Forgot your password?
typodupeerror
HP Hardware Technology

HP Reports Memory Resistor Breakthrough 141

Posted by samzenpus
from the flip-the-switches dept.
andy1307 writes "Hewlett-Packard scientists on Thursday will report advances demonstrating significant progress in the design of memristors, or memory resistors. The researchers previously reported in The Proceedings of the National Academy of Sciences that they had devised a new method for storing and retrieving information from a vast three-dimensional array of memristors. The scheme could potentially free designers to stack thousands of switches on top of one another in a high-rise fashion, permitting a new class of ultra-dense computing devices even after two-dimensional scaling reaches fundamental limits."
This discussion has been archived. No new comments can be posted.

HP Reports Memory Resistor Breakthrough

Comments Filter:
  • Finally something that sounds like it's actually patentable.
    • But the really exciting thing is accessing that third dimension.
      • by masshuu (1260516)
        so uuh? instead of 0x0001f57d 0x0001f57dx0005f566x0054faaa
        • by symbolset (646467)

          Well, they'll probably abstract it into the current one-dimensional addressing scheme for memory. But these things aren't used just for memory - you can build logic with them. Instead of 400,000 x 400,000 = (1,600,000,000 or 1.6 billion) transistors, 1,000,000 x 1,000,000 x 10,000 = (10,000,000,000,000,000 or 10 quadrillion) transistors in one package. Some for memory, some for logic, some for special purposes. You know, skynet. It gives us another 24 years of Moore's Law - though it probably won't take

    • Research! YES! (Score:5, Insightful)

      by Anonymous Coward on Wednesday April 07, 2010 @10:29PM (#31770966)

      Finally something that sounds like it's actually patentable.

      And not just patentable, but good solid research. It seems to me that lately, US companies have been cutting and cutting R&D budgets. The markets are so focused on who makes their current quarter earnings marks, and sinking money into innovation does not help towards making that profits goal. And because of this, it seems that we have lost touch with planning for the future.

      That always made me sick to my stomach. I am always thrilled when these big companies, that spun up and put technology where it is today, the HPs, the IBMs, the Xeroxs, the ATT/Bell/Lucents, etc., come out with something cool. I even like it when the small guys do something, but often they dont have the money to make it all the way to market.

      Anyway, my point is, I hope we see corporations (and everyone else, like NASA, etc) realize how important science and innovation are to our future. I hope that we can get back to the "old days" of (literally) shooting for the moon and achieving it, rather than spending money on fluffy marketing and trying to squeeze out margins with just barely passable work.

      This kinda stuff, I love. More please!

      (sorry for a horribly written post)

      • Re: (Score:1, Informative)

        by Anonymous Coward
        Why'd you post this anonymously? It was worth putting your (user)name on.
        • Why'd you post this anonymously? It was worth putting your (user)name on.

          Why would anyone write something that they didn't feel comfortable putting their (user)name on? To me, that's probably a sign that you shouldn't be writing it*...

          * whistleblowing and revolts against government excluded.

          • Why would anyone write something that they didn't feel comfortable putting their (user)name on? To me, that's probably a sign that you shouldn't be writing it*...

            Because your real name is "thePowerOfGrayskull", right? Either way, why do you need to know the identity of a poster? Isn't the important thing the quality of their argument not their identity?

            • Why would anyone write something that they didn't feel comfortable putting their (user)name on? To me, that's probably a sign that you shouldn't be writing it*...

              Because your real name is "thePowerOfGrayskull", right? Either way, why do you need to know the identity of a poster? Isn't the important thing the quality of their argument not their identity?

              Actually, my real name is "Marc Paradise" which is pretty easy to deduce since my web site and sig both point to "marcparadise.com" ;)

              That aside, you raise a good point and I agree. But in context of what I was replying to, OP said that the GP's post was "worthy" of using his registered name for. My reply to that is that if you're posting anything you feel you *must* hide or distance yourself from, perhaps there's a reason you feel that way and you should think twice before posting it at all.

            • The are two things I dislike about the dynamic forms on /., and one of them is that if I accidentally hit "back" I lose my post-in-progress. < /mini-whine>

              First - my real name is "Marc Paradise", which isn't exactly a secret considering that my sig and web site both link to "marcparadise.com".

              That aside, I agree with you. But in context of the conversation, I was trying to say that if you're making a post that you feel you *must* distance yourself from, then perhaps you should look at why that is

              • Shit. Alternatively, I already posted the first comment, and forgot that I did so when I hit "back" by accident... At least my reply was consistent ...
        • by Khashishi (775369)

          maybe 'a isn't a regular

    • Re: (Score:2, Insightful)

      Finally something that sounds like it's actually patentable.

