Comments Locked

25 Comments

Back to Article

  • lilmoe - Tuesday, July 10, 2018 - link

    30% better manufacturing productivity... Wonder how much of that will be passed down to the consumer as savings. If it's less than 20%, F U Sammy.
  • ZolaIII - Tuesday, July 10, 2018 - link

    Well cost is more dependent on the manufacturing process. If 96 layer brings 25~30% incres in capacity while cell size is down 20% it translates to 1.3x priced 1.5x capacity. 7nm FinFET first generation won't bring any price cut compared to the 10 - 14nm but second generation with denser rooting & EUV will cut the price roughly in half. So by the end of the next year you will have 3x capacity for 1.3x price compared to today advanced 64 layer one's.
  • SiSiX - Tuesday, July 10, 2018 - link

    Um, these don't use a 10-14nm, nor are they moving to a 7nm FinFET process. Part of the reason for the stacking in the first place, and the HUGE increase in overall endurance is because they're effectively using "old", very mature process size. (Their "21nm" process node, whatever that works out to be in real world terms.) By and large, Flash memory has more issues the smaller you make the features.
  • piroroadkill - Tuesday, July 10, 2018 - link

    It's not such a great idea to shrink flash; the endurance drops off unacceptably. It's why they've gone back to more mature processes and then stacked them in layers. Better density AND reliability.
  • FullmetalTitan - Thursday, July 12, 2018 - link

    Pretty sure these 96 layer parts are still on a 22nm architecture, flash doesn't follow the same roadmap as logic. Limitations like extreme aspect ratio etches to form vertical holes for connecting layers are difficult to scale.
    Samsung IS investing in 1Xnm scale memory technologies, but that will be limited to very high end products probably through the end of the decade.
  • iwod - Tuesday, July 10, 2018 - link

    that is you assuming demand being the same
  • MrSpadge - Tuesday, July 10, 2018 - link

    20% for us, 10% for their margins and future R&D - that's a win-win situation. If they passed the entire cost benefit to us, we'd probably still have to use 1 MBit devices because noone could afford further development.
  • B3an - Tuesday, July 10, 2018 - link

    Funny how none of you are thinking about price fixing, just like they undeniably do with RAM. So naive.
  • Ej24 - Tuesday, July 10, 2018 - link

    Nand prices haven't really gone up, they've stayed pretty flat for the last year or so, at least judging by consumer ssd prices. Whereas RAM has more than doubled for the same exact product in the same time period. There are also more producers of NAND than RAM with more viable competing designs, 3d, planar, mlc, tlc, and so on. Samsung can't pass on 100% of the savings to you because the initial R&D cost them money upfront and they need to make a profit to keep innovating.
  • iwod - Tuesday, July 10, 2018 - link

    You should understand more about market demand and supply before telling others about price fixing and calling other naive.
  • Altimmons - Wednesday, July 11, 2018 - link

    Then buy something else. It’s a free market after all. Unlike ram though I think the prices for storage are reasonable, affordable, and competitive. $200 for 500 Gb pro model. Not bad.
  • gfkBill - Tuesday, July 10, 2018 - link

    "20% for us, 10% for their margins and future R&D"

    Something like that, but don't forget their existing pricing has R&D and profit built in already, so no reason it can't be almost a 30% saving if it costs them 30% less to make it. Your 20% is probably pretty close to reality though.

    Ultimately competition will dictate how much of a saving we actually see, more than anything. They'll cream it while they can, as first mover.
  • sleepeeg3 - Tuesday, July 10, 2018 - link

    You are cursing Samsung for creating a better process? You have issues.

    Samsung competes with other companies so they can't arbitrarily charge more. Capacities will increase and price may come down. Everyone wins. That's the invisible hand of glorious capitalism.
  • Mikewind Dale - Thursday, July 12, 2018 - link

    We have no right to expect that the cost savings will be transmitted direct to the consumer.

    First, the price that is charged is going to depend on what Samsung's competitors can charge. Samsung's incentive to offer a net sale (price plus performance) that is just slightly better than what their next best competitor can offer. If the price Samsung charges is too high, then blame their competitors. It makes no sense to blame an innovator who is offering something new. You should blame all their competitors who can't keep up.

    Second, if savings in per-unit manufacturing costs were entirely transmitted to the consumer, then what would pay for the R&D? No company would ever engage in R&D unless it thought that it would keep some of the profits for itself. If every company transmitted all cost savings to the consumer, then there would never be any cost savings to begin with.
  • Atwork - Sunday, July 15, 2018 - link

    If the new equipment they have to buy has a life-cycle cost that's 40% higher (not unexpected) then 30% higher "manufacturing productivity" (yield?) would make it a wash. Those FAB tools don't just cost millions to hundreds of millions each, most need to be refurbished or replaced after every x number of wafers are processed. Consumable costs are likely higher as well. Each layer requires a pass through each of several expensive tools. Each pass through each tool is a chance to trash a wafer, or layer - it's amazing they get 90 to 96 successful layers out of any wafer!
  • bug77 - Tuesday, July 10, 2018 - link

    Waiting to see how many p/e cycles these have left.
  • Death666Angel - Tuesday, July 10, 2018 - link

    "without incurring the endurance and reliability costs that came with process shrinks for planar NAND flash memory"
  • bug77 - Tuesday, July 10, 2018 - link

    Yes, additional stacks won't affect durability, but we don't know whether the other manufacturing parameters will remain unchanged.
  • Mobile-Dom - Tuesday, July 10, 2018 - link

    1.8v to 1.2v? damn
  • CheapSushi - Tuesday, July 10, 2018 - link

    Stoked about 96 layer QLC NAND for bulk storage.
  • GTRagnarok - Tuesday, July 10, 2018 - link

    Will 3D NAND keep getting more layers for try foreseeable future? Is there a limit?
  • GTRagnarok - Tuesday, July 10, 2018 - link

    *the
  • haukionkannel - Wednesday, July 11, 2018 - link

    A kind of limit... the higher stack, the more wiring is needed to connect those stacks. Also the heat may become more problematic in lower part of the stack. But the wiring is the Main problem. Aka you need to have electronic router from each stack to each other stack and that Number increases when you make higher stacks.
  • FullmetalTitan - Thursday, July 12, 2018 - link

    There is a limit based on the current string stacking and cell design. Several companies expect to be able to hit 128 layers on their current technologies, but several are investigating new cell designs and layer topologies to move beyond 128L NAND.
  • tommo1982 - Friday, July 13, 2018 - link

    Is TLC the only one affected with reliability issues when shrinking the process or does it affect SLC and MLC as well?

Log in

Don't have an account? Sign up now