Intel, Micron start sampling 20-nm NAND flash

Slowly but surely, flash memory continues to get cheaper and more capacious. Intel and Micron have just taken another step in that direction with the announcement that an 8GB NAND flash chip based on their new 20-nm manufacturing process is now sampling. (The chip is the work of the two companies’ joint venture, IM Flash Technologies.)

According to Intel and Micron, the 8GB 20-nm chip measures a scant 118 mm² and enables a "30 to 40 percent reduction in board space" compared to 25-nm chips of the same capacity. The smaller die area allows IMFT to churn out "50 percent more gigabyte capacity" using the same fabrication plants. Best of all, the 20-nm NAND chip purportedly delivers performance and endurance on par with its 25-nm predecessor—a noteworthy achievement considering finer processes usually reduce flash longevity.

Left to right: Two 4GB 34-nm dies, one 8GB 25-nm die, and the new 8GB 20-nm die.

Intel and Micron expect to start mass producing the new 8GB chip in the second half of the year. The two also reckon they’ll be sampling a 16GB device based on the same 20-nm process in that same time frame.

You might recall that Intel premiered IMFT’s 25-nm flash technology inside its 320 Series solid-state drives late last month. The drives are held back somewhat by their older controller, and they’re not that cheap—Intel officially prices the 120GB 320 Series at $209, down from around $230 for the matching X25-M, but Newegg currently charges $239.99 for the newer drive. Still, SSD prices have been decreasing steadily over recent months and years, and moves to finer manufacturing processes do translate into cheaper drives—eventually.

Comments closed
    • Spotpuff
    • 9 years ago

    Are the 20nm chips any more reliable? I recall reading the 25nm chips are somehow less reliable.

      • PrincipalSkinner
      • 9 years ago

      Yes. They are much more durable than anything before.
      [quote<] the 20-nm NAND chip purportedly delivers performance and endurance on par with its 25-nm [/quote<]

        • NeelyCam
        • 9 years ago

        They can even make breakfast in only 20mins, compared to 25min the last gen took.

        • Meadows
        • 9 years ago

        [url<]http://bit.ly/eLMu5M[/url<]

      • DancinJack
      • 9 years ago

      [quote<]Best of all, the 20-nm NAND chip purportedly delivers performance and endurance on par with its 25-nm predecessor—a noteworthy achievement considering finer processes usually reduce flash longevity.[/quote<]

    • Sanctusx2
    • 9 years ago

    Would this mean that we’ll start seeing drives using the 20nm chips in the second half? Or does it take another 6 months for design, testing, production, validation, etc after production on the chips is already running?

      • DancinJack
      • 9 years ago

      Well, if we use the 25nm introduction as a timeline, it shouldn’t be long.

      I think it took about a year from 25nm sampling to production drives.

      • PrincipalSkinner
      • 9 years ago

      No. If the chips are out second half of the year it will take some time for them to reach consumers. Aside from validation there are yields to be considered. They could be low early on.

    • dpaus
    • 9 years ago

    Is the process technoloy for CPUs and memory different? i.e., does this mean we’ll soon see 20nm CPUs?

      • DancinJack
      • 9 years ago

      They’re different. The next node for CPUs is 22nm (Ivy Bridge). Then 16nm and then 11nm. At least I think that’s the plan right now.

        • dpaus
        • 9 years ago

        Thanks for that! And now, the inevitable follow-up question: Why? (my Dad is [i<]so[/i<] glad I don't ask [i<]him[/i<] that anymore!!)

          • sweatshopking
          • 9 years ago

          it’s a different fab process. they’re not made the same way.

            • Mourmain
            • 9 years ago

            “Why are they not the same?”
            “Because they’re different”
            “…”

          • indeego
          • 9 years ago

          I crave the why question from my kids. I frequently don’t know the answer so we can figure it out together. I never want my kids to stop asking.

            • srilumpa
            • 9 years ago

            Let me applaud you.

            I think the average human intelligence would be sharply raised if parents took the time to answer their kid’s questions, and teaching them how to find the answers themselves as they grow up, rather than trying to get them to ‘stop being annoying’ (though I know how hard it can be to find the time to do so).

            So this puts you in my great parent book.

          • David_Morgan
          • 9 years ago

          Flash memory is, in general, much denser and simpler from a circuitry perspective than a CPU, and can be made with smaller pathways quicker than more complex chips. There is also the fact that the machine and chemicals used to make them are similar, but slightly different than those used in processor manufacturing.

          One nasty side effect with flash memory however is that the smaller you make it, the less reliable it becomes. (Warning… following numbers are made-up as an example): For instance, at 25nm, you might get ~1000 write cycles before the material that holds the electrical charge representing a binary “1” can no longer conform to specification and is marked “dead” by the controller. at 20nm you have less of that material to store a charge, and therefore it is easier to drop outside of the tolerated voltage range… something like 500-600 write cycles.

            • SomeOtherGeek
            • 9 years ago

            But then Cyril did say in his post that the longevity is the same…

            [quote<]Best of all, the 20-nm NAND chip purportedly delivers performance and endurance on par with its 25-nm predecessor—a noteworthy achievement considering finer processes usually reduce flash longevity.[/quote<]

            • Waco
            • 9 years ago

            Whenever I read “on par” in a press document it usually means the engineering team told them the problem/decay/whatever was the same order of magnitude as the last generation. I’m sure they said the same thing with 25 nm versus 34 nm.

            • pragma
            • 9 years ago

            Quick google for “nand 25nm write cycles” gives several pages saying, and I quote
            [quote<]34nm MLC NAND is good for 5000 write cycles, while 25nm MLC NAND lasts for only 3000 write cycles.[/quote<] I'm sure Intel site has authoritative pdfs available, just run a google search for "25nm nand site:intel.com filetype:pdf" or similar.

            • Game_boy
            • 9 years ago

            Intel must have made some improvement to the process other than just shrinking it to maintain longevity. David is saying that, without changes, shrinking flash memory reduces lifespan.

            • SomeOtherGeek
            • 9 years ago

            Oops, re-read his reply and he did say that. My bad, David. BTW, I do agree with you 3 said.

            • thesmileman
            • 9 years ago

            Did you read the article?

            • DancinJack
            • 9 years ago

            What are you referring to? Which part of the article?

            • dpaus
            • 9 years ago

            Did you read the RP’s name?

          • NeelyCam
          • 9 years ago

          The processes are optimized for different purposes. In a flash process, focus is more on cost (through higher density, fewer metal layers, simpler/fewer process steps…).

          For a CPU process, requirements are vastly different. Speed/power consumption trade-off becomes more important (affecting the transistor optimization), and the complexity of the logic and wiring between the various logic blocks require several metal layers. Various I/O blocks require analog capabilities (resistors, capacitors..) All these together require a more complicated, more expensive process.

          Moreover, the transistor structure in a flash process needs to be able to store information indefinitely (i.e., keep the information after power is turned off), while a CPU transistor doesn’t have to do that, so the transistors themselves are very different. You could plausibly implement a flash-like transistor in a CPU process, but it would come with an extra cost and/or would likely be affect the ‘regular’ transistor performance significantly.

      • xeridea
      • 9 years ago

      I don’t claim to know specifics on anything (I would wiki/google it). CPU transistors are made for super high speed switching, and no data retention. Flash memory is made for data retention, so they are going to be fundamentally different. They both use transistors, but are for different purposes, so the manufacturing process is different.

Pin It on Pinterest

Share This