      Yeah, but it can't be that big of a breakthrough... Nobody's filed any lawsuits yet.

    • The devices, known as memristors, or memory resistors, were conceived in 1971 by Leon O. Chua, an electrical engineer at the University of California, Berkeley, but they were not put into effect until 2008 at the H.P. lab here.

      Except the idea isn't new, it's just the first time they can actually make one and test it.

      Regardless of that, this sounds very interesting. They are non-volatile, they are 1/th the size of a transistor, and they use far less power. Also (I assume), they should be cheaper to make. They also said that they tested them with hundreds of thousands of rw operations. That is pretty amazing at such an early stage of development.

      • Re: (Score:3, Funny)

        Once HP figures out a way to make sure that they won't work if either the black cartridge or the combined color cartridge are empty, Memristors will be ready for commerical release...
      • by hitmark (640295)

        if they can handle more RW then a SSD, without wear leveling, without breaking, it would be very interesting indeed.

        • since the memristor is a new fundamental componant of electronics, one would have to assume it has teh same durability as a transistor, resistor, etc.

          • by hitmark (640295)

            maybe so, tho the first design description talked about moving oxygen atoms around, or something like that...

  • by actionbastard (1206160) on Wednesday April 07, 2010 @09:47PM (#31770668)
    Has been resisting me for years. I'll be damned if I can remember where I put my keys.
  • Heat? (Score:4, Insightful)

    by pushing-robot (1037830) on Wednesday April 07, 2010 @09:51PM (#31770702)

    But is it so much more efficient that you could stack thousands of layers without turning your chip into a hunk of molten glass? That would probably be an even bigger breakthrough.

    • Re: (Score:3, Informative)

      by mswhippingboy (754599)
      According to TFA, the intended use for this is memory devices (possibly a follow-on to flash memory). Since it can retain it's state even without power, it would seem that this would result in an extremely low power device which should produce very low heat.
      • Re: (Score:3, Informative)

        by DJRumpy (1345787)

        It's more than that. This could have huge implications for no-power flash storage, and it has lower power requirements than the phase-change memory that is currently the top dog. I'm also curious about the 'data processing' blurb in TFA:

        "They are simpler than today’s semiconducting transistors, can store information even in the absence of an electrical current and, according to a report in Nature, can be used for both data processing and storage applications."

        "He said the company could have a competit

        • by beav007 (746004)
          If we're now stacking in 3d, why are we still using square (instead of cubic) measurements?
          • If we're now stacking in 3d, why are we still using square (instead of cubic) measurements?

            USB flash drives, SDHC cards, and especially microSDHC cards still have a maximum thickness. "Gigabytes per square inch" would refer to the typical thickness of a packaged memory device.

            • by Firethorn (177587)

              From my reading, I don't think the added layers are going to add significant thickness. I'm picturing more like PCB boards - dozens of layers, not thousands. Each layer is still thin enough that the base substrate is still thicker than the rest of the layers, combined.

              Still, even 10 layers would raise density by an order of magnitude, so it's not shabby. I'd think you'd have cost concerns if you're trying for hundreds, much less thousands of layers. Each layer would be another set of manufacturing steps

      • by Anonymous Coward

        Since it can retain it's state even without power

        Its = possessive.

        It's = "it is"

        Learn fuckin' English.

        Protip: That last sentence was a correct use of an apostrophe.

      • by Your Anus (308149)
        If this is going to replace the transistor, it's also going to be used like RAM. Perhaps they won't stack it in that case, but heat transfer will be an issue if they do. Maybe they can embed some heat pipes in the stack.
    • Re:Heat? (Score:5, Informative)

      by HiThere (15173) <charleshixsn AT earthlink DOT net> on Wednesday April 07, 2010 @10:13PM (#31770884)

      Yes. Memristors don't require that power be applied in order to retain memory state. Heat might limit write and retrieval rate, but it wouldn't limit the number of layers. I suspect that it might make heat pipes built into the memory boards to be a highly desirable option, but that would be to enable faster access, not to allow a greater number of layers.

      • Re: (Score:2, Interesting)

        I suspect that it might make heat pipes built into the memory boards to be a highly desirable option, (..)

        Hardly - the maximum amount of heat loss would be limited by the application.

        If you'd use this technology to build a SSD for a laptop or a portable media player, there are some hard upper limits on how much power (=heat) that SSD could draw. Things like battery life, the amount of heat a full system can deal with, acceptable noise levels for cooling fans, etc. If bandwidth = heat, the application would limit the maximum available bandwidth for a given power consumption.

