1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Build Advice Best 500GB SSD for under £180

Discussion in 'Hardware' started by Farting Bob, 25 Apr 2015.

  1. Farting Bob

    Farting Bob New Member

    Joined:
    21 Jan 2009
    Posts:
    469
    Likes Received:
    13
    Hi, looking to build a new system and ive been out the loop with SSD's, what is currently the best bang for my buck SSD for under £180?
    Ive been tempted by the Samsung 850 EVO since it generally sits near the top of most charts but i know it's still got long term performance question marks around it. Is there any better performing or substantially cheaper options i should consider?
     
  2. b1g-d0g

    b1g-d0g Well-Known Member

    Joined:
    30 Sep 2001
    Posts:
    2,532
    Likes Received:
    38
    Just bought the Samsung 850 EVO mate im very happy with it.
     
  3. IvanIvanovich

    IvanIvanovich будет глотать вашу душу.

    Joined:
    31 Aug 2008
    Posts:
    4,870
    Likes Received:
    252
    I'm not sure there is any real evidence the 850 evo has any issues like it's predecessor with performance deterioration. It consistently gets top reviews and recommendations so I would think it's a good choice.
    Crucial BX100 or MX200 seem to be the other top picks right now in that size/price tier.
    Some others that are considered to be quite good are SanDisk extreme pro or ultra II, Transcend SSD370, Plextor M6S, Corsair Neutron GTX, Plextor M5 Pro Xtreme, and ADATA Premier SP610.
    Whichever one can be found on best sale at the moment.
     
  4. murraynt

    murraynt Well-Known Member

    Joined:
    6 Jun 2009
    Posts:
    4,234
    Likes Received:
    128
    The BX100 is 129.99 on amazon.co.uk atm.
     
  5. PocketDemon

    PocketDemon New Member

    Joined:
    3 Jul 2010
    Posts:
    2,107
    Likes Received:
    139
    There's no inherent reason why the 850 Evo should have the same issues as the 840.


    Basically, the issue with the 840 is about the charges in the nand cells decaying over time - causing a need to aggressively re-read (& use the checksums & whatnot) for cells that haven't been refreshed for x months in order to ensure data integrity...

    ...hence the new solution is to, either manually (with Magician 4.6) or automatically as a background process with the new f/w, rewrite all of the cells periodically as this refreshes the cell charge.


    Now, charge state decay is a natural factor with all nand - indeed it's the reason why consumer nand is only rated to retain data for 1 year at 30C in a power off state (the rating improving as temp drops & significantly falling as temp rises - so increasing to 35C can typically cut the retention rate by half) - which is a cross brand JEDEC standard that everyone's signed up to...

    ...however it's more pronounced as you increase the number of bits held in a cell (as there's then significantly more overlap between the charge states at the same cell size) & reduce the nand size.

    That said, there is a design flaw with the TLC nand used in the 840 Evo as the (8) states per cell are narrowing far too quickly in a powered state - causing the read issue on older data... ...& hence the new solution being to periodically rewrite the data.


    Now, accepting that this shouldn't have happened in the first place (although there's now a solution that will work - albeit at a cost to longevity that should be irrelevant to almost all consumers), a key difference between the 840 Evo & 850 Evo is the nand cell size...

    So the 840 Evo used a 19nm process, whilst the 850 Evo uses a 40nm one (i believe it's actually 42nm)...

    These aren't necessarily straight figures though - as whilst for example, a 19nm nand cell is at least 19 x 19nm square... ...it can also be asymmetrical - so, say, 19 x 26nm nand which increases the cell's total area.

    This means that, for example, the cells in imft's (intel & micron's) 20nm nand which is symmetrical have a smaller area that Toshiba's 19nm nand which is 19x26nm - & so everything else being equal Toshiba's nand would be more robust in terms of maintaining charges as there can be more space or less overlap between the charge states.


    Okay, so there's no info that i can immediately see for Samsung's processes - however, the key metric is the decrease in the overlap between the charge states for the two processes when used for TLC - which is a factor of 10.

    [NB everything else being equal, this *should* logically translate to an average improvement of somewhere between >10 (prob 20 min) & ~50x before the issues could occur - though it would take pages & loads of diagrams & things to explain my thinking on this... & this post is long enough already... ...but even at say 15 times, this would take it beyond the 3 year warranty.]​

    Now, this therefore means that even *if* there was the same inherent fault with powered state charge decays (ie Samsung hadn't actually cured the issue with their V-nand), then the length of time before it would became an issue would be dramatically increased...

    ...and, if it did occur (ie that i was incorrect in my logic so that it 'could' occur during the warranty period), they already have a s/w solution that would need to run far less frequently than with the 840 Evo.


    Anyway, back to the original question then if you're after top performance for your budget without waiting on a deal then the 850 Evo is where it's at atm...

    ...unless, for some unknown reason, your build can't provide trim & you've got a heavy write erase usage as the consumer Samsungs really aren't good in that kind of environment - & you'd then need to be spending more money on something like a Plextor M5 or M6 Pro than any of the budget drives.

    it is always worth having a bit of a check for deals though as, for example, having a non-trim heavy write usage, i managed to find a new 512GB M5 Pro roughly 2 months back for ~£154, which was only £11 more than i paid for a new 500GB 850 Evo at roughly the same time.
     
    Last edited: 26 Apr 2015

Share This Page