back to article Never-never chip tech Memristor shuffles closer to death row

Martin Fink’s HPE Labs has been dangling the Memristor in front of us for years. With Fink retiring and HPE Labs losing its independence, becoming part of Antonio Neri’s Enterprise Group, inventing far out blue sky stuff will likely shift to devising technologies that can be realistically productised. The Memristor cannot. The …

  1. Mage Silver badge

    Memristor

    Deader than bubble memory.

    It at least actually got produced.

    1. John Smith 19 Gold badge
      Unhappy

      Deader than bubble memory.

      Bubble's issues were

      a) Very demanding (for the time) line widths

      b) Unusual (ceramic) substrates.

      c)Access times

      There were developments in the pipeline at Bell Labs that would have fixed the materials issues access time problems and time would taken care of the lithography.

      But time was one thing they didn't have

  2. inmypjs Silver badge

    What he and we..

    were saying 3 years ago

    http://www.theregister.co.uk/2013/11/01/hp_memristor_2018/

    same photos even :)

    1. Anonymous Coward
      Anonymous Coward

      Re: What he and we..

      Would love the idea of a story timeline, especially for "evolving" tech like this!

      From cutting edge technology to technology that got cut!

      1. DropBear
        Facepalm

        Re: What he and we..

        Unfortunately ALL cutting edge technology is like that. If you're first reading about "X" on a tech news site, it has no future. If you first see it show up in a product leak / ad / review, then it actually exists and has a future. Next up: boffins announce revolutionary new battery technology - and be sure to tune in same time next week for our in-dept look at fusion getting closer and closer... </sarcasm>

  3. Anonymous Coward
    Anonymous Coward

    The End of The Machine?

    Is this the end of the project from the sci-fi division, The Machine?

    1. Anonymous Coward
      Anonymous Coward

      Re: The End of The Machine?

      They've been playing Pink Floyd and talking about other memory technologies as a "step gap" until the memistors become available. There's been talk of a public demonstration machine running this year.

    2. Anonymous Coward
      Anonymous Coward

      Re: The End of The Machine?

      What are we going to rage against without The Machine?

      1. asdf

        Re: The End of The Machine?

        >What are we going to rage against without The Machine?

        No worries never a shortage of vaporware out there. On the wintel front Midori the wholly managed code OS and Larrabee the GPGPU were my recent favorites.

  4. Nigel 11

    Best Tech?

    Looks like it'll be Intel's Xpoint and/or other phase-change memory that reaches production. Does anyone know whether there's anything fundamentally better about Memristors over these others? If not, then it's just a race to production, with HP falling behind and dropping out.

    If HP has lost interest and if there is a fundamental advantage, one would expect some other company to buy the patents.

    1. asdf

      Re: Best Tech?

      > Does anyone know whether there's anything fundamentally better about Memristors over these others?

      Well some of the others have the advantage to actually work even after mass production.

    2. imanidiot Silver badge

      Re: Best Tech?

      Actually being able to mass produce the stuff for one.

      Real world litho processes are surprisingly finicky and what works in small proto batches might not work on full blown production. Memristors from my understanding are just too hard to produce large scale to be worth the effort.

  5. TeeCee Gold badge
    Black Helicopters

    Memristor is dead - edged off the table by XPoint.

    "Hmm, if HP ever get this Memristor stuff out, that ocean of cash we've hosed over XPoint will be wasted. Call HP and tell 'em they can go whistle for Itanium updates if this ain't killed now."

    1. Dazed and Confused

      Whistling

      > Call HP and tell 'em they can go whistle for Itanium updates if this ain't killed now."

      There ain't no new Itanium updates coming, we've had the last one already.

      I think HP already paid Intel for the last couple of revs anyway.

      One big issue I'd have thought was that HP doesn't have any fab capability. Didn't that sort of thing go to Agilent in the big split all those aeons ago?

      Surely HPE would be better off trying to persuade someone like Samsung to turn the dream into a reality. They got the expertise on actually making bit of silicon these days.

      1. Charles 9

        Re: Whistling

        Except Samsung's money right now is on its homegrown 3D Flash tech, which is rolling out right now and is going to be the tech in demand for the short term, giving them time to see which post-Flash technology matures first to close the capacity-price gap (since bulk capacity is the one niche Flash can't fill in a cost-effective manner at present, 3D Flash might close that gap, though).

    2. abufrejoval

      too good to live

      The videos with Martin Fink and Meg Whitman promising usable Memristor DIMMs in 2016 is still on YouTube (https://www.youtube.com/watch?v=33WPphbFeDY) and I'm quite astonished both are still around after failing to deliver.

      I found the stony faces of the Intel and Microsoft CEOs especially telling towards the end of that video, where Meg happily invited the other American IT giants to follow HP's technology lead.

      If the Memristor promise were to become true in every aspect just as Stan Williams hinted, it quite obviously would have destroyed more IT market value than it would have created. Essentially you'd have to rename Silicon Valley to Titanium Dioxide Valley, remove all spinning rust and dynamic from memory, rewrite all transistor logic in memristor cross-latches and generally go back to a single level address space whilst we wait for neuromorphic computing eat up what's left of IT niches beyond SAP and Excel.

      On the other hand, when you look at the price and performance points of 3D Xpoint, slotting almost perfectly in between DRAM and Flash, you can't help thinking that disruptive technology never naturally lands exactly where the sales guys want it and that we're being played, chipzilla scale.

