Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×
Media Hardware Technology

Samsung Announces Flash-Based Disk Drive 378

doc6502 writes "Samsung has announced flash-based disk drives with a 16 GB capacity, with an aim to get the drives to market by the end of the year. The (short) article suggests that this could be a big boost to laptop owners, as battery life could be seriously extended if there isn't a big high-speed motor to power constantly. The drives should be fast, too."
This discussion has been archived. No new comments can be posted.

Samsung Announces Flash-Based Disk Drive

Comments Filter:
  • Old News (Score:5, Informative)

    by TripMaster Monkey ( 862126 ) * on Tuesday May 24, 2005 @03:52PM (#12626684)

    Memtech [memtech.com] has been doing this sort of thing for a while now.

    Still, this is great news...the more companies that switch to flash technology, the more the technology itself will become mainstream. It's about time we did away with platter-based HDDs.
    • No SATA? (Score:5, Insightful)

      by Eunuch ( 844280 ) * on Tuesday May 24, 2005 @03:57PM (#12626757)
      If your're gonna jettison old crap, do away with PATA as well.
      • Re:No SATA? (Score:3, Interesting)

        by Anonymous Coward
        [shamelessplug]

        This is not news. My employer, Adtron, has been doing flash based "disk" and "tape" drives for years. And we were the first (and only?) with SATA: http://www.adtron.com/products/A25fb-SerialATAFlas hDisk.html [adtron.com]

        [/shamelessplug]

        Samsung announces and everyone goes gaga. Little guys do it for years and no one seems to know about it. [shrug]
    • Wouldn't this make the computer slower though ?, is this different to the normal type of flash that can't handle more than a certain amount of writes making it unusable for a page file.
    • Re:Old News (Score:5, Insightful)

      by rainman_bc ( 735332 ) on Tuesday May 24, 2005 @04:15PM (#12626988)
      Excuse me, but doesn't flash storage have a limited number of writes?
      • Re:Old News (Score:5, Insightful)

        by LordStraun ( 794808 ) on Tuesday May 24, 2005 @04:40PM (#12627251)
        Yes, and depending on how the writes are being spread across the media, the device could last a day or years. From the comments in TA, someone posted the following specs:

        MTBF specs vary based on the manufacturer and the calculation used; the following are some sample specs I have found:

        Pretec --> MTBF 500,000 hours (powered on)
        Simpletech --> MTBF 1M+ hours
        SanDisk --> MTBF 3M+ hours
        BitMicro --> MTBF 2M+ hours


        But the most reliable and experienced guys around are BitMicro, and this is what they pubish for one of their flash drives:

        http://www.bitmicro.com/...urces_flash_ssd_db2.php [bitmicro.com]

        Example #2: Write Frequency in MB/sec

        E-Disk® PB Size = 16 KB

        I/O Block Size = 64 KB

        Write Frequency = 6,016,204,800 KB per day (68 MB/sec)

        E-Disk® capacity = 155,648 MB
        Number of Flash chips = 608
        Size of Flash chips = 2048 Mbit or 256 MB or 262,144 KB


        Number of writes to Flash chip = 64 KB / 16 KB = 4
        Total E-Disk® physical blocks = (262,114 / 16) x 608 = 9,961,472
        Total max writes to E-Disk® drive = 9,961,472 x 1,000,000 = 9,961,472,000,000

        Endurance (in days) = 9,961,472,000,000 / (4 X (6,016,204,800/64)) = 26,492 days
        Endurance (in years) = 199,229 days / 365 = 72.59 years

        Samsung could also use a form of wear leveling [wikipedia.org] to increase device longevity.
    • Re:Old News (Score:5, Funny)

      by X1011 ( 819111 ) on Tuesday May 24, 2005 @04:24PM (#12627095) Homepage
      Memtech has been doing this sort of thing for a while now.

      Yea, they have the 35 inch SC3500 Sidewinder [memtech.com]. Imagine how much data you can put on a disk that big!
    • Re:Old News (Score:3, Informative)

      by LabRat007 ( 765435 )
      I've got some questions I hope others will weigh in on.

      1. Isnt flash media more suceptible to EMP? Wouldn't it be easier to damage with static electricty? What about common magnetic feilds?

      2. In the event of a hardware failure with a traditial hard drive you certainly can get get some of you data back given a clean room, a microscope and time. What can be done to recover data from a hardware failure on a flash drive?

      I'm going to hold on to my hard drives until I'm comfortable with the answers.
  • Wake me up when they're introduced.
  • Great news. (Score:3, Interesting)

    by Eunuch ( 844280 ) * on Tuesday May 24, 2005 @03:53PM (#12626701)
    This solution makes a lot more sense than those hybrid drives with both flash and platters. Keep it simple. I won't mourn the demise of the spinning discs. Speaking of KISS, going swapless when using this as your only drive makes a lot of sense too.

