Toshiba to start producing second-gen 19-nm NAND this month

Later this month, Toshiba will begin mass-producing NAND using a second-generation 19-nm process. According to the press release, the 64Gb (8GB) chips will be the world’s smallest with two bits per cell. Each one will have a die size of 94 mm², which appears to be 17% reduction from the previous generation. The presumably gen-one 19-nm chips discussed in this Toshiba presentation (PDF) from last year have the same 64Gb capacity and a 112.8 mm² die area.

Thanks to a “unique high speed writing method,” the new Toshiba NAND can program bits at up to 25MB/s. That might not sound like a lot, but the figure applies to a single chip. SSDs achieve their speedy transfer rates by addressing multiple chips in parallel, and a 256GB drive would have 32 individual dies accessible to the controller.

In addition to producing MLC chips with its updated 19-nm tech, Toshiba plans to use the process to crank out TLC NAND with three bits per cell. Mass production of that memory will begin in the firm’s second fiscal quarter, which ends in September. The TLC chips will first be sold in eMMC devices meant for smartphones and tablets. They’ll eventually make their way into solid-state drives targeted at PCs. Thus far, the only TLC-equipped SSD we’ve seen is Samsung’s 840 Series.

We first spotted Toshiba’s 19-nm MLC NAND in Plextor’s M5P SSD last year. Since then, the flash has popped up in a refreshed version of Corsair’s Neutron GTX and in Toshiba’s own SSDs. It will be interesting to see how long it takes the second-gen stuff to make its way onto the market—and whether the new chips will allow drive makers to cut prices even further.

Comments closed
    • ronch
    • 7 years ago

    THIS MARKS THE END OF INTEL’S PROCESS TECH LEAD!!!

    (I love my CAPS LOCK key too, except I’m using a tablet to post this so it’s just a virtual CAPS LOCK.)

    • smilingcrow
    • 7 years ago

    Assuming a 17% decrease in size leads to a 17% decrease in NAND cost the question is what percentage of the retail price is down to the actual NAND?
    It might be as low as 50% which would allow around a 8.5% reduction in retail cost.

    I know it’s not that simple as smaller die sizes leads to more efficient use of wafer space but yields might be lower initially to counterbalance that.

      • UberGerbil
      • 7 years ago

      Actually, most of the price of an SSD is the NAND. Spot prices bounce around a lot, but you can check them at various places like [url=http://www.dramexchange.com/<]DramExchange.com[/url<]. The SSD makers don't buy in the spot market, of course, but lock in contracts in advance (always risky with a depreciating commodity but you do what you have to do), and companies like Samsung and Intel/Micron have a captive supply so spot prices don't affect them at all. But I'd guess most of the SSD vendors are currently spending something like $5-$6 per 64Gb NAND; you need 16 of those to get 128 (or 120) GB, so you see that $80-$90 of the SSD price is NAND. They're probably not (yet) paying as low as $4 per chip, but even then it's more than half the total price.

        • smilingcrow
        • 7 years ago

        It is definitely the most expensive part but those numbers don’t add up for me.
        Plenty of 240GB drives at $180 which even at your lower figure of $5 would mean that the NAND was costing $150. That leaves $30 for everything else: controller, other components & packaging, manufacturing, design, validation, marketing, distribution, retailer/wholesaler profit and manufacturer’s profit.

        Spot prices have nothing to do with the actual manufacturing cost but is just another component that can potentially affect the retail price like inflation; albeit traded commodities are much more likely to show downward pricing pressure than inflation. Hence I’m ignoring it in this context.

        • willmore
        • 7 years ago

        Just having a captive supply doesn’t isolate them from spot prices. Someone at the company has to decide “Do we make SSDs with this FLASH or do we sell it on the spot market?” If they’re not doing that, then they’re not doing a good job of optimizing profits.

          • smilingcrow
          • 7 years ago

          I wonder how much of the total NAND production is traded via the spot market(s)?
          Sure the manufacturers can trade excess NAND on the market but if they start dumping on the spot market because prices are high that will normally bring the price down anyway.

    • brute
    • 7 years ago

    first pearl harbor, now this

    when will it end

      • SnowboardingTobi
      • 7 years ago

      wtf are you going on about?? If you’re trying to be funny… at least be funny. smh

        • brute
        • 7 years ago

        wow

      • smilingcrow
      • 7 years ago

      It already has; December 2012.

    • internetsandman
    • 7 years ago

    For some reason I started out reading this article thinking it was talking about DRAM and not NAND

    I got really excited when I saw 8GB chips

    • willmore
    • 7 years ago

    I’m curious to see if they’ve improved the write cycle longevity over the first gen 19nm parts. Especially for TbC.

      • UberGerbil
      • 7 years ago

      That’s always a concern with NAND shrinks. Some of the costs savings may be eaten up by extra reservoir bits to get write endurance where they want it.

        • willmore
        • 7 years ago

        Yeah, write speed and write margin are generally linked, so I’m hopeful that the latter has been imporved as we are being told that the former has.

    • Star Brood
    • 7 years ago

    [quote<]...whether the new chips will allow drive makers to cut prices even further[/quote<] Agreed. That is the real thing we are all watching for, isn't it?

      • CB5000
      • 7 years ago

      Yeah me too. I still only own a single 128gb Samsung 840 pro on a computer where “speed matters”…on my gamer wife’s computer to mitigate slow startup and load speed related complaining and nagging.

    • Mourmain
    • 7 years ago

    32 * 25 = 800 MB/s, in case anyone is trying to do the math.

      • hoboGeek
      • 7 years ago

      Of course, this is in theory. Not to mention, the bus must be able to support this speed

        • albundy
        • 7 years ago

        well, it seems like the z87 makers fell asleep at the sata express table, cus it could have supported it, and from what i understood, it should have been included.

      • UberGerbil
      • 7 years ago

      Not to mention, they may not put 32 channels in parallel — that’s going to be up to the implementer (and the controller they use). It’s quite possible to have just 16 channels, for example, and essentially do bank addressing to get higher capacities — or just wait until larger capacity chips are available to up the capacity of the drive. In fact, if the chips are that fast for writes, it would make a lot of sense (for the vendor) to do that: adding more channels is expensive, and writes are already fast enough for a lot of mainstream uses; at this point for non-server uses making a drive cheaper rather than faster probably has a better business case (certainly there have been commenters here in past SSD threads who have loudly requested that trade-off).

Pin It on Pinterest

Share This