthrough potentially enabling up arrows increase in computing performance."
  • From TFA:

    Racetrack memory is still years away from hitting the consumer market..

    In other words, maybe in the next 20 years, right?

  • So it goes really fast, but the article left out the answer to the quintessential question: does it turn left?

    • No, it's not an ambi-turner.

      Maybe someday, though, it may learn to, so that it can thwart the attempt on the Prime Minister of Malaysia's life.
      • No, it's not an ambi-turner.

        Maybe someday, though, it may learn to, so that it can thwart the attempt on the Prime Minister of Malaysia's life.

        Fortunately it WILL help with the development of really teeny-tiny cell phones; so there's that.

  • While this technology sounds great, I have a feeling this is more than five years away. Hell, I'd be happy if IBM delivered on the holographic storage [wikipedia.org] they've been promising for the past 15 years.
  • Can I ask Slashdot to not post any more stories about Racetrack memory until something interesting happens with it? I've been hearing about it for years, but thus far it's all theoretical or early experimental work. Just like Bubble memory, by the time this actually works conventional memory may be faster and cheaper and it will end up on the sidelines of history.

    I'll be intrested when they have something like a DIMM form factor that is actually better than existing memory.
    • Can I ask Slashdot to not post any more stories ... until something interesting happens

      you must be new here. low ID aside.

    • Comment removed based on user account deletion
    • I'll be intrested when they have something like a DIMM form factor that is actually better than existing memory.

      I'll be happy enough when it's up to competing with rotating memory, which is a lot more likely.
      Serial memory is serial memory, and promising to replace Random Access Memory in latency-critical applications like main memory is just nonsense. Either the people putting out these claims are stupid or they think we are.

      • I'll be happy enough when it's up to competing with rotating memory, which is a lot more likely. Serial memory is serial memory, and promising to replace Random Access Memory in latency-critical applications like main memory is just nonsense. Either the people putting out these claims are stupid or they think we are.

        You won't be asking to access the one bit at the end of a 8KB track (and stalling the CPU waiting for it). Modern chips move a whole line of cache at once - a whole 64 bytes for my current chip. And according to the wikipedia article on racetrack memory, the tracks are only 10-20 bits each - not terribly serial. If one track can be cycled around as fast as DRAM, then a bunch of tracks can be done in parallel to handle 64 bytes at once just as fast as DRAM. That's probably years away, but it's not as crazy

    • by geekoid ( 135745 )

      This was interesting. Determining whether or not the domains has mass is very exciting.

      Just because you only care about crap that you can buy doesn't mean others aren't interested in scientific breakthroughs.

      The pre-millennium jandrese called, he want's to know why you killed his curiosity.

      • by jandrese ( 485 )
        The domain mass thing is somewhat interesting, but the whole story is about how Racetrack memory is going to be totally awesome in the future because mumble mumble. A link to the paper about the magnetic domain experiments would have gone over much better IMHO.
  • a new technology design which has the potential to exponentially increase computing power

    P = NP

    QED

  • Instead of forcing a computer to seek out the data it needs, as traditional computing systems do, the information would automatically slide along the racetrack to where it could be used. The result: powerful and efficient computing.

    "Instead of forcing the computer to seek out data." (Meaning, at the address where it was stored?) "The data automatically slides to where it can be used." (Is the data omniscient?) "Powerful and efficient computing." (OK, perhaps w/regard to data retrieval.)

    I don't get it.

    • My understanding is that if your storage was one big racetrack the "address" of the bits you wanted correlates to the size of the pulse needed to "seek" it by a constant amount (this constant factor being the breakthrough discussd here). so if you want bit number 100 then simply pulse the memory 100*constant and it shows up on your sensor. No "seeking" flipping of transistors required

  • by LodCrappo ( 705968 ) on Wednesday December 29, 2010 @01:09PM (#34701036)

    why not connect the racetracks directly to the internet tubes. then the information could slide along the racetrack into a series of tubes and ultimately slide right into your own personal racetrack.

    • The domains move at about 100 m/s. Light moves through an optical fiber at about 200,000,000 m/s. Assuming that your scheme was possible, it would represent a 2-million-fold increase in latency (at the physical layer). It would take a packet 12 hours to get from New York to Los Angeles, giving a round-trip time of about a day.

      Aside from the fact that it would be completely useless, it's also impossible. The internet is packet switched, not circuit based. There's almost never a complete, continuous "tra

      • look, the internet is not a truck. you cannot just dump enormous amounts of material onto it.

        as long as I don't have to wait until Monday to get an internet I sent out on Friday, I don't see what the problem is.

      • Your facts do not in any way mitigate the tremendous WHOOOOSH.
        • Not every obviously stupid statement is a funny statement. Neither is every post that contains the phrase "series of tubes" hilarious. It only makes sense to imply that I missed the joke if there was a joke to miss.