    I'd be quite interested in this for a desktop. Would pair nicely with a passively cooled system.
    • Re:Great news. (Score:5, Interesting)

      by TrippTDF ( 513419 ) <{moc.liamg} {ta} {dnalih}> on Tuesday May 24, 2005 @03:57PM (#12626754)
      Install your OS and Applications to the Flash Drive (in my world, that's more than enough space for the installed apps) and then store your larger files (music, movies) on the Platter-based drive. It will save a ton of power on a notebook, and i bet it speed up load times.
      • Re:Great news. (Score:4, Interesting)

        by rainman_bc ( 735332 ) on Tuesday May 24, 2005 @04:34PM (#12627186)
        Great in theory, but with a limit on the number of writes, you might be hooped.

        The registry is too important to a Windows OS. The OS is constantly writing to and reading from that damn thing.

        I thought about the same thing too. A Linux OS might be more efficient though... You still have the problem of where to put the swap. On the drive with limited read/writes and isn't spinning, or on the one that's spinning and consuming power. Either way. I'd be concerned.
  • by PopeAlien ( 164869 ) on Tuesday May 24, 2005 @03:53PM (#12626702) Homepage Journal
    I'm so glad to hear they aren't announcing flesh-based disk drives. There may still be time to stop the robots from consuming us all!
  • by Anonymous Coward on Tuesday May 24, 2005 @03:53PM (#12626704)
    The have fast seek times but the slow rotational speed makes for low throughput.
    • The transfer rates are very impressive for something with such a low totational speed though. I imagine that lots of geeks will be buying these things and modding them with traditional hard drive motors to unleash their real potential. Imagine getting one these spinning at even 5,400 RPMs. Think of the possibilities!
  • by ProfaneBaby ( 821276 ) on Tuesday May 24, 2005 @03:53PM (#12626710)
    What happens to the frequently accessed parts of the drives? The standard flash drives/cards stop working after a few thousand writes per sector ... in an MP3 player, this isn't such a big deal. In a laptop, that failure could get ugly.
    • by TripMaster Monkey ( 862126 ) * on Tuesday May 24, 2005 @03:59PM (#12626790)

      Here's a great paper [bitmicro.com] about flash technology in HDD applications. The document is a bit lengthy, but the conclusion is that today's flash technology allows for enough erase/write cycles to make them more than viable for HDD use.

    • by crow ( 16139 ) on Tuesday May 24, 2005 @04:00PM (#12626809) Homepage Journal
      Yes, the life of the flash is a factor here, but you're missing a couple of points.

      First, the life of modern parts if much higher than you stated. I think it's in the hundreds of thousands if not millions of writes.

      Second, they can apply the same techniques as spinning drives to remap bad blocks so that when a block stops working, it gets replaced by a spare one that was never seen by the user. A similar remapping can be done to swap heavily-used and lightly-used blocks to even out the wear and extend the life.
      • I've always wondered, how do they store the map of the remapped blocks? Sounds to me that the flash area where the map is stored would wear out quite quickly instead. Anyone have some insights?

        /greger

      • First, the life of modern parts if much higher than you stated. I think it's in the hundreds of thousands if not millions of writes.

        What about swap? Seems that would still take it down relatively quickly.

      • by frazzlenz ( 882504 ) on Tuesday May 24, 2005 @05:00PM (#12627483)
        I've done some work with Flash in an embedded database application.

        Flash is specced for 100,000 erase cycles -- in a 'disk' application this probably equates to 100,000 writes. However, after about 10,000 erases, the write speed decreases significantly.

        In my application, I remapped blocks of data on a cyclical basis, so that all the blocks would get used the same number of times.

        At 100,000 cycles, if you erased and rewrote the entire disk every hour, it would last for 11 years. How many people are still using an 11 year old HDD? (That'd be, what, 1GB or so?)

        The key question is how much this will cost. The fact that its aimed at laptops suggests that it will be significantly more expensive than a HDD.

        Another question: how long do we keep calling Flash memory devices 'Flash drives'? Or will the name hang on, like 'dialling' telephone numbers?
  • 16GB? (Score:5, Funny)

    by Anonymous Coward on Tuesday May 24, 2005 @03:54PM (#12626713)
    16GB? How much is that in Libraries of Congress? Dammnit I can't understand these fancy units like these GBs!
  • by Anonymous Coward on Tuesday May 24, 2005 @03:55PM (#12626730)
    1) Not enough space to store my pr0n
    2) Not enough space to store my bittorrent downloads
    3) Not enough space to store my iPod MP3 collection
    4) Not enough space to store the web browser cache of various goatse.cx websites
    5) Not enough space for my MythTV
    6) Not enough space to store my archive of slashdot.org

    Nothing to see. Move along.

