Notebook SSDs disappoint

Notebook SSDs disappoint

Summary: An unsourced Engadget report says. . . flash-based laptops are being returned at a rate of 10 to 20 percent for technical failure, compared to the 1 to 2 percent of regular laptop returns due to HDD issues.


An unsourced Engadget report says

. . . flash-based laptops are being returned at a rate of 10 to 20 percent for technical failure, compared to the 1 to 2 percent of regular laptop returns due to HDD issues. . . . Other consumers are sending 'em back just due to lackluster performance -- particular applications like streaming video and, strangely, Microsoft Outlook . . . .

Low-end suckage? The high rate of failure is surprising, since millions of flash chips get wave-soldered on PCBs every day. Likely problem: the flash translation layer chip isn't fully compatible with notoriously finicky disk drivers.

The lackluster performance problem is well known to regular Storage Bits readers - search on Solid State Disk if you aren't. I'm still working on unraveling the issues in detail, but the basic problem is that flash was never intended for frequent small random writes.

Can the low-end flash drive be saved? High-end flash drives - the kind that cost $10,000 a copy - throw money at the problem. DRAM buffers, fancy pre-erase algorithms, multiple data paths and smaller block sizes all play a role.

But for your cheapie $1,000 64 GB flash drive, fuhgeddaboutit. Those drool-worthy performance specs are all large-block transfers on a virgin drive - not a crowded, fragmented drive. Vendors aren't lying. You'll just never see those numbers after a couple of weeks.

Your PC OS throws small random writes around like confetti. Open a file and get 2 writes: update the file metadata and the journal. Since the flash controller can only handle so much I/O even your nominally fast random reads get slow-w-w.

The Storage Bits take High-end flash drives seem likely to improve mid-range and high-end storage arrays, where money is not much of an issue.

But consumers face a rocky road. Even at 10x-20x the GB cost, consumer flash drives just aren't delivering the goods. Without high volumes they probably never will. Unless a deep-pocketed company steps up with aggressive forward pricing, those volumes won't materialize.

Net-net: don't hold your breathe waiting for high-performance, affordable notebook flash drives. They may never arrive.

Comments welcome, as always. Looks like the hard drive industry is gutting yet another potential competitor.

Topics: Laptops, CXO, Hardware, Mobility, Storage, IT Employment

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Eee PC - OS makes a difference

    One of the problems when using the Eee PC with XP is the logging, writes to the registry, page filing, etc. that XP normally does on a conventional hard drive. Of course you can reduce some of these writes like turning off the page file if you have enough ram or some other registry hacks, but beyond this there is not much else you can do. Without some specific changes to the OS you're going to burn out the drives in these units faster.

    Now with the custom Xandros install there is no swap file, logging is turned off, no registry to content with, and a host of other modifications have been made to reduce the number of writes in an attempt to maximize the life of the drives.

    So in order to use current SSDs in notebooks there has to be a modified OS specific for this type of storage until the technology for less degradation from multiple writes advances.
    • Score 1 for Linux

      Thanks for making me aware of the modifications made to
      the Xandros distro. They make sense and point out yet
      another advantage of an open source OS.

      The one part of today's flash controllers that does seem to
      work as advertised - probably because it was the first
      major issue vendors needed to address - is the wear
      leveling. So while the performance may be terrible, the
      extra writes probably won't be an issue for the useful life
      of the Eee.

      R Harris
      • One correction

        The custom xandros install on the Eee PC does some logging, but limits the number of archive log files plus it logs fewer services and applications.

        I did some additional checking and found out that most of the temp files are written to the tmpfs which is a pseudo drive in memory:

        tmpfs on /tmp type tmpfs (rw)

        So minimum logging, but temp files are written to memory. :)
      • One more thing ...

        The Eee PC also uses the ext2 drive format because it lacks journaling to cut down further writes to the drive. You would have to use FAT32 with XP to accomplish similar results.
  • Seems like

    Other than yourself, this is a problem nobody saw coming.
    If these special considerations have to be taken into account, why not just use a hard drive and forgo SSD entirely?

    One shouldn't need to handle things differently with SSD and it should be as transparent to the user as HD technology.

    This is not good. Thoughts?
    D T Schmitz
    • I didn't see it coming, until

      A Caltech PhD who follows these things professionally
      turned me on to the issues. Then I started researching it

      I'd had a CF card based notebook in the 90's that I loved,
      so I was thrilled to see to technology coming back into

      Once I'd done the research I saw where the problems might
      be. The performance data has only confirmed them.

      The failure rate is a surprise though. Too bad for the
      vendors and the customers.

      R Harris
      • SSD Lifetime...

        It always struck me as odd that the old flash cards had a (very generalized average here, don't kill me) *rough* lifetime of ~100000 writes... I'd assumed that this had been addressed for the new SSD's, but it would appear not... Am I missing something or did people forget ?
    • You ask the very difficult

      You stated "[i]One shouldn't need to handle things differently with SSD and it should be as transparent to the user as HD technology.[/i]" That is very hard to do. It's like asking that a 1 ton truck handle the same as a compact car. The underlying physics and reality make it very hard.

      Besides, we now have decades of experience with rotating platters in hard disk drives. That means we have decades of experience writing and tuning drivers, operating systems and applications to exploit hard drives' strengths and hide their weaknesses. It's very hard for a new physical foundation layer to step into that without some disturbance of the user experience. Very hard indeed, unless you want to spend billions duplicating the historical knowledge of hard drives that we now take for granted. Then the tech would be far too expensive to deploy at all.
  • I'd'uv never thunk this!

    Thanks Robin. As a casual reader I would not have known
    about the downside to this technology without your
    heads up.
  • RE: Notebook SSDs disappoint

    I was wondering what people (i.e. Apple) had done to rectify the finite read / write cycle situation. Looks like they haven't. So what may happen if / when these drives fail? Are there data recovery methods available?
  • Not what I read

    Darn it, can't remember where I read about this just a few days ago, but that article went into SSDs in considerable depth. Bottom line: some power savings, very fast reads, relatively slow writes, for a few specialized apps in industry they were a bad idea, for all other apps they were wonderful and fast, operating life pushed back considerably in the past few years so that now you're looking at 10+ years of heavy use under typical conditions before the drives start losing capacity.

    Whew, a pretty long sentence, but I think I summarize that article accurately. Now if I can just remember where I saw it...
    • Found it

      This is the article I saw the other day. It speaks of a spec called "write endurance" and calculates that an SSD device would offer a service life of many decades before failure. Interestingly, the article also mentions that the MTBF specs offered by hard drive manufacturers are much more theoretical than real-life.
      • oops

        Guess it would help if I included the link to said article:
        • Too bad they didn't understand Windows ...

          and it's appetite for continually writing to the storage device. From my new laptop behavior, Vista is even worse than XP was in making jillions of writes to the disk even when you are browsing or using email.

          Combine that with the bloat of Windows, taking a major portion of your expensive SSD storage for the base OS + patches, and you have a technical and economic disaster.
          terry flores