Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
Hardware Technology

NAND Flash Shrinks To 15/16nm Process, Further Driving Prices Down 60

Posted by Soulskill from the smaller-better-faster-cheaper dept.
Lucas123 writes: Both Micron and Toshiba are producing NAND flash memory based on 15 and 16 nanometer process technology, which reduces die area over a 16GB MLC chip by 28% compared with previous die technology. Additionally, Micron announced its upcoming consumer USB flash drives and internal SSDs will also use triple-level cell NAND flash (a technology expected to soon dominate the market) storing three bits instead of two for the first time and further reducing production cost. The advancement in NAND flash density has been driving SSD pricing down dramatically over the past few years. In fact, over the last year, the average price for 128GB and 256GB SSDs have dropped to $50 and $90, respectively, for system manufacturers, according to DRAMeXchange. And prices for consumers have dropped to an average of $91.55 for a 128GB SSD and $164.34 for a 256GB SSD.
This discussion has been archived. No new comments can be posted.

NAND Flash Shrinks To 15/16nm Process, Further Driving Prices Down More

NAND Flash Shrinks To 15/16nm Process, Further Driving Prices Down

Comments Filter:

  • single, multi, triple (Score:1)

    by Anonymous Coward

    Whadayamean, "for the first time"? We've had three bits per cell for a while.

    Note how the prices sink while the fabs get ever more expensive. Can't be much margin left in that game. Also note that the number of available cycles per cell drops with the process density AND with the number of bits per cell. You can try and paper over it with ever more clever wear leveling tricks, but the basic problem remains the same: It's not good for data longevity.

    • Whadayamean, "for the first time"? We've had three bits per cell for a while.

      Indeed. Samsung has had TLC flash (3 bits or 8 levels) since at least 2012.

      Note how the prices sink while the fabs get ever more expensive. Can't be much margin left in that game.

      The fab cost goes up, but as the density increases, so does the capacity of the fab. So they make it up on volume.

      Also note that the number of available cycles per cell drops with the process density AND with the number of bits per cell.

      In general, it is better to have dense, yet unreliable, storage, and then fix the reliability problems with higher level error correction and redundancy. This will often give you more capacity, and more overall reliability.

      • Also note that the number of available cycles per cell drops with the process density AND with the number of bits per cell.

        In general, it is better to have dense, yet unreliable, storage, and then fix the reliability problems with higher level error correction and redundancy. This will often give you more capacity, and more overall reliability.

        Or, you can do things better, and move to vertically-stacked, 3D NAND, like Samsung. Each individual NAND chip is built on a larger process (28 nm I think) providi

        • Re: (Score:1)

          by Anonymous Coward

          Die shrinks will be coming to 3D NAND too. It's true that the flash cell longevity decreases as the process tech gets "smaller" but the increase in density makes it worth it. Any problems are mitigated with advanced wear leveling available in modern controllers, and increased provisioning of spare cells.

        • Re: (Score:2)

          by mlts ( 1038732 )

          There are always tricks to getting more bits stuffed in a cell. Each bit doubles the capacity, so (for example) if one bit causes more errors per cell, with ECC, it might be a net win.

          I do agree, eventually moving to 3D NAND storage is going to be a must, but shrinking dies is another way to increase capacity at this point and time.

          This is how Moore's law keeps up. Once one technology starts having diminishing returns, another can be used to keep things going.

      • The fab cost goes up, but as the density increases, so does the capacity of the fab. So they make it up on volume.

        Are NAND fabs much easier than IC fabs at the same mask size? There are companies offering interesting IC's that are hampered by their 60/90/130nm parts, and it seems interesting that the profit is falling out of NAND while those are still on the larger fabs.

    • Re: (Score:2)

      by swb ( 14022 )

      Doesn't improved density and lower cost also mean that they can just add more spare cells to make up for any reliability issues?

      Obviously the math has to work out in terms of the reliability decline being smaller than the increases in density and cost/GB.

    • Re:single, multi, triple (Score:4, Informative)

      by phoenix_rizzen ( 256998 ) on Friday June 05, 2015 @12:20PM (#49849595)

      Reading comprehension fail.

      The sentence clearly states that Micron will be using TLC for the first time. Not that the SSD industry will be using TLC for the first time ever.

    • Re: (Score:2)

      by PRMan ( 959735 )
      First time for Micron is how I read it.

    • Whadayamean, "for the first time"? We've had three bits per cell for a while.

      Note how the prices sink while the fabs get ever more expensive. Can't be much margin left in that game. Also note that the number of available cycles per cell drops with the process density AND with the number of bits per cell. You can try and paper over it with ever more clever wear leveling tricks, but the basic problem remains the same: It's not good for data longevity.

