Super-heating flash memory reportedly extends endurance

The inconvenient but unavoidable truth about flash memory is the fact that individual cells degrade every time they’re written or erased. These cells will burn out eventually, leaving users with dead drives; the only question is how long it will take.

To combat this issue, flash maker Macronix has developed a process that purportedly extends the life of the cells by heating them to extremely high temperatures. The approach uses redesigned memory chips that include tiny heaters capable of cooking the flash at around 800°C. Heating the flash and then letting it cool slowly apparently repairs the cell structure, making it possible to extend the life of the flash past 100 million write-erase cycles. For reference, the MLC NAND in typical solid-state drives is good for less than 10,000 write-erase cycles.

According to Hang-Ting Lue, a Macronix deputy project director quoted by IEEE Spectrum, heating the flash also shortens erase times. Lue says this side effect may enable a “thermally-assisted” mode that can improve both endurance and performance. Heating the flash does consume a “substantial” amount of power, but only short, infrequent bursts are required to repair the flash. Macronix plans to report its findings at the 2012 IEEE International Electron Devices Meeting later this month.

The prospect of self repairing flash memory is certainly tantalizing, especially if the endurance measures up to Macronix’s claims. However, it’s unclear how Macronix plans to bring this technology to market or when we could see it deployed in actual products. As shrinking fabrication processes produce flash with shorter lifespans, the interest in such a technology only seems likely to grow. Thanks to Engadget for the tip.

Comments closed
    • Wirko
    • 7 years ago

    Do semiconductor structures even withstand such a high temperature? I thought that they start to decompose soon above 150°C (that’s usually the maximum allowed junction temperature for power semiconductors).

      • cegras
      • 7 years ago

      Aren’t there other components besides the silicon that would suffer?

      Anyways, the article mentions that there are tiny heating elements (presumably the size of copper vias) that sit right above the flash device, probably heating by passing a lot of current in a short time through such a via.

      • ludi
      • 7 years ago

      During the annealing process in commercial manufacturing, the silicon wafers may be heated in excess of 1000C for several seconds.

      Other effects are a combination of temperature and electrical effects (e.g. secondary breakdown) or sustained exposure to high temperatures.

    • Arclight
    • 7 years ago

    Would hate to call it BS and then see it become the predominant tech for the next decade. Will wait and see.

      • just brew it!
      • 7 years ago

      Well, the physical mechanism described sounds plausible, so in that sense I don’t think it is complete BS.

      OTOH it is also entirely possible that it will turn out to be impractical from a commercial standpoint, effectively relegating it to the “vaporware” category.

        • bcronce
        • 7 years ago

        I was reading on Slashdot that the way they’re heating sounds a lot like how PCM(Phase Change Memory) heats its storage, which has had many years of development.

        It is possible that the tech required to make this main stream may already be mostly ready.

          • just brew it!
          • 7 years ago

          That’s only half the battle to making it commercially viable. If there’s a significant price premium compared to normal flash chips, most people aren’t going to bother.

      • Stranger
      • 7 years ago

      I know other industries use similar tech. for example I believe the detectors in ATLAS at the LHC are designed to be heated to fix defects induced by radiation.

      [url<]http://webcache.googleusercontent.com/search?q=cache:W1p4UkZcwgAJ:hep.ph.liv.ac.uk/~martis/indet_rad.ps+&cd=3&hl=en&ct=clnk&gl=us[/url<]

    • JDZZL
    • 7 years ago

    i’m guessing this would have to be performed as some sort of “service” from the company as everything else in the drive would likely be destroyed by the intense heat, not to mention have a power supply capable of directing that much juice…

      • Helmore
      • 7 years ago

      Read the article. Then you’ll see that this won’t even draw so much power as to really affect battery life of smartphones for example.

      • ludi
      • 7 years ago

      Bro, do you even skim?

        • bitcat70
        • 7 years ago

        Seriously…

        • JDZZL
        • 7 years ago

        Ahh, well I didn’t read the actual article and most definitly spoke without knowledge, my bad. I’ll read the entirety of articles before I do that again.

      • Meadows
      • 7 years ago

      Do not confuse temperature with heat.

        • cynan
        • 7 years ago

        But the relationship between them is often intuitive by context. (IE, The amount of energy required to raise the temperature of substance of specific type and quantity).

        Here the oversight was probably not realizing that a) the heating “elements” are only active for a few milliseconds and b) that the high temperatures are only achieved at the actual transistor gate, which collectively comprise a tiny quantity of substance to actually heat.

        Not really much to do with confusing temperature and heat in my opinion. More like failure to accurately appraise what’s actually being heated.