  • by keraneuology ( 760918 ) on Tuesday May 24, 2005 @03:55PM (#12626737) Journal
    Any word on the MTBF of these things? And would they ever need to be defragmented?
    • Any word on the MTBF of these things? And would they ever need to be defragmented?

      Don't know about MTBF, but as they're not mechanical I'm sure they can live much longer than spinning disks (except for the write issue, but that can be buffered with more spares). As for defragging - don't think so, as defragging is only useful to reduce seek times while accessing the same file (the same file isn't physically scattered on disk). As there are no seek times here, why bother defragging.. file systems could b

    • Fragmentation (Score:4, Insightful)

      by ari_j ( 90255 ) on Tuesday May 24, 2005 @04:04PM (#12626856)
      The MTBF question has been asked a dozen other times, and I don't see any answers or know of any to contribute. But as to fragmentation, I would think it will not be an issue. Since there are no moving parts, there should be no waiting time to get from sector 0 to sector 8 billion. Of course, I may be wrong, particularly if there is complex circuitry to route requests to the drive, seeing as there are probably quite a few individual flash memory chips involved in this and addressing that many different chips could require a memory processor (replacing the drive controller circuitry that traditional hard drives have) which would take some time to access a given piece of the drive.
    • Well flash media doesn't have seeking like harddrives. so the disadvantage of being fragmented is that you can't do burst i/o. Although there are file systems that actively avoid fragmentation. Or that fragmentation is less of an issue because of small file sizes (unix) and a good block cache.

      FAT stinks, but NTFS is okay when it comes to fragmentation. Ext2/3, Reiser, XFS, FFS, UFS, etc are all quite good at dealing with fragmentation. I don't know about HFS+, but I suspect it's simular to UFS, but with re
    • seek time should be low enough for fragmentation not to matter.
  • The (short) article suggests that this could be a big boost to laptop owners, as battery life could be seriously extended if there isn't a big high-speed motor to power constantly.

    ...I would've thought this is an Enzyte ad.

    Talk about subliminal marketing..."This is Bob. Bob is springing large and laughing easy"... I've gotto mute the TV during those back to back Enzyte commercials during Southpark.

  • mp3 players (Score:2, Interesting)

    by DustyShadow ( 691635 )
    This should be nice for HD based mp3 players since I'm sure most of the battery life goes to spinning the platter.
  • by nweaver ( 113078 ) on Tuesday May 24, 2005 @03:58PM (#12626782) Homepage
    The hard drive is 1/3rd of a notebook's power budget, so thanks to Amdahl's law, this can increase your runtime by no more than ~50%. And probably a bit less.

    The BIG use is for ruggidized laptops: You can, combined with a passively-cooled CPU, make a laptop with no moving parts and which could stand being dropped, kicked, and shaken to a great degree without damage.
  • Looks like... (Score:3, Interesting)

    by grumpyman ( 849537 ) on Tuesday May 24, 2005 @03:59PM (#12626792)
    ... 8 sticks of 2GB USB FLASH stick with an USB hub?
  • by Scrameustache ( 459504 ) on Tuesday May 24, 2005 @03:59PM (#12626795) Homepage Journal
    Is with an animated sing and dance number.

    That's where the bar has been raised, and I won't stand for sub standard hard drive technology announcements! ;-)
  • CompactFlash (Score:5, Interesting)

    by slashdot.org ( 321932 ) on Tuesday May 24, 2005 @04:02PM (#12626837) Homepage Journal
    You can, of course, do this today by getting a CompactFlash and a CompactFlash to IDE adapter. You can get at least 8GB.

    I ran WinXP off of this for a while. It was interesting to note the different behaviour in terms of performance; sustained transfers are considerably slower, seeks are considerably faster. Over all CF is slower than a 5400 RPM notebook drive, but the overal feel seems smoother somehow.

    The unfortunate thing with CF is that they don't support UltraDMA modes, so you end up with more overhead on the CPU side, as well as a slower datapath.

    Sometimes people bring up the limited write cycles of Flash. Well, yes, I did turn off the swap file. But most modern CompactFlash perform a sort of 'load balancing' of writes, which means that if you write to the same sector twice, the write may physically happen to two different sectors.
    • Re:CompactFlash (Score:4, Informative)

      by NerveGas ( 168686 ) on Tuesday May 24, 2005 @04:08PM (#12626905)

      The CF+ and Compact Flash specification 3.0 includes UDMA 33 and UDMA 66 support. I've seen references to certain cards and CF->IDE adapters that support DMA, so that problem is partially solved, and will get better.

