SSDs with next-gen NAND approach

At the Computex trade show in Taipei, Taiwan earlier this year, the folks at SandForce told us to expect drives based on next-generation NAND in the fall. Now, a report out of China suggests the first offering could come from Intel. VR-Zone’s Chinese site has posted an official-looking roadmap that has an Intel 335 Series SSD scheduled for the third quarter of this year. Only a 240GB model will be released at that time, with 80GB and 180GB variants supposedly following in the first quarter of 2013. The 335 Series will reportedly use the same SandForce SF-2281 controller as the Intel 330 Series, except with newer 20-nm NAND.

According to the same roadmap, the Intel 525 Series will debut in the fourth quarter. This new addition looks like an mSATA version of the 520 Series, which pairs the SandForce controller with 25-nm NAND. The 525 Series won’t make the jump to 20-nm NAND, VR-Zone says.

Of course, Intel isn’t the only firm with next-gen SSDs in the works. EXPreview says the Plextor M5P will debut next month with 19-nm Toggle DDR NAND. The SSD will purportedly tap Marvell’s 88SS9187 controller and be available in 128-512GB capacities. We can expect sequential throughput up to 540MB/s for reads and 450MB/s for writes, EXPreview says. The M5P’s random I/O ratings max out at 94,000 read IOps and 86,000 writes.

Although Google’s translation of EXPreview’s Chinese report is a little hard to understand, it looks like the M5P will be priced in the same range as the existing M3P. Newegg is selling that model at $180 for 128GB, $280 for 256GB, and $680 for 512GB. Those prices are a little on the expensive side considering the Blue Light Specials we’ve seen on some other SSDs lately.

Comments closed
    • flip-mode
    • 7 years ago

    I’m a fan of the Marvell controllers. I’d pick a Marvell controlled drive over a Sandforce controlled drive, all else equal or even somewhat unequal.

    • FranzVonPapen
    • 7 years ago

    Sweet, I can’t wait for this upcoming NAND with 1000-write-cycle endurance!

    Flash RAM is such a dead end. I refuse to support it with my dollars.

      • UberGerbil
      • 7 years ago

      You know conventional hard drives are a dead end too, right? What are you using for storage?

      LCDs are a dead end. So is DDR3. [i<]Every[/i<] technology is dead end if your perspective is long enough. The question is whether you get sufficient value out of it while it's in its prime. For lots of people, SSDs offer plenty of reasons right now for them to spend their dollars.

      • FranzVonPapen
      • 7 years ago

      At least the “dead end” or final evolution of mechanical hard drives is, you know, USEFUL and USABLE. Even if they can’t make the magnetic domains any smaller, the drives continue to have a long service life.

      But, I’ll grant you that Flash SSDs fit in much better with our fetish for disposable goods.

        • mganai
        • 7 years ago

        If you don’t mind a slower-loading OS. 🙂

          • ptsant
          • 7 years ago

          How much time do you spend loading the OS in a work (or play) day? I start up, go get a coffee and after half an hour everything is cached in 16 GB RAM. Media (photo, video etc) are never on an SSD anyway…

            • mganai
            • 7 years ago

            What about apps?

        • eofpi
        • 7 years ago

        The large proportion of 1-egg ratings on large hdds suggest otherwise.

          • FranzVonPapen
          • 7 years ago

          I see more 1-egg ratings on SSDs of all stripes than I do on mechanical hard drives.

          Not to mention mechanical hard drives don’t have a narrowly-defined, near-term wear-out point. How many SSDs have 5-year warranties, whereas Western Digital’s entire Caviar Black and Velociraptor lines do.

            • mganai
            • 7 years ago

            SSDs have been getting better and better though.

            Plus, I’m not sure if mechanical hard drive reviews tend to be as finicky.

        • bcronce
        • 7 years ago

        “USEFUL and USABLE”

        Depends on your definition. If mechanical HDs take 10min to boot a new OS and SSD take 1s, could you still define mechanical as “USEFUL and USABLE”?

        It has been estimated that NAND SSDs will plateau around 16TB drives. Even with limit writes, they will still out-live their usefulness for non-enterprise users.

        Not to mention that when an SSD dies from too many writes, they fail in a read-only state and only the last transaction to the HD is incomplete. The bulk of your data is still accessible, assuming the controller has decent firmware.

      • My Johnson
      • 7 years ago

      You’re wrong in your thinking. What you leave out is consideration of an SSD as a boot drive and a HDD for bulk storage. That combination is quite a winner, IMHO.

      • TO11MTM
      • 7 years ago

      While some people seem to just say (tl;dr version here) ‘you’re wrong, SSDs rock,’ I think someone should provide numbers to back them up. So I did.