        With that constraint as a given, I

      • Yes. Memristors don't require that power be applied in order to retain memory state. Heat might limit write and retrieval rate, but it wouldn't limit the number of layers. I suspect that it might make heat pipes built into the memory boards to be a highly desirable option, but that would be to enable faster access, not to allow a greater number of layers.

        No reason we can't use the peltier effect to move the heat to the surface and sap it away with heat sinks like we do now.

        • If I'm reading this right, external cooling may not be the issue. The problem would be the thermal transfer coefficient of the memory chip itself. Imagine are large number of these modules being stacked on top of each other, they now take on a "cube" form. While the outside is nice and cold, the center of the cube could instantly spike in temp (causing damage) before it has a chance to migrate to the surface for heat dissipation.

          The obvious solution would to limit the density and/or how many modules to stac

        • by drinkypoo (153816)

          No reason we can't use the peltier effect to move the heat to the surface and sap it away with heat sinks like we do now.

          Please share with us your method of inserting peltier cooler layers between silicon layers. We're talking about removing heat from the middle of a three-dimensional structure, not a flat plane.

          • Please share with us your method of inserting peltier cooler layers between silicon layers. We're talking about removing heat from the middle of a three-dimensional structure, not a flat plane.

            Buildings have floors. They don't all have the same stuff in them.

            • by drinkypoo (153816)

              Please share with us your method of inserting peltier cooler layers between silicon layers. We're talking about removing heat from the middle of a three-dimensional structure, not a flat plane.

              Buildings have floors. They don't all have the same stuff in them.

              Your plan is to put the peltier cooler on another floor? I don't see how that will help.

              • Your plan is to put the peltier cooler on another floor? I don't see how that will help.

                Silicon wafers have layers. So do buildings. Put the cooling in between the active circuits. Heat moves upward. Use a tree-like structure to channel heat.

                • by drinkypoo (153816)

                  I had hoped I wouldn't have to spell it out, but the primary appeal of the memristor is the ability to layer circuit on circuit without anything in between, requiring no interconnects. Your plan eliminates this benefit.

                  There is a possibility though; MEMS-built heat pipes built into the chip. I seem to recall hearing that someone was researching this but I forget who, or if it was even true.

                  • I had hoped I wouldn't have to spell it out, but the primary appeal of the memristor is the ability to layer circuit on circuit without anything in between, requiring no interconnects. Your plan eliminates this benefit.

                    The primary benefit is the meristors isn't size, it's that they don't decay after a certain number of rewrites. They are durable solid-state components. The benefit isn't eliminated -- the layers can still be stacked. I'm just saying that you can use the peltier effect, possibly by aligning the meristors themselves in a lattice structure, to channel heat towards the conduits. In any semiconductor, there is a 'hot' and a 'cold' side, as it were. By aligning them all in a similar fashion, you can channel and

          • by Shotgun (30919)

            Wrap the peltier cooler around a copper sheet. The composite is inserted between the layers. The heat is highly concentrated in the copper, which is topped with a heatsink.

    • by Sulphur (1548251)

      Should you still call it a chip?

      Would cube be better?

      • Re: (Score:3, Insightful)

        by h4rr4r (612664)

        Chip, we called floppies floppies long after they had ceased to be floppy.

        • More to the point, we still produce core dumps, although the times when there was core memory to dump are long gone.

          And of course, solid state disks usually are not disk shaped.

          • We still produce core dumps? are you crazy? Blasphemous! My applications never core dump, ever! (Yeah, Friday is the official release date of v 1.5 of one of the apps I develop. There'll be a big presentation, lots of people, and I'm freaking nervous). I mean, I'm not nervous! My applications are bug free. (crawls back into the corner, continues countdown to Friday 10 A.M while swinging back and forth).

      • Re: (Score:1, Insightful)

        by Anonymous Coward

        Block, surely?

  • by judolphin (1158895) on Wednesday April 07, 2010 @09:58PM (#31770756)
    This is the real difference between genuine R&D (actual breakthrough in computer science) and Cupertino R&D (Let's remove the floppy drive! Let's remove the optical drive! Let's remove the keyboard! I can't believe we're acutally being paid for this!)
  • There wouldn't be a excuse for tiny amounts of space even on the lowest of the low end phones.
    • Sure there would. My phone supports SMS and calling, and not a heck of a lot else. One excuse would be "it can't actually make use of gigabytes of storage".
  • by patlabor (56309) on Wednesday April 07, 2010 @10:50PM (#31771106)

    This is a really big deal. Since our brains work in much the same way as an array of memristors, this brings the possibility of an artificial brain (and perhaps artificial intelligence) much closer to reality.