      All the belly-aching we read about all the time about reliability, data retention, wear leveling, et al - are due to the multi-bit per cell - be it 2, 3, 4 or whatever. When you have only 2 voltage levels - VIH, VIL, VOH and VOL where V*H == VDD and V*L == 0, it's easier to get more reliable memory. Note that with each of the shrinks that we've been having, internal voltage levels too have been dropping - from 5V to 3.3V to 1.8V to.... where is it now? In other words, it's far more difficult to have fin

  • Cheaper than that (Score:4, Insightful)

    by pushing-robot ( 1037830 ) on Friday June 05, 2015 @11:42AM (#49849213)

    I just picked up a Samsung 850 EVO 500GB for $149, and prices could be driven down further in the coming months as 32-layer and 48-layer chips show up.

    It's getting harder to justify spinning disks at home, especially as the traditional data hogs (backups, videos) are largely moving to 'the cloud'.

    • Yeah I built a NAS with 'disks' and all the new machines just get SSD for local stuff. If/when I am faced with having to be in scenarios where I'm disconnected or poorly connected for long periods of time again I'll have to reconsider, unless we're all sporting 4tb SSDs by then.

    • It is pretty incredible how the larger capacity SSDs have come down in price. $149 is in the affordable range. However, 500GB used to be sufficient to store multimedia, but that is no longer the case. Each of my daughter's school performance videos is a 30-50GB file, and they quickly add up. The cloud is okay as a backup, but for primary file storage it's more costly, higher latency, and less convenient. For non-multimedia storage, we already had affordable SSDs in usable capacities, so I'm not sure I

      • Each of my daughter's school performance videos is a 30-50GB file, and they quickly add up.

        Is that a copy for watching or for editing? An extended DVD is 8 GiB, and it uses an obsolete codec (MPEG-2). I'll grant that video production needs more disk space, but I imagine that "most" people won't be doing that. Besides, I was under the impression that external interfaces (USB 3, eSATA, Thunderbolt) have become fast enough to support editing video, so you could leave the SSD inside the case and plug in the HDD only when needed.

        • Each of my daughter's school performance videos is a 30-50GB file, and they quickly add up.

          Is that a copy for watching or for editing? An extended DVD is 8 GiB, and it uses an obsolete codec (MPEG-2). I'll grant that video production needs more disk space, but I imagine that "most" people won't be doing that. Besides, I was under the impression that external interfaces (USB 3, eSATA, Thunderbolt) have become fast enough to support editing video, so you could leave the SSD inside the case and plug in the HDD only when needed.

          Yes, I could transcode the video to something much more compact. I don't have a camcorder with a newer interface, so I wouldn't want to use it as a storage device. However, the main reason I don't do either is convenience. I already have an HDD, so I can simply copy the file. Why go to all the hassle just to fit the file into a smaller device?

          • Re: (Score:2)

            by tepples ( 727027 )

            I already have an HDD, so I can simply copy the file. Why go to all the hassle just to fit the file into a smaller device?

            So that you can carry the smaller device with you and leave the external HDD at home.

    • Comment removed based on user account deletion

    • Cap (Score:2)

      by tepples ( 727027 )

      It's getting harder to justify spinning disks at home, especially as the traditional data hogs (backups, videos) are largely moving to 'the cloud'.

      This is true so long as your home Internet connection either lacks a monthly cap or has a cap so high you're never likely to run into it. For example, people who don't torrent or stream HD video to multiple devices are unlikely to hit cable's 300 GB/mo cap. But people who rely on satellite, cellular, or Iowa DSL [slashdot.org] have to pinch their megabytes.

  • Call me when prices reach $90 for 2TB+

  • but also driving reliability down.

  • I thought - where the heck have I been that we've gotten down to sub-nanometer processes? Then I read the summary.

  • FTFS: "prices for consumers have dropped to an average of $91.55 for a 128GB SSD and $164.34 for a 256GB SSD"

    Perhaps I'm missing something, but why would an additional 256 GB flash cost $300 and an additional 768 GB cost $800 when you buy a 15" MacBook Pro [apple.com]? Or perhaps there's a boardroom in Cupertino where they're laughing and shoveling money the whole day :D

    • You usually shouldn't get your RAM or storage upgrades directly from Apple, but since they've started integrating both directly in the hardware, you can't do that anymore on some of their computers.

    • Re: (Score:3)

      by tlhIngan ( 30335 )

      FTFS: "prices for consumers have dropped to an average of $91.55 for a 128GB SSD and $164.34 for a 256GB SSD"

      Perhaps I'm missing something, but why would an additional 256 GB flash cost $300 and an additional 768 GB cost $800 when you buy a 15" MacBook Pro? Or perhaps there's a boardroom in Cupertino where they're laughing and shoveling money the whole day :D

      Well, the MacBook Pro doesn't use SATA SSDs, because SATA is bottlenecking SSD performance. SATA3 SSDs are only getting 540MB/sec - a really magic figu

  • I can get, at very low volumes from Silicon Power, 120GB for $47 and 240GB for $77. The "average" price on newegg out of the 500 or so SSDs I bought there in 2014 was $60 and $105.

Slashdot Top Deals

Keep up the good work! But please don't ask me to help.

Close