Results 1 to 10 of 10

Thread: Memristor Computing

  1. #1
    Join Date
    May 2008
    Posts
    55

    Thumbs up Memristor Computing

    HP is rapidly advancing in their research into memristors. Cnet and MIT just published a couple of articles on the new research from HP on memristors. Reading about just the known capabilities of them is impressive. When the methods of using them to process data (compute) as well as store it is worked out they have the potential to dramatically alter the capabilities of the computers we use.

    A new discovery by HP could lead to faster, more powerful computers and other devices in the near future.

    Hewlett-Packard's HP Labs research branch has discovered that the memristor, a new electric circuit developed by HP in 2008, has far greater potential than initially thought, the company announced on Thursday.

    Previously seen mostly as a new type of storage similar to flash memory, HP found that the memristor can also perform its own logic. Such a discovery can pave the way for chips that can both perform calculations and hold data, potentially eliminating the need for a traditional core CPU.


    HP plans new nonvolatile memory for 2013.
    HP has begun testing samples of a new kind of nonvolatile memory based on memristors--circuit elements that are much smaller than the transistors used in flash memory. The company plans to introduce the first commercial memristor memory product in three years' time.

    Memristors have been the subject of much interest because, in theory, they're capable of activity that's analogous to what happens in a synapse in the human brain. So far, however, all the experimental demonstrations of memristors have been accomplished by forcing them to behave more like transistors. Instead of switching between hundreds of states, these memristors have been made to switch between two states with a high and low resistance--a digital zero and one.

    This week, in the journal Nature, Williams and colleagues reported a major step forward for memristor logic with the fabrication of circuits capable of full Boolean logic. The circuits are still digital, but Williams says his team has "shown that anything that can be calculated on silicon can be done with memristors," and in a smaller space. Demonstrating digital logic with the devices is an important first step toward more exotic computing, says Strukov.

    The memristor circuits reported in Nature are also capable of both memory and logic, functions that are done in separate devices in today's computers. "Most of the energy used for computation today is used to move the data around" between the hard drive and the processor, says Williams. A future memristor-based device that provided both functions could save a lot of energy and help computers keep getting faster, even as silicon reaches its physical limits
    .

  2. #2
    Join Date
    Feb 2008
    Posts
    210

    Default

    Quite interesting piece of hardware technology. AMD is in the train too, but not Intel.

    Funny because HP and Intel are both joined in the Itanium platform...

  3. #3

    Default

    I am surprised that HP still does this kind of research. I thought it ended when they sold their assets from DEC to Intel.

  4. #4
    Join Date
    May 2008
    Posts
    55

    Default

    I find it truly amazing that a circuit element as basic as a resistor or inductor was not discovered until now, makes you wonder if there aren't others one we've missed. HP is a very big high technology company. Their lifeblood is tech development. I'm sure they have a large R&D department. The surprising part is that this element was unknown in the 21st century.

    This will be a HUGE development. You can be sure that every fab and tech company in the world is looking into making and using them. I'm sure Intel is ramping up R & D on them now, though given HP's lead and patents it won't be publicized until they're near product introduction. I would expect some patent fights as well. I expect ALL the big tech companies, Sony, Samsung, TSMC, etc started programs to development this as soon as HP announced it, if not before. It's quite possible that HP has been keeping this secret for and only made the announcement after it was discovered or leaked to a competitor.

    The first applications will be in memory, as HP said it will have products in 2013, but using memristors for logic or computation will be the biggest advance. The ability to build them in 3 dimensions is very important too as it will let them stack CPU cores and memory to get to Terabyte and hundred core levels. Imagine a computer with 16 or a hundred or more cores and a Terabyte or more of non-volatile memory contained in a single chip!!! No need for a hard drive, the RAM, GPUs, and CPUs are on the same chip and instant on. Very low power so loooong battery life too. A mainframe in your pocket. The external connection ports (USB, DVT, etc,) will take up most of the computer.

    The computer architecture will be very different as well. Big changes in OS's will be necessary to fully take advantage of all the memristors capabilities. It will be interesting to see if open or closed source comes out with the first apps harnessing it's abilities. Could be a turning point for Linux too.

    This will be a very disruptive technology. On par with the laser and transistor, maybe more so. Like them when they were first discovered, most of the uses for it have not even been imagined yet.

  5. #5
    Join Date
    Jan 2008
    Posts
    772

    Default

    Do the hysteresis resistances get high enough to make memristor-based processors reasonably power-efficient?

