back to article Micron's glass memory monster chews up slowcoach flash

Micron has demonstrated Phase-Change Memory (PCM), enabling an app to run around 50 times faster than it would on NOR memory. PCM is a post-NAND and post-NOR contender in the non-volatile memory arena; it's said to combine the speed and simple data access of DRAM and the non-volatility of NAND and NOR flash memory. It stores …

COMMENTS

This topic is closed for new posts.
  1. ideapete

    I always wondered What my wife has in her head

    Must be Phase-Change Memory (PCM) its like lightening changes her mind in milliseconds

    1. TeeCee Gold badge
      Happy

      I wonder if it's got the same error correction system too?

      If a discrepancy is detected, all the conflicting versions held elsewhere in the world are assumed to be invalid.

  2. This post has been deleted by its author

  3. david 12 Silver badge

    more info required

    Does it have a lifetime limit? How does the read spead compare? Does it require queing and look-ahead (like both flash and dynamic RAM) to achieve read speed? What is the timetable?

  4. Hungry Sean
    Boffin

    ITRS is your friend

    international technology roadmap for silicon is put together by reps from various major silicon manufacturers. They all put on their mystic turbans, stare into their crystal balls, and create from a common vision of the future. It's a great resource for detailed information about next-generation technologies and the expected demise of existing ones.

    This document, starting on page 13, goes into their most recent predictions on capabilities of different non-volatile technologies, and details PCRAM on page 19. Another major revision will probably be out next year which should have more up to date details, but:

    http://www.itrs.net/Links/2009ITRS/2009Chapters_2009Tables/2009_PIDS.pdf

    I thought the most interesting bit is this:

    The major challenges for PCRAM are the high current (fraction of mA) required to reset the phase-change element, and the relatively long set time. Since the volume of phase-change material decreases rapidly with each technology generation, there is hope both the above issues become easier with scaling.

    A technology that becomes easier to use with smaller scales is pretty darn neat if you ask me. . .

    1. John Smith 19 Gold badge
      Unhappy

      @Hungry Sean

      "A technology that becomes easier to use with smaller scales is pretty darn neat if you ask me. . ."

      Not really.

      The actual question is does the power requirement scale *faster* than the thickness of the wires carrying it.

      If it does not you also have to produce power conductors with higher aspect ratio to carry the *same* power you needed for the last generation.

      1. Hungry Sean
        Pint

        good point

        I'm not a circuits or process expert, so can't say for sure, but it sounds like the change in power is driven by a change in volume of material with scaling. Assuming the depth of the material isn't changing, that'd be a quadratic reduction in power to go along with what would presumably be a linear reduction in supply current from reduced wire cross section (I'm pretty sure metal depths are not being reduced in general, so it should just be the thickness).

  5. Anonymous Coward
    Anonymous Coward

    Wake me up when someone makes

    Terabyte memory chips using this material.

    What would be nice is a write once read many (WORM) material using this, the plan being that people can download their data onto it and then to "erase" just overwrite with 1's.

    When it runs out of space buy another chip, hopefully the cost will be lower than Blurays.

    Seems that high density flash sacrifices rewrite longevity , so why rewrite at all?

    Just write once and do error checking, use the erase circuitry area saved for more cells.

    AC/DC

    1. Anonymous Coward
      Anonymous Coward

      "What would be nice is a write once read many (WORM) material using this, the plan being that people can download their data onto it and then to "erase" just overwrite with 1's."

      If it only lets you write once, how do you "overwrite" anything?

      1. Toby 10

        I assume it starts as all zeros.

        So you write ones where necessary and then change all the spare 0's to 1's when you want to trash it.

        Of course this makes the slightly crazy assumption that that is how it all works.

  6. a_mu

    how things change

    once there was eeprom,

    450 ns cycle times,

    then flash,

    45 ns cycle times,

    now pcm,

    4.5 ns cycle times,

    works for me

This topic is closed for new posts.

Other stories you might like