Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
×
Data Storage Intel Businesses Software Hardware Technology

Intel Launches Flurry of 3D NAND-Based SSDs For Consumer and Enterprise Markets (hothardware.com) 145

MojoKid writes: Intel launched a handful of new SSD products today that cover a broad spectrum of applications and employ 3D NAND technology. The SSD 600p Series is offered in four capacities ranging from 128GB, to 256GB, 512GB and 1TB. The drivers are targeted at consumer desktops and notebooks and are available in the M.2 form-factor. The entry-level 128GB model offers sequential reads and writes of up to 770 MB/sec and 450 MB/sec respectively. At higher densities, the multi-channel 1TB model offers sequential reads and writes that jump to 1,800 MB/sec and 560 MB/sec respectively. The 128GB SSD 600p weighs in at $69, while the 1TB model is priced at $359, or about .36 cents per GiB. For the data center, Intel has also introduced the DC P3520 and DC S3520 Series SSDs in 2.5-inch and PCIe half-height card form-factors. Available in 450GB to 2TB capacities, the range-topping 2TB model offers random reads/writes of 1,700 MB/sec and 1,350 MB/sec respectively. Finally, Intel launched the SSD E 6000p (PCIe M.2) and SSD E 5420s Series (SATA). The former supports Core vPro processors and is targeted at point-of-sale systems and digital signage. The latter is aimed at helping customers ease the transition from HDDs to SSDs in IoT applications.
This discussion has been archived. No new comments can be posted.

Intel Launches Flurry of 3D NAND-Based SSDs For Consumer and Enterprise Markets

Comments Filter:
  • Math is hard (Score:4, Informative)

    by Anonymous Coward on Friday August 26, 2016 @06:46AM (#52773857)

    the 1TB model is priced at $359, or about .36 cents per GiB

    Really? REALLY?

    • Re: (Score:2, Informative)

      by Anonymous Coward

      It'd be roughly $0.39, anyway. If you want to be clever and use GiB instead of GB, use it correctly.

    • by Anonymous Coward

      Hah. Off by a factor of 100 *and* fucked up the TB -> GiB conversion.

      1TB = 10^12 bytes
      1GiB = 2^30 bytes
      therefore, 1TB = approx 931.32 GiB

      $359 / 931.32 GiB = approx 38.5 cents per GiB

      Seriously though: who the **** uses cents per GiB? Drives have always been marketed in GB and TB.

      $359 / 1000 GB = exactly 35.9 cents per GB

      • Uh, GiB is used for 1024, instead of 1000. So your above number would be $359/1024, which would be 35.05 cents per GB. Almost a cent difference. Miniscule at this level, but blows up once one starts talking TiB. Incidentally, the reason GiB is used is so that people distinguish b/w the scale varying by a factor of 1000 vs 1024, which blows up when you start squaring and cubing them
      • by jbengt ( 874751 )

        Drives have always been marketed in GB and TB.

        No, they used to be marketed as MB, and even lower. And that was when a MB was typically understood to be 1024 KBs and a KB was known to be 1024 bytes.

        • That's why they came out w/ KiB, MiB, GiB, TiB et al. To distinguish b/w factors of 1000 and 1024. 1,000^3 is 1,000,000. 1,024^3 is 1,073,741,824. There's quite a bit of difference.
        • Was MB ever 1024K? (except for memory)

          I didn't ever see/use a 8 inch floppy.

          Early 5 1/4 floppy was 40 tracks, 18 sectors/track - 360K. That was 1024 bytes/K

          Early 3 1/2 floppy was 80 tracks - 720K.

          Double sided - 1.44M (we'd already started confusing multipliers)

          At some point the remaining factor of 1024 got dropped - probably when we stopped thinking about heads, tracks, sectors/track. Prior to that the first 1024 was baked into the disk geometry. I don't remember enough detail of the early hard disks to rec

          • Floppies and hard disk storage was never 1024. But with SSDs now stepping in to replace them, the story is different. SSD is made out of NAND flash, which is addressed bitwise, and so will always be a power of 2, w/ some memory set aside for things like redundancy. So stating SSD memories in terms of TB instead of TiB doesn't make much sense.
      • You think that's bad, TFS went from "a flurry" to "a handful" in a heartbeat and I want to see the conversion chart on that!
        Don't mind me, I read the RSS headline as "Intel Launches Furry of 3D NAND-Based SSDs..." and was disappointed when the story turned out to be something else entirely...
      • 1 TB = 1099511627776 bytes.