          I contend there was nothing humorous. Can you point out anything in the original post that was funny?

          The post doesn't contain any puns, explicit mockery, surprise or elements with multiple interpretations, which are some basic aspects of humor. It might be argued that the post is so stupid th

          • Responding to jokes like Debbie Downer because you don't think they are funny does not add anything to the conversation. Responding, as you did, in perfect sincerity is worthy of a /. WHOOSH.

            Defending your post with a checklist of what a joke requires to be funny just makes you look even MORE stiff and grumpy. If you don't like /. humor, set your preferences to down-rank +funny mods or just scroll down.

            I'll give you the benefit of doubt though. We all have bad days. I've groaned at horrible jokes and typed

          • For some reason, I read your original post in the voice of "Debbie Downer" and I actually thought you were being pretty darn funny.

            Then I read your response, and realized you are being serious.

            ...or ARE you?

            Maybe your sense of humor is overly-strict. Maybe you're just having a bad day. Maybe you're the next Andy Kaufman, goofing us all with a lecture on what makes jokes funny. I'm too tired to figure it out, but I'll give you the benefit of the doubt.

    • by c0lo ( 1497653 )

      why not connect the racetracks directly to the internet tubes. then the information could slide along the racetrack into a series of tubes and ultimately slide right into your own personal racetrack.

      I imagine the Pentagon and US State Dept would strongly object, even sue... and please think at the Swedish court and women population: do you think they'd be able to handle the entire inter-tubes connected population?

  • by olsmeister ( 1488789 ) on Wednesday December 29, 2010 @01:16PM (#34701140)
    Here is [ibm.com] a press release from a couple of years ago basically trumpeting the same thing. I think it is policy to recycle this every so often to prop up their stock price. [nyse.com]
  • by DCFusor ( 1763438 ) on Wednesday December 29, 2010 @01:20PM (#34701194) Homepage
    I guess the patent has finally run out on the original, which I played with in the early '70s for some military EE work I was doing then for a beltway bandit. Just a big bunch of shift registers moving magnetic "bubbles" or "domains" round and round. The thing had a (for the time) decent capacity and storage capability, for example, you could get just about floppy drive performance out of a chip (and some other parts to make all the clocks)....It was of course still far slower than the ram of the time.

    To this old fart, it looks the same, just a different way to fab the thing. But hey what do I know?

    One thing I do know. Current scientists aren't very well educated on what has gone before. About a year ago I saw the "breakthrough" development of a "plasma transistor" that I also had in a 1950's book on my shelf....happens pretty frequently these days. These guys are so specialized they don't even know the history of their own fields anymore, much less a broad history.

    Reminds me of Hari Seldon and "the galactic empire is crumbling" to be frank. Not even up to Heinlein standards!

    • by vlm ( 69642 )

      One thing I do know. Current scientists aren't very well educated on what has gone before.

      Computer scientists / IT people have the same problem. Nothing is really new. Personally I'm waiting for "implicit typing" to be in style again. Python whitespace is conceptually pretty close, probably why I find it repulsive.

      • by Desler ( 1608317 )

        Personally I'm waiting for "implicit typing" to be in style again.

        WELCOME TO THE WORLD OF TOMORROW! [developer.com]

        • Oh BS.

          Python's type checking is done at run-time, which is why it's so horrible (though it's better than the weak typing as present in Perl and Javascript... PS, I like both of those languages for various reasons, I just hate weak typing).

          Type inference as present in C#, Haskell, Ocaml, and others, is done at compile time, and so is perfectly safe. It just means the programmer can spend less time casting things, which is a huge pain in the ass in a strictly type language like C#, particularly when you thro

          • by Desler ( 1608317 )

            Oh BS.

            My post wasn't being serious...

            Type inference as present in C#, Haskell, Ocaml, and others, is done at compile time, and so is perfectly safe. It just means the programmer can spend less time casting things, which is a huge pain in the ass in a strictly type language like C#, particularly when you throw generics and lambdas into the mix.

            Gee no shit? It's almost like I already posted that [slashdot.org].

      • Python whitespace is conceptually pretty close, probably why I find it repulsive.

        Python whitespace in a nutshell: Indent your code as always. Next, don't bother with any brace brackets because you're done.

        Why this bothers anybody is utterly beyond me. On the other hand, when I write C these days I sometimes wonder why I have to babysit the compiler by entering a bunch of brace brackets when the indentation is already there showing the code structure plain as a day.

    • before magnetic disks and tapes were perfected
    • by geekoid ( 135745 )

      The break through isn't the technology, it's how they are going about it, or certain aspects of the technology. Like, does a domain have mass.

      Seriously, you need to to think a bit more. It's like someone finding a away to make better rubber tires and all you can say is 'Tires? hell those where around 100 years ago, this isnt new at all. I guess these scientist don't know their history."