      Intel is saying their 20nm Flash should have the same P/E cycle count as their existing 25nm Flash.
      That’s 3000-5000 Cycles. (Source: [url<]http://www.anandtech.com/show/5193/intel-and-micron-imft-announce-worlds-first-128gb-20nm-mlc-nand[/url<] . Note this is Not 'enterprise grade' either.) So, assuming you have a 128GB SSD, with a 1.25 Write amplification factor (Which is fairly pessimistic given a lot of the newer controller technology.) Also we'll assume that the controller has decent wear leveling (Although really that's part of the Write amplification factor anyway...) (128*3000)/1.25=307200GB, or 300TB of Data, would have to be written to the drive before you're done with it. And that's again, a pessimistic look. Run the numbers on the upper end and you have 500TB of write life. If it is a dead end, that sounds like a pretty long road at least. =)

    • tootercomputer
    • 7 years ago

    Geoff, exactly what is the “next-gen NAND approach”, will it improve performance, and if so, in what way?

      • continuum
      • 7 years ago

      Should result in cheaper pricing. I think that’s it at the moment, performance will need new controllers…

        • Mourmain
        • 7 years ago

        Well, it looks like it’s way more expensive right now. I don’t see the attraction… Maybe it has low power usage?

    • demani
    • 7 years ago

    I thought with the shrink to smaller feature sizes new controllers were going to be needed to combat the problems inherent to the shrink (less write endurance). Maybe that’s why the 525 series isn’t going down to 20nm?

      • Madman
      • 7 years ago

      This is what I was thinking about as well.

    • chuckula
    • 7 years ago

    Want: 1 TB SSD for (at most) $800. That price is twice the going rate for a single 512GB M4 right now, so it’s not too unreasonable to ask. Interestingly, on a per-GB basis SSD prices usually get cheaper up to 512GB and then spike through the roof for larger capacities.

      • yogibbear
      • 7 years ago

      Same thing happens with RAM. Gets cheaper up to ~4GB per stick of DDR3 then increases linearly

        • superjawes
        • 7 years ago

        Margins and markups. 2GB, 4GB, and 8GB sticks require a lot of the same things (PCB is a big one), and then most of the cost is in how dense/how much silicon you get, but it makes it easy to mark up larger capacities. “If you’re getting twice the memory, why shouldn’t you pay twice as much?” –Manufacturers

      • ImSpartacus
      • 7 years ago

      Just get the 512GB m4. What’s so special about 1TB?

        • Madman
        • 7 years ago

        Yep, and stack ’em in raid 0 or JBOD. Good thing about SSD is that they can be joined in arrays.

        They are still to expensive for my liking though…

          • Airmantharp
          • 7 years ago

          You lose Trim, unless it’s an OS-level JBOD. It’s not like having two drives is going to make your desktop experience twice as good.

    • OneArmedScissor
    • 7 years ago

    Is anyone other than Intel using high-k gates for 20 / 19nm? While that was done for write endurance, and toggle memory already hardly uses any power, I’m still curious if that will accomplish anything for battery life in phones / tablets / laptops. Phones, especially, could benefit from cutting even a tiny fraction of a watt.

      • Oberon
      • 7 years ago

      Nevermind, misunderstood your post.

        • OneArmedScissor
        • 7 years ago

        [url<]http://www.dailytech.com/IMFTs+20nm+HKMG+NAND+Flash+Memory+in+Mass+Production+Intel+Plans+New+SSDs/article23351.htm[/url<]

          • Oberon
          • 7 years ago

          Yup, for some reason I thought I had read something about FinFETS or tri-gate in your post. That’ll teach me to respond to posts before I’ve had my morning caffeine….

      • BobbinThreadbare
      • 7 years ago

      I don’t think tiny fractions of watts help all that much. ARM chips are generally using multiple watts, and screens use way more than that. Maybe standby time could be improved but I don’t think by much unless you start cutting at least 1/2 a watt.

        • OneArmedScissor
        • 7 years ago

        Phones, with the screen on, shouldn’t even be using 1 watt. The current iPad 2 generally uses 2 to 3 watts. The power use is almost fixed, regardless of the exact task, because of how many dedicated processors there are.

        Because of that, whatever leakage can be mitigated is likely going to be tangible in all cases. Even if it’s 0.01 watt in a phone, that could be something.

        • ludi
        • 7 years ago

        ARM chips, in mobile device applications, often have TDPs of [i<]less[/i<] than 1W, and use power in the low mW range under most operating conditions. My Google Nexus One, which uses a 1GHz "Snapdragon" ARM processor and has a battery rated at 3.7V and 1400mAh, or 5.18W-h. If it was burning power at anything close to your estimates, the battery would die in as little as 15-20 minutes of casual use.

Pin It on Pinterest

Share This