        1 TiB = fuck you, go back to CS 101.

        1 TB on the label of a storage device = fuck you, marketers, I know that's just over 0.909 TB.

        1 Mb on any networking interface = fuck you, marketers (as usual) and retarded standards bodies (you confused baud and bit ages ago and you're too ashamed to admit your fuckup).

    • yes, really, you poseur. Math *is not* hard, but it seems the basic arithmetic operation of approximation is very hard indeed for you. dividing by a 1000 produces about the same result as dividing by 1024, and rounding up is just more of the same dark magic, I assure you. By the way, what part of "about" is confusing you?
      • I believe that this:

        .36 cents

        is one of the issues. It is .36 dollars, or 36 cents. .36 cents is 1/3 of a single cent, which is a factor of 100 off.

        Also, the mismatch in a 1TB drive (1000^5) and GiB (1024^4), which isn't what hard drive prices are EVER measured in.

  • by Anonymous Coward

    "1TB model is priced at $359, or about .36 cents per GiB"

    36% of a cent per gigabyte? Care to multiply that out for me?

    • The linked article gives the same numbers, so the blame lies with the author of the article; Brandon Hill [hothardware.com].

      • Re: (Score:2, Informative)

        by Anonymous Coward

        Actually, no. The linked article doesn't include the word *cents*, hence the .36 refers back to the full price unit of $, and is therefore correct. The summary is wrong.

  • Speed or density? (Score:4, Interesting)

    by jez9999 ( 618189 ) on Friday August 26, 2016 @07:05AM (#52773887) Homepage Journal

    So is the main advantage of 3d NAND technology going to be access speed? I thought it was going to be able to enormously increase capacity, but with the drives coming in between 128GB and 1TB (similar sizes to existing drives), maybe I got the wrong idea.

    • Re: (Score:3, Interesting)

      by Anonymous Coward

      Planar (2D) NAND is getting smaller and smaller in order to accommodate increases in required density per module. This leads to bigger SSDs, but has a downside - smaller cells are more fragile, which decreases durability (we're down to thousands of erase cycles nowadays), and it's harder to measure multiple levels of voltages reliably.

      The main advantage of 3D NAND is the ability to have big cells while still having steady increases of density per module. Durability is also back to "old levels". That's why 3

      • You talk about a 4TB SSD and then link to a picture of a 2TB SSD?

        (Those modules on the board are 500GB or 512GB modules from what Google tells me)

    • by Z80a ( 971949 )

      Piling up chips and keeping em cool is not that easy.
      It's a lot easier than keeping shrinking the chips, or making the die bigger, but still not easy.
      I wonder if in the future, we will have those weird chips floating in a thermal paste, with copper posts holding em in place and also serving as the communication vias.

      • How exactly do they do Si in 3D? I used to do product marketing of multi-chip packages at one time - where 2 or more die are stacked within a package: that was package level integration. But within a die itself, how do they do multiple layers? Especially since MOSFETS, which look 2D on a diagram, are actually 3D when one comes to the gate, source and drain layers.
        • by Z80a ( 971949 )

          Well, i was thinking the "3D package" they were talking about is a big stack of dies.
          Because otherwise you can arguably call any chip 3D, not only due the gates being 3D as you mentioned, but also all the different layers of vias etc..

          • Stack of dies is what I described above - multi-chip packages. The way Intel has been describing this, I got the impression that they are talking about at silicon level. You have multi-chip packages, you have POP - package on package, where 1 chip is larger and taller than the other and mounted on it, so that they share the same real estate on the board - sorta like living in a multi-storeyed building.
          • Traditionally transistors are made by doping the surface of a silion wafer. The interconnect is then built on top by laying down alternate layers of oxide and either metal or polysilicon. So while you get multiple layers of interconnect you only get one layer of transistors.

            "3D" ICs aim to have not just multiple layers of interconnect but also multiple layers of transistors.

        • http://www.semi.org/en/node/38... [semi.org]

          One layer at a time?

      • by Gr8Apes ( 679165 )
        Apparently Seagate got it right. They announced a 60TB SSD [slashdot.org] a couple of weeks ago.
      • by AcquaCow ( 56720 )

        Eh, keeping NAND cool isn't really an issue. NAND likes heat. It's the controllers that you have to keep cool.

        Source: I just spent the last 6 years working for Fusion-io/SanDisk

      • http://www.semi.org/en/node/38... [semi.org]

        They don't stack packages, the package contains an actual 3d structure of NAND that holds the data.

    • Or cheaper. We've been hearing about SSD under 30 cents a GB "real soon now" for, what, five years now? At ten cents it replaces hard drives in all small capacities. The slope [pcpartpicker.com] still puts that many years out.