  6. #6
    Join Date
    May 2008
    Posts
    55

    Thumbs up

    Quote Originally Posted by Ex-Cyber View Post
    Do the hysteresis resistances get high enough to make memristor-based processors reasonably power-efficient?
    I'm not sure. I haven't seen that value mentioned. If there was a problem with that I think it would've been discussed. You need to keep in mind too that:

    Memristors are nanoscale devices with a variable resistance and the ability to remember their resistance when power is off.

    So the value of resistance at current shutoff is more important than some arbitrary number. They're very new, so there's a lot that's still unknown. HP just discovered how to use them to make logic circuits. I haven't seen this mentioned, but since their resistance is variable each memristor can encode much more than a "one" or "zero". Potentially hundreds or even more values. Enabling a form of quantum computer you can put in your shirt pocket! They can be made very small: 3nm, so the density will be high and the voltages low. Also from the MIT article:

    "The memristor circuits reported in Nature are also capable of both memory and logic, functions that are done in separate devices in today's computers. "Most of the energy used for computation today is used to move the data around" between the hard drive and the processor, says Williams. A future memristor-based device that provided both functions could save a lot of energy and help computers keep getting faster, even as silicon reaches its physical limits."

    With memristor circuits substituting for Hard drives, RAM, CPUs, and GPUs their will be very little data moving around, at least at millimeter distances. There'll be no need to refresh the Ram either, so that'll save energy as well.

  7. #7
    Join Date
    Apr 2010
    Posts
    141

    Default

    I think artificial intelligence is a bit step closer now, if those can be used as a synapse.

  8. #8
    Join Date
    Jun 2008
    Posts
    16

    Default

    This link contains some more info

    http://highscalability.com/blog/2010...verything.html

  9. #9
    Join Date
    May 2008
    Posts
    55

    Smile

    Quote Originally Posted by Dosfish View Post
    Good Article. Fleshes out some of their possible uses and development. I find this especially interesting:

    It Talks Using Light

    The kryptonite for large pools of storage is moving chunks of data around fast enough so cooperative work can be done. The performance bottleneck is in the interconnects, when data has to flow over wires. To get around the tyranny of the wire HP is working on an optical backplane using photonic interconnects. Every time a bit has to travel more than a 100 microns it will travel as a pulse of light. Over the next 10 years they project memristors + on chip photonic interconnects will improve the overall computational throughput of a computer system by two orders of magnitude per unit of power, far outpacing what Moore's law and transistors can accomplish.


    If HP can develop a photonic backplane to use light as an inter/intra chip interconnect it would be a major feat in itself. Much faster data exchange, lower power consumption,and very little heat generated.

  10. #10
    Join Date
    Jan 2008
    Posts
    772

    Default

    Quote Originally Posted by LenS View Post
    I'm not sure. I haven't seen that value mentioned. If there was a problem with that I think it would've been discussed.
    I think it's too early for that sort of thing to really be discussed even in the trade press. If it's anywhere, it's probably in the journals/papers. I should be able to get at the Nature article once I actually get around to finding it (sometimes I wish the popular press made a habit of giving proper citations).

    Quote Originally Posted by LenS View Post
    So the value of resistance at current shutoff is more important than some arbitrary number.
    When the current is shut off, resistance becomes irrelevant. What matters more for computing is the energy it takes to change from one state to another, which depends on what the range of resistances is during switching operations.

    They're very new, so there's a lot that's still unknown. HP just discovered how to use them to make logic circuits. I haven't seen this mentioned, but since their resistance is variable each memristor can encode much more than a "one" or "zero". Potentially hundreds or even more values.
    Transistors aren't actually limited to two values either, but the more values you encode, the more complex (i.e. slow, expensive, and unreliable) it becomes to decode them and the less likely it becomes that mass production will result in a working product.

    Enabling a form of quantum computer you can put in your shirt pocket!
    I think you're confusing quantum and analog. Quantum computing requires rather specialized devices.

    With memristor circuits substituting for Hard drives, RAM, CPUs, and GPUs their will be very little data moving around, at least at millimeter distances. There'll be no need to refresh the Ram either, so that'll save energy as well.
    Outside of suspend/resume, why would there necessarily be any less data moving around than there was before? Even if everything is switched to memristors, it's likely that there will still be a memory hierarchy (i.e. it's likely that there will still be tradeoffs between access speed and density).

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •