10 ms·

This is the promise of memristors. Despite innumerable articles being written about neuromorphic architectures like they'll be something miraculous, this abili

• NegativeK 9y ago
  That's why. Memristor articles were all over the news for a bit, but have died down as either companies realized they're too hard to manufacturer or companies have quietly plodded toward making them competitive.
• xellisx 9y ago
  Well with most claims that generalize computing power, such as the title of this, it's mostly aimed at something more specialized. "The researchers believe this new prototype technology will enable ultra-dense, low-power, and massively parallel computing systems that are especially useful for AI applications." So it seems that would benefit AI/Machine Learning more than GP processing, from that statement.
  • lightedman 9y ago
    The researchers are those focused in AI applications and as such I expect them to focus advertising that. This technology has far more potential than just that, right down into simple stuff like realtime waveform or spectrographic analysis. Being able to store and process on the exact same 'silicon' without having to run out a few inches to retrieve data will boost latency-sensitive applications.
  • dTal 9y ago
    I'm not so sure. CPUs are stupidly fast, now - the bottleneck is virtually always the time it takes to move data in and out of memory. What if, instead of trying to bring the data to the calculation, you brought the calculation to the data? You could speed up everything. Fun to think about: in this scenario, the main job of an "optimizing compiler" would be to organize things spatially so that, as a calculation moved around in memory, it always found itself right next to the data it needed, when it needed it.
    • coliveira 9y ago
      It makes perfect sense, because this is already how good programmers look at the task, especially considering the penalties for memory access on secondary devices.
    • austin_y 9y ago
      Brings to mind images of Conway's Game of Life, with computations moving about a grid of memory as they create and destroy new bits of data.
      • zardo 9y ago
        Or Core War. https://en.m.wikipedia.org/wiki/Core_War https://en.m.wikipedia.org/wiki/Core_War
      • Someone 9y ago
        And that brings to mind http://www.conwaylife.com/w/index.php?title=HashLife http://www.conwaylife.com/w/index.php?title=HashLife, an example (of many) of an algorithmic improvement that gives a way higher speed up than parallellizing and throwing hardware at a problem.
        • tripzilch 9y ago
          Hashlife is a brilliant algorithm. For anyone not familiar (and vaguely interested in Conway's GOL), highly recommend reading up about it. I found it mildly mindblowing :)
    • 9y ago
  • otakucode 9y ago
    AI stuff would be easy to implement on it, but general computation will almost certainly benefit just as much once research is done. It's a fundamentally different architecture when your assembly language can include instructions that essentially re-wire the FPGA which is your processor/storage. That makes possible entirely new classes of algorithms.
• farresito 9y ago
  I can't talk for other people, but I personally don't try to get over-excited over anything unless I see a working proof of concept where the advantages are clear. It's kinda unfortunate, because there's great work being done in research that I should probably feel much more excited about, but I'm continuously bombarded with new discoveries that I never hear about again.
  • stcredzero 9y ago
    I personally don't try to get over-excited over anything unless I see a working proof of concept where the advantages are clear. Those are the two things you need to look for. No proof of concept? Well, talk is cheap! Have a proof of concept, but the advantages aren't clear? Again, talk is cheap. Refine it until the advantages are demonstrable!
  • CalChris 9y ago
    That's fine but it is a very conservative approach. I also can't talk for other people, however I personally get excited about new things without a working proof yet and where the advantages aren't clear. It makes for a lot of disappointment and isn't for the faint of heart. I'd rather see the future as it happens, false starts and all. So the bleeding edge is where I want to be.
• Illniyar 9y ago
  I thought memristors is RAM that doesn't lose state when losing electricity. Is this article about memristors? That wasn't clear. Can memristors perform computations?
  • ericd 9y ago
    I'm not an expert on this, but I think you might be thinking of nvram/non volatile RAM, "memristor" is a portmanteau of memory and transistor, so I'd expect that they have properties of both.
    • digi_owl 9y ago
      Not quite. memristors are resistors with a memory. Specifically as current is run across it, it builds up more and more resistance. Then as the current is reversed (the plus and minus ends of the resistor is switched) this resistance will go back down. Thus it can be used as non-volatile memory in the sense that certain tiers of resistance can be designated as a bit. Note that as it is fundamentally an analog component one can potentially also used this property to emulate a neuron.
      • markdog12 9y ago
        Currency? Is it a crypto-currency that runs across it? :)
        • digi_owl 9y ago
          bleeping butterfingers... Fixed.
      • ericd 9y ago
        Thanks very much for the correction! Removing my foot from my mouth, now. That's incredibly interesting, especially the idea that it could be used to emulate neurons. I expect that it'd be hard to fabricate them with a very high branching factor, but as I've amply demonstrated, I know very little about this topic.
    • 9y ago
• deepsun 9y ago
  Maybe for the same reason as people don't get more excited about quantum computers -- they can replace basically everything as well, yet we don't see them around.
  • paulddraper 9y ago
    And of course, people do get excited about quantum computers...millions and millions of $ a year excited. But yeah, their mostly non-existance is exactly why people aren't more existed.
  • ChuckMcM 9y ago
    If you open the box their wave state collapses and they are either excited or bored. :-) (ok that was a long way to make a Schrödinger joke) There are a lot of people who are very excited about Quantum computers and working on them, as they get more 'real' people on the fence move over, etc. I watched this on a smaller scale with 3D printers, as only a few people were excited about them and then as they demonstrated more and more interesting things the wave of excited people expanded outward from that core.
  • otakucode 9y ago
    Quantum computers require both very specific configuration for each problem, and cryogenic cooling. Both of those issues mean there's no reason to suspect they will find more than use as a daughtercard any time soon. Memristors, on the other hand, could replace CPUs, RAM, SSDs, bulk storage, FPGAs with one consistent lattice, a totally unified memory architecture, and an economy of scale that I don't believe has ever existed before. And that's BEFORE you do any of the exotic research of moving the computing elements around during operation. Just imagine if there was no memory hierarchy, where accessing a byte in mass storage is as fast and as easy as accessing a register. Quantum computers can't replace commodity hardware and there's no reason to think they will. Memristors, on the other hand, are one of the fundamental circuit elements and are destined to become as widespread as resistors, capacitors, and transistors.
    • adrianN 9y ago
      The speed of light dictates that there will always be a memory hierachy. A nanosecond is just a couple of inches of wire [1]. Communication overhead and heat issues are what ultimately limits computational power [2]. [1] http://www.youtube.com/watch?v=JEpsKnWZrJ8 http://www.youtube.com/watch?v=JEpsKnWZrJ8 [2] https://en.wikipedia.org/wiki/Limits_of_computation https://en.wikipedia.org/wiki/Limits_of_computation
• kensai 9y ago
  The best thing about memristors is that actually this is how the brain really works (at an electronical equivalence). There is no von Neumann architecture in our cortex, rather neurons in networks which both compute and store memories.
  • stephengillie 9y ago
    Neurons are routers that apply weights to their inputs.
    • kensai 9y ago
      I talked about the bigger picture, neurons IN NETWORKS.
• nyolfen 9y ago
  > I've never understood why people don't get more excited about memristors. personally, mostly because i've been seeing articles about how they're just about to totally upend computing for the last ten years
  • cududa 9y ago
    Pretty sure I first read about them in a PC World magazine I got at the airport on a family vacation in like 2003
    • Alex3917 9y ago
      As far as I know I submitted the first ever article about memristors to HN, which was this in 2008: https://news.ycombinator.com/item?id=177865 https://news.ycombinator.com/item?id=177865
      • 1024core 9y ago
        There was a world of media before HN, you know. Just saying.
        • andrepd 9y ago
          I don't think he was implying otherwise.
      • squeaky-clean 9y ago
        HN has only been around since 2007
        • kbenson 9y ago
          Ambiguous language. I thank what was being implied is that it was the first occurrence of that type of article on HN, not that the article was the first ever about memristers and was posted to HN.