    Maybe I will live to see Data in my lifetime.

  • I read about the original research and hadn't heard anything for a while and wondered if HP was still working on this.

    Personally, I feel this is OUTSTANDING news. RTFA, they think they'll get 20GB on a square centimeter. And have a viable competitor to flash memory in 3 years.

    Instead of "coming on Blu-Ray Tuesday!", it'll be "coming on MR-Chip Tuesday!"

    Of course, when they get the 4D version working, that'll change to "coming last week on MR-Chip!"

    ( And yes, I just copyrighted "MR-Chip"...)

    • ( And yes, I just copyrighted "MR-Chip"...)

      You're too late [imdb.com].

    • by gtall (79522)

      Just wait until Mark Hurd hears about it..."You mean we still have researchers we haven't outsourced or fired yet? Flunky, get me a stack of pink slips, I'll fix this straight away!"

  • It seems like Joule heating would be a problem if the memory consists of resistors. In particular, a three-dimensional stack would build up heat fairly quickly. Of course, switching transistors requires a good deal of energy so the prospect of not needing to constantly refresh each element may be a huge advantage in this respect. Also, it depends on the on and off resistances and the currents required to read and write bits.

    Does anyone know of a link to a more technical description of the technology?

  • I think HP's research could pave the way for solid-state disk (SSD) drives in the 3-4 TB storage capacity range about the size of today's 2.5" laptop drives, which will essentially end the reign of hard disk drives with spinning disc platters and moving read-write heads. Imagine being able to boot Windows 7 completely in about 5-6 seconds from the time the "disk" starts its boot sequence, or read-write access at essentially RAM speeds.

    Alas, we'll start to run into this problem: current disk drive interfaces

    • by symbolset (646467)
      The end of magnetomechanical media has been drawing nigh for quite a while. When this storage comes it will still be very expensive for quite a while for no better reason than they can get a lot for it. SATA3 drives are here, as is SAS 6G. Both are 6Gbits/second theoretically. Lightpeak supposedly only starts at 10 - less overhead. This is new technology that's potentially much faster than that and there's no good reason to pretend it's a spinning disk when it's not. When we get the performance up, it
    • Alas, we'll start to run into this problem: current disk drive interfaces won't be able to keep up, unless we use the Serial ATA Revision 3.0 spec.

      Why should we use a spec made for spinning disks at all? The logical thing would be to access this as what it is: Memory. After all, you don't use SATA for reading your BIOS either, do you?

      • by MtViewGuy (197597)

        Actually, I think for compatibility reasons when these new solid-state high-capacity drives arrive they'll still be using the Serial ATA-II interface, if only for one reason: compatibility with current hardware.

        Sure it may not be as fast as RAM itself, but with no more access speed limits imposed by the speed of the spinning disk and the seek times of read-write heads, we get boot times, data access times and data indexing times that will still be many times faster than Western Digital's VelociRaptor Serial

  • Since spinning off Agilent and Avago. Always thought that those would be the divisions involved with something like this?

    Figured since then all HP did was slap far east junk into cases. Does the server/etc portion still do a lot of RnD?

  • Current microSDHC cards are already 16GB [lexar.com], so 20 GB in a square centimeter in 3 years isn't impressive at all.

    • Re: (Score:1, Informative)

      by Anonymous Coward

      Starts getting impressive when you stack those square cm layers. Something you can't do with conventional flash.

      • by Yvan256 (722131)

        If it's 20GB per 1cm square layer, they should have said so. They should also have said something like "we will probably be able to start at around 10 layers minimum, giving us 200GB per square centimeter", which is much more impressive.

  • Never mind how dense or how long it'll be until we have these devices in our gadgets. How about something about how long it'll be until they can show a functional part? There's a HUGE difference between "proof of concept" in a lab and an actual manufacturable part. Most companies wouldn't even dream of announcing their partly baked research this early.
  • From TFA:

    "We believe that [20 GB/cm^2] is at least a factor of two better storage than flash memory will [] have [three years from now]"

    I wouldn't be too thrilled with that proposition. This is an ambitious new technology introduction. It won't take much for that time frame to slip by a year or two, in which case your edge is shot and your price structure is unlikely to be competitive.

    Suppose you get 5% initial market share where the difference in feature set is "just right" for some set of early adopters. Now with 5% of the revenue base, your business requirement is to scale faster than a mature competitor sitting on 95% of the

The trouble with opportunity is that it always comes disguised as hard work. -- Herbert V. Prochnow

Working...