      Maybe 3DXpoint will depress the NAND prices for existing fab utilization next year. Here's hoping.

    • No you got the right idea but just the wrong approach for the same reason as Intel followed the tick-tock approach to chip design. New design followed by new process.

      3D NAND is a new process. Out of the box the first generation of the technology has already shown to be cheaper than the previous generation.
      Now that this technology is here we can look at expanding it with new designs.

  • I'm getting old. (Score:5, Insightful)

    by dr.Flake ( 601029 ) on Friday August 26, 2016 @07:13AM (#52773913)

    What amazes me most, but is probably because i'm getting old is:

    looking at the announcement, it seems like the SATA drive is just an obligatory part of the line-up. Its all M.2 and PCIe.

    Sure, SATA is getting old quickly and starts to become the bottleneck, but the way this is going, motherboards will soon have some SATA port somewhere for the occasional DVD / old spinning drive, and M.2 for the rest. Did i just recently buy my last SATA drives to fill up my NAS? I'm not planning on buying more for the next couple of years.

    Man, i remember buying my first ATA drive. And i was late to the party, it already was a stunning 20MB (imagine how many WP files were needed to fill that sucker up to the rim). And man, that thing was fast as lightning! ;-)

    • by wbr1 ( 2538558 )
      Hell I remember MFM/RLL drives. First drive I worked on a lot was the ST-225 AT. Great drive that one.
    • by Anonymous Coward

      Huh. My first thought was: SATA? In the data center? Who the fuck uses SATA in the data center? Where are the SAS versions.

      • SATA has it's uses in the data center. Think: cheap nearline storage where performance isn't the concern, but density and cost is.

      • I do. 4 SSDs in RAID 10 covered that server's performance needs easily and saved a buttload of money.
        We didn't buy consumer SSDs, we bought the datacenter ones with the longer warranty, extra over provisioning, higher performance, and better TBW. But they're still SATA instead of SAS.

      • by Dadoo ( 899435 )

        Where are the SAS versions.

        My guess is that Intel will never produce SAS drives. Unlike SATA, SAS is big-endian, and Intel has made it very clear, over the years, they don't approve of that.

    • by swb ( 14022 )

      I get the M.2 format's advantages, but I don't understand why they wouldn't offer the same drives in SATA packaging. It seems to me there's a hell of a lot more devices that accept SATA devices than M.2 devices.

      Has anyone heard of NAS or SAN devices that now feature rows of M.2 slots instead of SATA sleds? I like the idea, I just don't see anyone making them at this point.

      • by Anonymous Coward

        Has anyone heard of NAS or SAN devices that now feature rows of M.2 slots instead of SATA sleds? I like the idea, I just don't see anyone making them at this point.

        QNAP announced one earlier this year [anandtech.com], although they couldn't seem to come up with a precise use for one. "SSDs are quiet! You could use it for presentations or karaoke!" is about the best they could come up with. I'm sure some people will find reasons to need the speed of M.2 drives, but aside from that I'm not sure why SATA SSDs wouldn't suffice, it's not like there is a lot of demand for NAS boxes to be teeny.

      • I get the M.2 format's advantages, but I don't understand why they wouldn't offer the same drives in SATA packaging.

        If you need the SATA packaging to fit existing hardware you can get M.2 to SATA adapters for $8-10:
        Oley Laptop SSD NGFF M.2 to 2.5" SATA 3 PC Converter Adapter Card [amzn.com]
        AD905A SATA III 3 to M.2 (NGFF) SSD 7+5 pin Connector Converter Adapter Card [amzn.com]

        Here's a higher-end dual-M.2 to SATA adapter with integrated hardware RAID for $40:
        Ableconn ISAT-M2SR 2.5" 7mm SATA III to Dual M.2 SATA SSD Adapter with Hardward RAID [amzn.com]

        Has anyone heard of NAS or SAN devices that now feature rows of M.2 slots instead of SATA sleds?

        They don't appear to be commonplace yet, but here's one example:
        Qnap 4-Bay M.2 SSD NASbook with Bu [amzn.com]

        • by swb ( 14022 )

          Somehow the reliability of knockoff aftermarket adapters is less appealing than OEM SATA packaging.

          • I understand your concerns, but these adapters are basically just wiring and physical supports. There are hardly any electronics involved (perhaps a discrete voltage regulator, judging from the images). If you would be willing to trust a non-OEM SATA cable and mounting bracket then I wouldn't see any reason not to trust a non-OEM M.2 to SATA adapter.

