SSD life expectancy - what's the status?

Discussion in 'Memory and Storage' started by Gaco, Apr 28, 2009.

  1. Gaco

    Gaco Junior Member

    Joined:
    Apr 28, 2009
    Messages:
    2
    Likes Received:
    0
    I recently talked with a friend about SSD's and I mentioned having read Anand's excellent article "The SSD Anthology: Understanding SSDs and new drives from OCZ" (http://www.anandtech.com/storage/showdoc.aspx?i=3531).

    I was talking about the the declining performance problem while he was focusing on the life expectancy of SSDs, something that he considered to be the biggest problem. If it's still so much of the issue, why wasn't it mentioned in Anand's article? Or has a solution been been found, leaving it as an early issue of early-gen SSDs?

    Thanks,
    - Gaco
     
  2. Modelworks

    Modelworks Lifer

    Joined:
    Feb 22, 2007
    Messages:
    16,237
    Likes Received:
    0
    Nobody really knows how long a SSD life span will be. The only sure way to know is with time. All we have right now are lab test simulating what the expected wear will be. Like any tech there are going to be problems that nobody foresaw. Remember when recordable optical media was released they said it should be good for 100 years, then dropped it to 50, then 20 , then 5.

     
  3. Gaco

    Gaco Junior Member

    Joined:
    Apr 28, 2009
    Messages:
    2
    Likes Received:
    0
    Are there no indicators yet? Is 2 years as good a guess as 10 years? And why no mention of it in Anand's recent mega article if it's something to worry about?
     
  4. Elixer

    Elixer Diamond Member

    Joined:
    May 7, 2002
    Messages:
    7,812
    Likes Received:
    42
    As was mentioned, we just don't know yet, until these things get out to more people.

    It also depends on how you use it. The more writes it does, then the less life it has.

    Just have a good backup plan with this (or any storage device) and you will be fine.

     
  5. alcoholbob

    alcoholbob Diamond Member

    Joined:
    May 24, 2005
    Messages:
    5,273
    Likes Received:
    1
    Intel stated 5 years at 100GB/day for their 80GB X25-M. This was before they made some firmware changes that likely reduce the lifespan. However, chances are you won't use more than 30GB/day and we are being generous--in that case you are looking at least 15 years of use.

    But if you are running a supersmall SSD like those 4-8GB netbook ones, don't expect much lifespan...

    If you buy an SLC drive (10x the expected writes compared to a MLC drive) for mainstream use...the drive will outlast you.
     
  6. taltamir

    taltamir Lifer

    Joined:
    Mar 21, 2004
    Messages:
    13,574
    Likes Received:
    0
    for a home user its HUNDREDS of years. don't worry about it.
     
  7. shabby

    shabby Diamond Member

    Joined:
    Oct 9, 1999
    Messages:
    5,622
    Likes Received:
    2
    3 of my hd's already failed this year, i hope ssd's start picking up the pace in the storage area.
     
  8. taltamir

    taltamir Lifer

    Joined:
    Mar 21, 2004
    Messages:
    13,574
    Likes Received:
    0
    um, just to clarify, hundreds of years before it RUNS OUT OF WRITES and becomes READ ONLY...

    It can fail at any moment due to the internals frying, usually from a power surge... but a regular drive can fail either from the motor failing (most common failure) from a power surge ruining the internals. an SSD only from such a power surge. (which would have identical failure results on an SSD and a spindle drive)...
    Regular drives are also significantly more prone to damage from impact or vibrations...
     
  9. Dazex

    Dazex Member

    Joined:
    Oct 9, 1999
    Messages:
    38
    Likes Received:
    0
    Anand mention this info in his Intel SSD review on page 4, "How Long Will Intel's SSDs Last?" In his anthology article, he referred back to the Intel article a lot.

    But basically, the OEMs wanted to write 20GB of data per day and guaranteed that this could go on for five years. Intel went one step further and delivered 5x what the OEMs requested. So you can write 100GB everyday for the next five years and your data will stay intact. But in the end, we're still in the early days of SSD...so what other poster wrote is probably correct. In that, time will tell. I am sure in the next year or two, the advancement in this space will be tremendous, and the state of the art drive I purchase today will most likely collect dust in my closet along with the 400GB IDE hard drives. So I wouldn't worry about it too much.

    -David


     
  10. taltamir

    taltamir Lifer

    Joined:
    Mar 21, 2004
    Messages:
    13,574
    Likes Received:
    0
    well, you can actually measure those things, unless intel is lying about its measurements. You can easily measure how much write multiplication you have, and you can easily test how many writes a cell can take by rewriting a specific cell again and again and seeing how long it takes it to wear out.
    This is different than determining when a mechanical device will fail.

    Intel is not the only one who measure it, and no home user does 20GB every day 24/7, 365 days a year. sure you get 50 gigs on some days... but then not even 1 GB on most others. I have actually done the math in detail in another thread. if i find it ill post it.
     
  11. taltamir

    taltamir Lifer

    Joined:
    Mar 21, 2004
    Messages:
    13,574
    Likes Received:
    0
    found it: http://forums.anandtech.com/me...=2297124&enterthread=y

    lemme quote myself:
     
  12. IntelUser2000

    IntelUser2000 Elite Member

    Joined:
    Oct 14, 2003
    Messages:
    4,073
    Likes Received:
    5
    The new firmware likely increases the lifespan. The claims of "adjusting" to the usage to maximize performance was in the presentation but only with the new firmware it is capable of doing that. Reliability features and performance are closely related for flash storage.

    Calculating lifespan is really not that simple. The first generation EEEPC users with 4GB SSD had some of its blocks failing after mere 6-7 months. See with ANY people spending ridiculous amounts of money to get SSDs are likely enthusiasts or at least heavy computer users. And please know data retention capability decreases as the flash cells reach closer to their lifespan.
     
  13. taltamir

    taltamir Lifer

    Joined:
    Mar 21, 2004
    Messages:
    13,574
    Likes Received:
    0
    6-7 months is the EXPECTED and CALCULATED lifespan for the beginning sectors of a drive with no wear leveling and a journaled filesystem (like NTFS)... actually even FAT would have the fie allocaton table on the first few blocks and rewrite it EVERY TIME that ANY file is modified. constant rewriting of the same single cell with 2x write amplification it will wear it out in a year and a few months according the the very same calculations i did. if it wore out in 6-7 months thanit probably has higher write amplification and no wear leveing. Modern SSDs all have wear leveling so this will not occur with them.

    there can always be an undiscovered FAULT that causes the drive to break early, but MLC chips, as a technology, have a known tolerance of 10k writes which allows you to calculate accurately how long it will take to run out of writes on a wear leveled drive.. that being said, the drive can fail for OTHER reasons.
     
  14. Idontcare

    Idontcare Elite Member

    Joined:
    Oct 10, 1999
    Messages:
    21,130
    Likes Received:
    0
    And even that is just a bare-minimum spec from a reliability lifetime characterization point of view during the flash cell development phase.

    The reality is that in the worst case situation the cells wear out shortly after 10k writes, but only a small percentage of them will wear out so early. And with cell-remapping on SSD's the fact is you will get to continue using the remaining good cells on the chip well past the 10k point.

    Since the fundamental failure mechanism in Flash are activation barrier based kinetics (Arrhenius in nature) there will be some 80-90% of the cells that survive to 20k writes, and another 80-90% of those will live to see 30k writes...your standard half-life decay mathematics apply.

    Its not to say it can't be a problem if wear-leveling is not handled properly, but when handled as Intel has handled it (and as it appears that OCZ and others have done) it is a concern that is pretty much relegated to the non-issue file folder.