      As for the problem of sustained speeds, there's always RAID 0...

      steve
    • Re:CompactFlash (Score:3, Informative)

      by kawika ( 87069 )
      If the OS still hits the drive then you're not going to get the most out of hybrid drives. There was an interesting presentation at Microsoft WinHEC last month. The presenter said that Samsung's new flash was significantly faster so it eliminated a lot of the flash performance penalty. You can see the slides here:

      http://www.microsoft.com/whdc/winhec/Pres05.mspx# t oc10 [microsoft.com]
      "Hybrid Hard Drives with Non-Volatile Flash and Longhorn [WinHEC 2005; 207 KB]"

      The presentation slants towards Longhorn but you can see
  • what's new here ? (Score:4, Informative)

    by Anonymous Coward on Tuesday May 24, 2005 @04:03PM (#12626848)
    check m-systems http://www.m-sys.com/ [m-sys.com] they have a 176G flash scsi disk there, also a 'low cost' 8G ide flash drive in 1.8 and 2.5" so how is this news exactly ?
  • by rnturn ( 11092 ) on Tuesday May 24, 2005 @04:06PM (#12626873)

    I want a couple of handfuls of these to use in my home system. These aren't all that big so making a one or more RAIDsets would be nice, especially come backup time. Added plus: No spinning drives or the auxiliary fans to keep them cooled == nice quiet system.

  • Naming? (Score:5, Insightful)

    by Winterblink ( 575267 ) on Tuesday May 24, 2005 @04:07PM (#12626887) Homepage
    Why is it called a disc drive if it's based on flash memory? :)
    • Re:Naming? (Score:3, Insightful)

      by yagu ( 721525 )
      When you make a phone call, why is it called "dialing"?
    • Re:Naming? (Score:2, Insightful)

      by DrSkwid ( 118965 )
      probably by the same people that gave us the cable modem
    • Why is it called a disc drive if it's based on flash memory? :)

      Because the chips are cut from a round flat disk of doped silicon, NowSitDownKidAndPleaseShutUp. Next question....

    • Because people are already familiar with associating the hardware with an application. This is the same reason why we call them "floppy disks" when they have long since lost their floppy-ness (for those who never saw 5.25" floppies, must less 12" floppies..)
      • Because people are already familiar with associating the hardware with an application. This is the same reason why we call them "floppy disks" when they have long since lost their floppy-ness (for those who never saw 5.25" floppies, must less 12" floppies..)

        12"? I remember 8" floppies quite well. They were, indeed, quite floppy. They were named as such to differentiate them from rigid discs, a term that is hardly ever used anymore, at least not in its original context.

        As long as people don't use

  • What about servers? (Score:3, Interesting)

    by __aaitqo8496 ( 231556 ) on Tuesday May 24, 2005 @04:10PM (#12626922) Journal
    what kind of implication might this have for servers? a big performance boost to servers is caching data in ram (to reduce read access time from the hard drive). what if that read access time was minimal? would this have an impact on the need to stock servers with LOTS of ram?
  • by JustNiz ( 692889 ) on Tuesday May 24, 2005 @04:18PM (#12627027)
    Flash read/write performance is terrrible compared to DRAM, and has a very limited number of possible rewrites, too. Depending on the flash technology:(NAND=100,000 NOR=10,000).

    Other than for laptop use, I'd rather have a DRAM-based drive that optionally gets backed-up/restored to conventional HD at power-off/on. It would give much better performance than flash, last much longer and probably cost much less per Gb.

    If you just used it for /temp and the swap partition, you'd get good performance gains and it wouldn't even need to be backed-up/restored. It would save wear on your conventional HD's too.

    Unfortunately the only such drives I've found are ludicrously expensive.
    • How would a "DRAM disk" solution possibly get cheaper than normal RAM? Of course, you may "afford" a lower frequency, but I don't think you really gain that much below a certain level, as the yield can't get any better than 100 %. Of course, it's impossible to connect 16 GB of DRAM to most mainstream memory controllers today, and a specific controller + memory pack could be cheaper than a server board allowing that, but I don't think that an "all-DRAM" solution is very feasible; especially not swap on it?

      W

  • Flash memory has limited write cycle life (10,000 or 100,000 depending on technology). Excessive writes to any one location will render it useless.

    Cache and virtual memory will eat them up in no time.

Get hold of portable property. -- Charles Dickens, "Great Expectations"

Working...