            There are some higher-end models [amzn.com] which provide a full 2.5" enclosure for your M.2 drive for $20-30, if you want the extra peace of mind.

      • There are M.2 to SATA enclosures available now.

        http://www.newegg.com/Product/... [newegg.com]

        But, you lose a considerable amount of speed because SATA 3.0 isn't anywhere near as fast as a direct PCIe connection used in M.2.

      • "I don't understand why they wouldn't offer the same drives in SATA packaging"

        The "packaging" is just an PCB with an edge connector, so you can't just make 1 board. (You could use an adapter I suppose.)

    • SATA and SAS drives will still be around for some time, specifically for the data center. There are countless 4U disk shelves with 16+ SATA or SAS ports on them out there that will outlive the drives that are currently spinning in them.

    • But once you go SSD you do not go back.

      I have a raid 0 SSD with samsung pros and a regular samsung pro. Besides benchmarks there is no noticable difference unless you sping up 4 VM's at the same time :-) Even then it is only a few seconds.

      It is IOPS and not how many megs per second for the user. Speed in bandwidth is irrelevant as a PC needs lots and lots of read dependent on data from other reads in tiny small batches like reading hte registry, loading daemons/services, etc.

    • looking at the announcement, it seems like the SATA drive is just an obligatory part of the line-up. Its all M.2 and PCIe.

      For SSDs is this really so surprising? A common configuration at the moment is SSD for the primary drive, spinning rust for the secondary large storage. With many motherboards shipping with M.2 slots and all of them with PCIe, it would make perfect sense to not have to run cables to your primary part of your system, and only for the secondary high density storage.

      This is common in my systems at home. They all have M.2 drives in them. My desktop also has a 2TB spinning rust, and my NAS has one M.2 drive and

    • by Solandri ( 704621 ) on Friday August 26, 2016 @01:20PM (#52775891)
      It's important to understand that while we benchmark storage in MB/s, those units are actually the inverse of how we perceive their speed - wait time. Wait time would be sec/GB. To see what the consequences of this are, imagine loading up a game involves reading 1 GB of data, and for simplicity imagine you can read that 1 GB at max speed.

      33 MB/s = 30 sec - old IDE HDD
      66 MB/s = 15 sec - newer IDE HDD
      125 MB/s = 8 sec - SATA HDD
      250 MB/s = 4 sec - SATA2 SSD
      500 MB/s = 2 sec - SATA3 SSD
      1000 MB/s = 1 sec - early PCIe SSDs
      2000 MB/s = 0.5 sec - newer PCIe SSDs

      Notice how every time MB/s doubles, wait time is only cut in half. This means perceive speed increases are the inverse of MB/s, and thus not linear in terms of MB/s. The difference between SATA and SATA3 (125 MB/s and 500 MB/s) is "only" 375 MB/s. While the difference between SATA3 and newer PCIe drives is a whopping 1500 MB/s. But that doesn't mean that upgrading from SATA3 to a newer PCIe SSD will feel 4x faster than upgrading from a HDD to a SATA3 SSD felt.

      The reduction in wait time going from the SATA HDD to a SATA3 SSD was 8 sec vs 2 sec - a 6 sec reduction. But the reduction in wait time going from SATA3 to newer PCIe is only 2 sec vs 0.5 sec - a 1.5 sec reduction. So upgrading from a SATA3 SSD to a newer PCIe SSD will only give you 1/4 the perceived speed increase you got when you upgraded from a HDD to a SSD. Not 4x. Compared to a SATA HDD, a SATA3 SSD gives you 80% the wait time reduction of the newest PCIe SSDs (6 sec vs 7.5 sec).

      In other words, for the typical amounts of data we need to read off of storage, SATA3 SSDs have already given us most of the speed benefit we can expect by making our storage media faster. (The same problem plagues cars and using MPG to measure fuel efficiency. MPG is actually the inverse of fuel efficiency. It's the metric you want to use if you have a fixed amount of fuel and need to know how far you can travel, like if you're in a boat. The vast majority of people's driving is the other way around - they need to travel a fixed distance, and want to do it using as little fuel as possible - which is GPM. So the biggest fuel savings actually comes from making fuel hogs like tractor trailers, buses, and SUVs more efficient, not from econoboxes like the Prius. Despite how big 50 MPG sounds, going from 25 MPG to 50 MPG actually only represents half the fuel saved of going from 12.5 MPG to 25 MPG.. The rest of the world measures fuel efficiency in liters per 100 km for this reason - equivalent to GPM.)
    • by tlhIngan ( 30335 )

      Sure, SATA is getting old quickly and starts to become the bottleneck,

      No, SATA IS the bottleneck.