          • JDZZL
          • 7 years ago

          Spot on.

      • Goofus Maximus
      • 7 years ago

      It uses the same heaters as the new phase-change memory. They are very SMALL heaters, that only heat the actual junction that stores the electric charge.

      So you’re not heating up the entire flash chip, nor are the heaters turned on for a very long time at all.

    • Chrispy_
    • 7 years ago

    I’m curious how they’d stop the 800°C (localised) heating from conducting down the interconnects and melting all the solder holding the PCB components in place.

    I’d also expect a lot of magic smoke escapes from their initial prototypes.

      • chµck
      • 7 years ago

      heatsinks?
      Maybe they do this before they assemble the drive.

        • just brew it!
        • 7 years ago

        No, it needs to be done periodically to “heal” the floating gate before it wears out. But as already noted, it doesn’t need to be done to the entire chip at once, so the total amount of heat should be manageable.

        I suppose making sure other nearby components on the chip don’t get damaged may pose some challenges though…

          • Chrispy_
          • 7 years ago

          This. All this talk of “miliseconds” and “quickly dissipates” is pointless if you read the article.

          [quote<]Heating the flash[/quote<] Okay, they've got that part sussed. [quote<]and then letting it cool slowly[/quote<] meaning high [b<]sustained[/b<] temperatures that [i<]will[/i<] conduct away from the localised hotspot, even for a poor conductor like Si or SiO2. This is what I'm curious about them solving.

            • just brew it!
            • 7 years ago

            It all depends on the definition of “slowly”. We’re dealing with circuits that normally operate on a timescale of microseconds or nanoseconds. The linked article indicates that the heat pulses are only a few milliseconds long, which in this context is indeed “slow”.

            • ludi
            • 7 years ago

            Silicon annealing requires a temporary excursion to high temperature, then a comparatively slow cooling to prevent unequal stresses from forming. But as JBI says, this self-healing process is being discussed in terms of tiny fractions of a second.

            Speaking from experience with hand soldering at 350-400C and occasionally higher, the heat propagation time — even down a tiny wire — is not instantaneous, nor does it immediately concentrate in the nearest solder joint if there are other places for it to go.

      • willmore
      • 7 years ago

      Because the area heated would likely be a sector or block–which is to say a very small portion of the chip. So, the absoute amount of heat energy required would be very low and would easily dissipate to the surrounding chip area. Si and SiO2 are pretty poor conductors of heat, so it’s not hard to insulate these sections from each other.

      Also, as others have pointed out, this isn’t done often–probably when the FLASH controller detects a region is getting too many bit errors, that chunk could be scheduled to be reallocated and baked.

      They might have to use other metalizations for the lower layers–like tungsten, maybe. The thermal isolation problem doesn’t strike me as all that hard to solve. With billions of dollars of value at stake I’m sure some clever team will solve it. 🙂

      • ludi
      • 7 years ago

      [quote<]The modified structure enables current to pass through the transistor’s gate to generate pulses of heat a few milliseconds long. Researchers found that temperatures exceeded 800 °C but that the hot spot was restricted to the area near the gate.[/quote<]

      • GTVic
      • 7 years ago

      They would be heating the semi-conductor area. It heats up in a very short time so there may not be much transfer.

      Chip makers are experts in this area as the chips have to be designed to withstand the heat of the reflow solder process during the assembly process.

      There are issues of maximum temperature and maximum rate of change of temperature that they can withstand. This is a much higher temperature BUT applied over a period of milliseconds vs the 3-5 minutes that reflow soldering takes.

    • MadManOriginal
    • 7 years ago

    800C…going to need a little more than the case of a 2.5″ form factor drive for heatsinking this one!

      • wierdo
      • 7 years ago

      It’s only heating a tiny area, I don’t think it’d be that bad considering we’re talking about “x atoms” measurements, more so by the time this actually reaches market if things work out.

      Plus an area that small should dissipate heat extremely rapidly so not even a hazard concern one would imagine.

        • ChronoReverse
        • 7 years ago

        More importantly, this isn’t something that needs to be done very often or even across the entire chip at once.

          • bcronce
          • 7 years ago

          Or very often.

            • jensend
            • 7 years ago

            Or even across the entire chip at once.

        • MadManOriginal
        • 7 years ago

        Do I need to start putting <humor> tags on my posts? 🙂

          • willmore
          • 7 years ago

          If you do, keep in mind that some parsers ignore that tag.

          • cynan
          • 7 years ago

          Well, I guess that all depends on how (overtly) humorous they are ;-P

Pin It on Pinterest

Share This