• jerf 9y ago
  "this ability to change from functioning as a bit of memory to being a bunch of functional logic on the fly at the speed of a memory read?" My excitement is tempered by considering the areal demands on the silicon for the putative "smart memory". Suppose, just for the sake of argument, you want your smart memory to be able to take a 4K block of 64-bit integers and add them together. It happens incredible quickly, sure, though you'd have to get an expert to tell me what the fanin can be to guess how many cycles this will take. But you're now looking at massively more silicon to arrange all that. And adding blocks of numbers isn't really that interesting, we really want to speed up matrix math. Assuming hard-coded matrix sizes, that's a whackload of silicon per operation you want to support; it's at least another logarithmic whackload factor to support a selection of matrix sizes and yet more again to support arbitrary sizes. In general, it's a whackload of silicon for anything you'd want to do, and the more flexible you make your "active memory" the more whackloads of silicon you're putting in there. It may sound weird to describe a single addition set of gates as a "whackload", but remember you need to lay at least one down for each thing you want to support. If you want to be able to do these operations from arbitrary locations in RAM, it means every single RAM cell is going to need its own gates that implement every single "smart" feature you want to offer. Even just control silicon for the neural nets is going to add up. (It may not be doing digital stuff and it may be easier to mentally overlook, but it's certainly going to be some sort of silicon-per-cell.) Even if you were somehow handed a tech stack that worked this way already, you'd find yourself pressured to head back to the architecture we already have, because the first thing you'd want to do is take all that custom silicon and isolate it to a single computation unit on the chip that the rest of the chip feeds, because then you've got a lot more space for RAM, and you can implement a lot more functionality without paying per memory cell. And with that, you come perilously close to what is already a GPU today. All that silicon dedicated to the tasks you aren't doing right now is silicon you're going to want to be memory instead, and anyone making the smart memory is going to have a hard time competing with people who offer an order of magnitude or three more RAM for the same price and tell you to just throw more CPUs/GPUs at the problem. RAM that persists without power is much more interesting than smart memory.
  • vvanders 9y ago
    Yeah, this seems like the exact same problems that we already have with FPGAs. 1. They're really hard to program for in a way that is easy to understand and scale. 2. They eat a ton of power. Heat = power = max speed. If you can't make it better then today existing ASICs are still going to be used. The intersection of ASIC + reconfigurable serial processes(aka sequential programming) strikes a really sweet point between power and flexibility that I think is going to be hard to unseat. I think FPGAs are incredible but if this were true I think we'd already see them taking over the world.
• AndrewKemendo 9y ago
  I've never understood why people don't get more excited about memristors. Probably because they have been promising them too long. I remember watching a video lecture in 2008 about how IBM was on the verge of revolutionizing everything with them. Meanwhile in the same timespan GPU/DNN have upended the computing landscape.
• BugsJustFindMe 9y ago
  > I'm still waiting to be able to buy a 2000 ppi display like IBM's R&D announced creating back in the late 1990s or so... You're not thinking of this and mistakenly adding an extra 0, are you? https://en.wikipedia.org/wiki/IBM_T220/T221_LCD_monitors#History https://en.wikipedia.org/wiki/IBM_T220/T221_LCD_monitors#His...
  • hinkley 9y ago
    They were right. Retina displays aren’t even the high PPI ones anymore. They are the equivalent of second generation laser printers.
    • BugsJustFindMe 9y ago
      > They were right. Who was right about what? > They are the equivalent of second generation laser printers. What does this even mean?
      • hinkley 9y ago
        Roentgen and Retina displays are in the neighborhood of 200 ppi, which is about when laser printers became 'clear' and legible, which was the thesis behind Roentgen. But now you've got smart phones pushing in the region of 400 PPI. It's smoother, and make Apple displays look like laggards. But they're beyond the point where most eyes notice the difference without magnification.
        • giobox 9y ago
          The PPI required for the "Retina" effect is largely dependent on viewing distance, which is why Apple vary the PPI according to the device, they aren't a blanket ~200 PPI as you suggest. The Apple Watch and most iPhones (including the original Retina display on the iPhone 4) are around 300-400 PPI, with the X being 458. Only Apple's "Retina" laptop/iMac displays have been around 220, which normally looks great given the increase in the viewing distance. Wikipedia has a pretty nice breakdown of the PPI of all Retina displays Apple has shipped: https://en.wikipedia.org/wiki/Retina_Display https://en.wikipedia.org/wiki/Retina_Display
• tasty_freeze 9y ago
  Transistors provide something that resistors, capacitors, inductors and memristors don't, and thus they can't be replaced. That thing is gain, ie, amplification (of either voltage or current). All the others are lossy, so if you imagine some building block, then cascade N of them in a row, eventually the signal will become too weak (at some N) to be useful.
  • hinkley 9y ago
    Do you remember when IBM sold off their hard drive research division? I was convinced this meant they knew memristors were going to kill hard drives soon and they were profit taking before killing them off. That’s been more than ten years now. I am still waiting, but made other plans.
    • deelowe 9y ago
      In all fairness, the hd industry as a whole has shrunk considerably since. Same for servers and PCs. I think they made the right decision, but I'm not sure I'd go as far as to say they had a good strategy for what they would do once they got rid of these product lines.
    • Aloha 9y ago
      They sold it when hard drives became a commodity item - generally commodity items don't get major R&D attached to them.
      • manigandham 9y ago
        There's still plenty of R&D being done with storage: https://www.anandtech.com/show/11925/western-digital-stuns-storage-industry-with-mamr-breakthrough-for-nextgen-hdds https://www.anandtech.com/show/11925/western-digital-stuns-s...
• sliverstorm 9y ago
  this ability to change from functioning as a bit of memory to being a bunch of functional logic on the fly at the speed of a memory read? That's going to be crazy. It will open up possibilities that we probably can't even imagine right now. There might be clever ways to make a particular circuit dual purpose or something (Storage element that can do math on itself? A math pipeline that doesn't need pipeline flops?) But arbitrary reprogrammability is already here. It's called an FPGA, and it sees only niche use, not in the least because developing a bitstream to program it with is a huge chunk of work. So the days of a chip that constantly morphs like a chameleon to unrecognizable new forms is probably a long way in the future. I like to remind myself the industry is called VLSI- very large scale integration. In other words, the job is not about transistors. More than anything else, it's about managing the complexity of billions of transistors. Digital CMOS is a great example; transistors can already do so much more than CMOS, but it's a massively simplifying design scheme.
  • otakucode 9y ago
    FPGA reprogramming is slow. It's like saying we can make more oil because all we have to do is wait a few hundreds million years for all our plant matter to rot. And correct me if I'm wrong, but can you even reprogram an FPGA while it is actively being used? I don't believe so. Much less reprogram it from the gates which are active.
    • blattimwind 9y ago
      > And correct me if I'm wrong, but can you even reprogram an FPGA while it is actively being used? You're wrong, this is called partial reconfiguration.
      • smilekzs 9y ago
        Can you reprogram the part of an FPGA that is actively being used?
        • CamperBob2 9y ago
• coldtea 9y ago
  >I've never understood why people don't get more excited about memristors For the same reason they aren't excited about flying cars? Because they wait until something is actually delivered that in a form they can use?
  • eradicatethots 9y ago
    Excitement happens before things are delivered by definition. This comes off as a pretty rude comment to me, with the excessive question marks implying that obviously people aren’t so stupid as to get excited about anything that isn’t 100% guarenteed to happen. Would like to see less comments like this on HN.
• deepnotderp 9y ago
  Well, first off, their endurance is currently not even good enough to replace DRAM, let alone replacing critical path transistors. They also take significantly more power to switch and are much slower. They also tend to be highly unreliable (they even have variable write time!) and hard to manufacture. Really, the only useful advantage they do have is their theoretically incredible density, but even most of this comes from the possibility of 3D stacking (also possible with transistors) and from the possibility of a crossbar memory, but crossbars have epic sneak path issues.
• the_jeremy 9y ago
  I was not technologically inclined in high school, but I distinctly remember a peer talking about how cool they were and how much they could be used to do. This article is the second time I've ever heard the word. So... I guess the fact that it's been 7 years and they're still just "going to do awesome things" is probably a good reason why.