      If you read specs and they all say 540MB/sec, that's the SATA3 limit. And benchmarks of practically every SATA SSD has it pegged at 540MB/sec.

      Its why Apple pioneered PCIe for storage, and brought everyone a 1GB/sec SSD read and 750MB/sec write at the beginning. Nowadays a NVMe PCIe SSD can easily do 1.5GB/sec reads and 1GB/sec writes, while the top end can do 2.5GB/sec reads and 1.5GB/sec writes.

      The other reason

  • How does this compare to 3d-xpoint stuff?
    • by swb ( 14022 )

      Yeah, where IS 3D-Xpoint?

      A push into the MLC market with a miracle storage technology "just around the corner" seems an odd initiative. If 3D-Xpoint is as good as they say, I would think they would want to focus on stealing the market with a unique and superior product rather than trying for slivers of an existing market.

      Of course the cynic in me assumes that 3D-Xpoint is nowhere near ready and if it is, Intel just want to milk the existing NAND technology for maximum profit and dribble out the new stuff a

      • Fact is Intel/Micron and Toshiba were years behind Samsung on 3D-NAND technology. The 3D Xpoint press release smelled a lot like vaporware when I heard about it. Intel and the industry has been working on PCM for decades. Remember Ovonyx? Intel announced a large investment in it around the time the *Pentium 4* came out and it was old even then... The industry has been working on PCM since the 1960s-1970s.

        Intel/Micron and Toshiba are manufacturing 3D-NAND this year so there should be a price drop soon as com

        • by swb ( 14022 )

          It's funny, but I could have sworn I read Intel actually demoing the technology at a media event, that it was already production ready and that it was beating NAND in all the significant measures, density, speed and durability.

          The chatter was that it was *so* good that it was being considered as potential augmentation for RAM, allowing for huge RAM cuts in lower end devices since swapping to it would be largely indistinguishable from actual memory access on low end systems. Marginally believable as I have

    • by erice ( 13380 )

      How does this compare to 3d-xpoint stuff?

      You can actually buy the NAND SSD's. Who knows when 3D-Xpoint will actually ship?

      When it becomes available, 3DXpoint is expected to be faster than NAND Flash but also more expensive. To make use of that speed it needs a lower latency interface than PCIe. Unfortunately, it is not quite fast enough to comfortably mix with DRAM on the DDR bus. It remains to be seen how it will actually be connected.

    • by ganv ( 881057 )
      3d Xpoint is a fairly different technology. It is much faster than NAND and much cheaper than DRAM while still being non-volatile. Initially some people may use it in expensive high speed SSD configurations like Optane, but I think the real potential is in new architectures with huge non-volatile fast memory. Maybe it will replace Flash in mobile devices that currently operate without off-processor DRAM. It is possible that manufacturing becomes cheaper and it will compete with NAND Flash for non-vo
  • Pedantry fail (Score:2, Informative)

    by Anonymous Coward

    the 1TB model is priced at $359, or about .36 cents per GiB

    Look, I've got nothing against being pedantic with "GiB" = 2^30 bytes, and I can divide by 1000 on my own, but if you're going there at least get it right.

    1TB is 1,000,000,000,000 bytes and is only 931.323 GiB, so the cost per GiB is 39 cents to 2 significant digits. Note that's nowhere near .36 cents, which is less than half a cent. I presume OP meant "$0.36" or "36 cents".

    • OP subscribes to the obscure branch of mathematics called Verizon Math. By Verizon Math, OP's calculation checks out, I double checked.

  • So, how much am I going to pay for a 512GB M.2? And does it have 4 lanes?

  • ...what does this mean? Does it imply that they'll have an easier time keeping their fabs at 100% utilization, regardless of how the market for their CPUs and other stuff is doing? By flooding the market w/ so many SSDs?
    • Flash typically uses different manufacturing processes than the logic processes used for CPUs. So it's not like you can switch a fab from manufacturing CPUs to manufacturing memory like that.

  • The limitation of SATA is the bandwidth. NVMe can go to 2 gigs so you won't see much difference on consumer pcs. But it would be a nice relief as I pay for hte premium prices of the samsung pros in my rig.

    • NVMe scales. The high end devices today use 4 lanes of PCIe 3.0. There's nothing preventing someone from using (and benefiting from) 16 (or more) lanes of PCIe 4.0 the day that shit is ready to go.

He has not acquired a fortune; the fortune has acquired him. -- Bion

Working...