      It's quite clear Intel doen't want the Memristor (unless it's theirs), Micron wants what Intel tell them, Samsung would love the memristor without paying and Hynix is struggling badly to stay alive.

      I hope that we'll hear the story behind the "failure" of the memristor eventually and my feeling is that it didn't fail for technical reasons, but was stopped because too much IT spending would have become unnecessary.

      And where IT giants suffer, consumers must suffer, too, right? So Intel did the proper thing (and product placement), where HP failed to respect it's position in the pack and it's not even an anti trust issue!

      Intel quite cleary said that they wouldn't support any other NV-RAM technology than Apache Pass aka 3D Xpoint in their x86 chipsets and CPUs just like you won't see Power-8 or ARM supported 3D Xpoint DIMMs, because Intel has the right to protect its home territory.

      It would probably take a China size opponent to challenge that.

      The concept of hundreds of terabyte of non-volatile DRAM speed storage in a smart phone form factor certainly got my creative juices flowing.... and then left me with few meaningful outlets, because no current battery would survive a single compute pass across the complete range.

      1. asdf

        Re: too good to live

        >and I'm quite astonished both are still around after failing to deliver.

        Well Meg still hasn't bought a company (AFAIK) that they had to write almost all of its value down immediately so I guess its more a matter of the bar being pretty low. I do miss the days of the incompetent board members getting arrested, the CEO being grabby with the help or telling the world they no longer make PCs, etc. The LOLHP days were more entertaining.

      2. Matt Bryant Silver badge

        Re: abufrejoval Re: too good to live

        "....It's quite clear Intel doen't want the Memristor...." That's a bit harsh on Intel. What this actually is (hopefully) is hp learning the lesson that it can't buck the market. Remember hp's HyperFabric? They got bored waiting for a clustering interconnect standard and so made their own - great tech, eventually stable and fast, but hideously expensive due to it only selling with hp's servers. But hp persisted, pushed on with a faster development, HyperFabric2, even in the obvious threats of cheap 10Gb Ethernet and Infiniband (some would say in denial of). At the time they ranted on about it being a technically better product, but they didn't convince any other vendors to join the HyperFabric party. They made the mistake of not getting partners onboard and HyperFabric became the Betamax to a range of cheaper and more widely available offerings. Unfortunately, without a major partner onboard (especially Intel's preference for their own solution), the Memristor has become a scientific wonder zombie, and maybe it's better that someone at hpE makes a decision based on market logic and pulls the plug now.

  6. luis river

    mass production

    The Tech Memristor speciment device "Works", only this tech is faulty on mass-bulky production, I dont doubt what HPE- others to reach this final purpose

  7. EddieD

    WTFingF?

    "...that can be realistically productised"

    PRODUCTISED??? What the fuck? "realistically produced"

    Journalists should know and love their language, not invent synthetic words to appease a market fragment. (I noticed "ised" not "ized" too, which is even more painful - if a Brit is coming up with pathetically crap neologisms like this, it's too late for heaven help us, Satan is already on the march)

    1. Charles 9

      Re: WTFingF?

      Ever thought the writer doesn't have English as a first language?

    2. Pascal Monett Silver badge
      Coat

      Re: WTFingF?

      Satan has preceeded you and his marketroids have done their best to nullify your argument by taking control of the Cambridge dictionary.

      Once he controls the Oxford dictionary, his dominion will be complete, the portal will open and we will all be doomed.

    3. deeso

      Re: WTFingF?

      It does feel rather awkward on the tongue. How do you feel about 'produlate'?

      "...engage in the realistic produlation of technologies"

      1. Anon

        Re: Produlate

        I don't think it'll stick.

      2. Ralph B

        Re: WTFingF?

        > How do you feel about 'produlate'?

        Professor Nebulous? Is that you?

    4. CRConrad
      Holmes

      WTFingF yourself.

      EddieD goes: "PRODUCTISED??? What the fuck? 'realistically produced'"

      Different concepts.

      "Produced" = Made; manufactured. "This processor was produced, and this one, and... Thousands of them every day in this factory."

      "Productised" = Made into a product; developed from theoretical invention into something actually (technically, profitably) producable. "That one thing that was productised last year, we produce thousands of it each day now."

      (There, now go off on a rant about "producable", whydoncha.)

  8. Dave 13

    Honey, I shrunk the company..

    As EMC has shown with the DSSD product, you don't need to reinvent memory, just intelligently use what already exists. HPE has shrunk itself to just another box vendor - something the world can happily live without.

  9. SeanC4S

    Certainly very poor choices where made in allocating research money. It was not exactly a risk managed balanced portfolio. But others have actually delivered, such as Micron with its automata chip and then completely failed to market the product. Weird things happen in large companies and they often are not efficient.

    Anyway, whatever you can give me in terms of memory products I can use. I can make vast decision trees with random projection neural predicates.

  10. Christian Berger

    Actually that's not the reason why you should be worried

    The reason is that it apparently cannot get traction in areas where you could already use memristor memory. After all you can already buy 16 kilobyte chips, and there probably are usecases for this.

  11. dubina

    This topic should be refreshed.

    https://ece.umass.edu/news/research-opens-way-low-cost-mass-production-silicon-based-memristors-building-blocks-next

    Any thoughts?

    1. Charles 9

      Let's wait until a serious firm actually employs the process to produce actual large-capacity memristor chips and puts them through the wringer.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like