Interesting article on SSD lifespans

Discussion in 'Storage & Backup' started by metamorphosis, Oct 27, 2018.

  1. metamorphosis

    metamorphosis Member

    Joined:
    Feb 25, 2002
    Messages:
    1,760
    pinchies likes this.
  2. Catweazle

    Catweazle Member

    Joined:
    Aug 12, 2003
    Messages:
    2,867
    Location:
    Sale, Vic.
    Quite interesting to see WD Blue 3D NAND score 82, whilst SanDisk Ultra 3D scores 514.

    They're the exact same thing with different labels stuck on them.
     
    Hive and terrastrife like this.
  3. PsydFX

    PsydFX Member

    Joined:
    Jan 1, 2011
    Messages:
    3,556
  4. Catweazle

    Catweazle Member

    Joined:
    Aug 12, 2003
    Messages:
    2,867
    Location:
    Sale, Vic.
    When one measurement can be 600+% of another measurement of same thing, I'd disregard all other provided info and go elsewhere for illumination :p
     
  5. PsydFX

    PsydFX Member

    Joined:
    Jan 1, 2011
    Messages:
    3,556
    Could come down to firmware revisions, or could have jut been a bad drive.
     
  6. Catweazle

    Catweazle Member

    Joined:
    Aug 12, 2003
    Messages:
    2,867
    Location:
    Sale, Vic.
    Either way, it's "ignore that reading coz it's stupid", and every other reading is "doesn't matter how much bigger the other numbers might be, coz it's already good enough to cope with realistic usage over a realistic lifespan."
     
  7. Aetherone

    Aetherone Member

    Joined:
    Jan 15, 2002
    Messages:
    8,530
    Location:
    Adelaide, SA
    Sample size of each drive n=1?
    Yeah, not exactly a powerhaus of scientific rigor.
     
  8. HobartTas

    HobartTas Member

    Joined:
    Jun 22, 2006
    Messages:
    749
    True, but given there are so many SSD's with pretty dismal results then I very much doubt any of them are too far from the ballpark figures and this very large spread was evident very early and is still evident today but for most home users it really doesn't matter as I'm still running on a 256 GB Samsung 840 Evo for a couple of years now and I've only written 28 TB to it so far. These figures only really matter if you are running a business and hammering them with Windows Server or an active database and you need to work out longevity for the dollars you are paying.
     
  9. Skitza

    Skitza Member

    Joined:
    Jun 28, 2001
    Messages:
    3,743
    Location:
    In your street
    Out of all the ones they tested the Samsung 860 EVO isnt there. :(
     
  10. NSanity

    NSanity Member

    Joined:
    Mar 11, 2002
    Messages:
    17,501
    Location:
    Canberra
    Did they actually test? or are they comparing manufacturers MTBF's?
     
  11. OP
    OP
    metamorphosis

    metamorphosis Member

    Joined:
    Feb 25, 2002
    Messages:
    1,760
    Those're some good quotes.


    Read the article - they tested.
     
  12. Catweazle

    Catweazle Member

    Joined:
    Aug 12, 2003
    Messages:
    2,867
    Location:
    Sale, Vic.
    The problem with comparisons like this is that they infer "bigger number = better in every way". But that just isn't true, because in reality there's a concept called "more than good enough", and when you attain that any further improvement is really irrelevent.

    I just did some quick figuring. Unless my math is wrong, even for that lowly '82Tb' figure (which I discount as a probable anomoly) For the drive to fail in less than 5 years you'd have to be averaging over 44Gb of writes per day. 22Gb per day will give a 10 year lifespan. And they're 256GB SSDs. They're not going to be used for scenarios which involve massive amounts of data being written on a daily basis. They're a component suited to acting as System Drive. Repository of OS and Programs, and subjected to very little ongoing write activity.

    In other words, even if that unlikely bottom outlier number is correct, it's still a device which is good enough to last well beyond its probable actual useful working lifespan. It'll become redundant before it becomes dead.
     
  13. elvis

    elvis Old school old fool

    Joined:
    Jun 27, 2001
    Messages:
    35,673
    Location:
    Brisbane
    How many of each drive did they test?

    Google released their spindle drive tests some years back where over a million drives were tested. One thing they discovered is that drives are either likely to fail quickly (in the first few weeks of life), or last to 5 years where their chances of failure increased exponentially from there.

    This high level of variance means any testing of a single drive isn't conclusive at all. Should you be unlucky enough to test a bad unit, that will skew your stats completely.
     
    Catweazle likes this.
  14. HobartTas

    HobartTas Member

    Joined:
    Jun 22, 2006
    Messages:
    749
    So I presume you'll be wanting information like this then instead which you would use to make those decisions?
     
    elvis likes this.
  15. Catweazle

    Catweazle Member

    Joined:
    Aug 12, 2003
    Messages:
    2,867
    Location:
    Sale, Vic.
    Yep. That's a good basis. Negligible return rates pretty much right across the board indicates that product sold is of good quality for all makes/models.

    Couple that with real-world usage figures, and you find that hardly anyone even needs to worry about which drive they buy. They're all unlikely to be a dud delivered, and they're all good enough to serve most people's needs for many years.
     
  16. Mathuisella

    Mathuisella Member

    Joined:
    Dec 3, 2002
    Messages:
    6,230
    Location:
    in your gearbox...grindin
    wonder how it compared to mechanical drives ?
     
  17. Ratzz

    Ratzz Member

    Joined:
    Mar 13, 2013
    Messages:
    7,865
    Location:
    Cheltenham East 3192
    As already said by others, unless I'm not looking hard enough, I see no discussion anywhere in the article of sample size, which makes the whole thing completely irrelevant.
     

Share This Page