    • To this old fart, it looks the same, just a different way to fab the thing. But hey what do I know?

      It is the same thing, but the scale is far different, with much the same consequences as going from discrete transistors to nanoscale transistors etched on silicon, i.e., it can (theoretically) store more data and retrieve it faster.

      About a year ago I saw the "breakthrough" development of a "plasma transistor" that I also had in a 1950's book on my shelf...

      I know what you mean. I was going through a mass-market encyclopedia of science from the 1960's the other day, and stumbled across an article promising that holographic memory was right around the corner.

      To be fair, though, the basic principles of most of the technology we use t

    • by ErikZ ( 55491 ) *

      A book on your shelf?

      If it's not on the Internet, it doesn't exist.

  • by MarkvW ( 1037596 ) on Wednesday December 29, 2010 @01:29PM (#34701314)

    If I can carry around all my data in a little pod, then all I'll need is access to input and output devices.

    That would be far out. Thanks IBM, for the neat science fiction story of the day!

  • by the_rajah ( 749499 ) * on Wednesday December 29, 2010 @01:37PM (#34701434) Homepage
    I worked on magnetic bubble memory at T.I. in the Dallas corporate research labs back in the mid-70s and it used a "racetrack" architecture where magnetic bubbles (domains) were stored in very long shift registers with the shifting accomplished by rotating magnetic fields. I hope it does better this time around.
  • I watch car races on TV for the same reason most people do. To see the crashes. What happens if the data in these memory chips fails to make the turn? Getting implaled by ones and zeros doesn't sound like much fun. I'm just glad we're not using Roman Numerals, because those dots on the i's flying about, and those x's look a lot like those Japanese surikans, and those L's winging around like boomerangs.

    Just how safe are we?

  • Riiight. I welcome our long-lasting, battery powered overlords...if they ever transcend marketing fiction and appear IRL.

  • I read this:

    Racetrack memory works by storing data as magnetic regions (also called domains), which would be transported along nanowire "racetracks." Instead of forcing a computer to seek out the data it needs, as traditional computing systems do, the information would automatically slide along the racetrack to where it could be used."

    And I'm focused on the word "automatically". Um, so racetrack memory is clairvoiant? No, it appears to be a FIFO method. So it appears "automatically" when it's time for it

    • Sounds like the tape in a Turing machine. They accelerate the data in one direction or the other, and can modify it when they stop.
  • I didnt realize they were still in the business of developing new stuff.

    Seems the core of their business is acquiring other technology companies, and injecting red tape and excess bureaucracy into other enterprises through proliferation of their "Architectural Thinking" workshops.

  • http://www.sciencemag.org/content/330/6012/1810.abstract [sciencemag.org]

    Courtesy of a better writeup at:
    http://www.computerworld.com/s/article/9202379/IBM_s_racetrack_memory_moves_closer_to_the_checkered_flag?taxonomyId=147 [computerworld.com]

    In a paper published in the Dec. 24 issue of Science Magazine, the IBM researchers report that domain walls have mass and do indeed take a bit of time to speed up to peak velocity, and to slow down. Knowing this, they'll be able to move and retrieve data on a racetrack trip accurately. There's still a lot of work to be done before racetrack becomes a reality, but according to Parkin, the biggest questions -- whether an electric charge would move these domain walls, and whether or not they have mass -- have now been answered. Now the problems are more practical and less theoretical: how do you build a racetrack chip that works reliably with millions or even billions of these racetracks, for example. "Those are the questions that we can only address by building prototypes and testing them for a period of time," Parkin said.

    And the official IBM press release:
    https://www-304.ibm.com/jct03001c/press/us/en/pressrelease/33291.wss [ibm.com]


    I see more data center utilization for this technology rather than consumer devices. Be nice if I could get a home NAS on one of these in 5-10 years.

  • So what's the difference between this and the old TI bubble memory concept?

  • Sounds promising but would like a lot more information. I doubt they could give out more info at this time due to competitors looking into their type of racetrack memory right away. With the limitation of solid state drives, we need a breakthrough in some form or fashion.
  • Delay line memory [wikipedia.org].

    Completely different form of energy, of course.

  • Didn't RTFA to know this: At least IBM is still doing R&D, and has been for quite some time. That is usually the first budget item to be slashed in economic downturns and I am grateful that big blue (notice not capital B on those) is still doing it's part to better our computing environment. On a personal note I joined a banking company years ago, one of the reasons was because they had an R&D division (not to date myself but that was the Windows NT era). That was an early goal of mine to strive
  • How does racetrack memory compare to that other favourite - the memristor? What are the main disadvantages/advantages of either?

I had the rare misfortune of being one of the first people to try and implement a PL/1 compiler. -- T. Cheatham

Working...