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

Storage Which SSD to get?

Discussion in 'Hardware' started by Frohicky1, 29 Nov 2010.

  1. Frohicky1

    Frohicky1 Awaits his moosey fate . . .

    Joined:
    16 Apr 2009
    Posts:
    1,661
    Likes Received:
    9
    Hi all,

    My 150gb velociraptor has served me well but it's starting to chug a bit, so methinks SSD time. I haven't been keeping with the times very much, and am wondering which SSD of around the 128gb mark (or multiple smaller ones) would be fastest. Something less than £200 would also be nice :D For example I know that some SSDs (Sandforce based?) use on the fly caching, which sounds too variable for me, and that others (eg SLC) are too expensive.

    So, what say you learned people? Where should my money go (and, in true English Lit form, why?)
     
  2. PocketDemon

    PocketDemon Modder

    Joined:
    3 Jul 2010
    Posts:
    2,107
    Likes Received:
    139
    Rather than re-typing everything, i'd suggest you look at this thread (a couple of ones below yours) as i believe i've given a reasonable appraisal of the C300 vs V2 vs raid argument...

    (plus there's the link in there to my 'disagreement' with bittech about their trim vs GC testing methodology & conclusions)


    [EDIT]

    Oh, & the SFs don't use "on-the-fly caching" as they don't have a cache... Afaik, every other gen2 SSD does use a cache though (it's what differentiates them from Gen1)


    They instead use "on-the-fly compression" - which is entirely different & is about writing to fewer nand cells in order to reduce wear/increase longevity...

    ...simply put, this gives a range of (predominantly - unless you really abuse it by copying all the files in the world on & off for the sake of it - in which case the reads will slow down until it's had sufficient time to recover) write speeds between ~270MB/s for 100% compressible data & ~100MB/s for 100% incompressible data for the 50/60GB models & ~110MB/s for 100% incompressible data for the 100/120GB ones.

    in r.l. usage (rather than b/ms) this will average somewhere in between - for most usages, faster than the 128GB C300 for writes...
     
    Last edited: 29 Nov 2010
  3. Frohicky1

    Frohicky1 Awaits his moosey fate . . .

    Joined:
    16 Apr 2009
    Posts:
    1,661
    Likes Received:
    9
    Whoops, mistype, meant on-the-fly compression. Thanks for post, didnt' see that one, will check it out . . .
     
  4. Pot Of Jam

    Pot Of Jam What's a Dremel?

    Joined:
    6 Sep 2010
    Posts:
    58
    Likes Received:
    0
    I have 2 64GB C300's, one in my desktop and one in my EEE both running 7, and granted I have only ever had these SSD's, I still think there a best buy as there SATA 3 and future proof.

    Also, there pretty nippy even in my SATA2 EEE.

    Crucial are selling the 128 for under £200 direct from there site, i ordered my 64gb on friday and turned up on sat!
     
  5. PocketDemon

    PocketDemon Modder

    Joined:
    3 Jul 2010
    Posts:
    2,107
    Likes Received:
    139
    Just because they're SATA3, that unfortunately doesn't make them "future proof" in the slightest...


    Well, forgetting that nothing is ever future proof a couple of techs that are known to be on the way -

    SF have already said they've reached 500MB/s r/ws from their new 2000 series of SSDs (should be early next year) -> which 'should' at least give something like ~480-90MB/s sequential reads & a range between ~480MB/s & 210MB/s for 100% compressible & 100% incompressible sequential writes for the 60GB model (writes improving on the higher models by ~10MB/s per size increase for 100% incompressible data).

    [whilst intel seem to be initially going for capacity rather than speed increases next year, it would be 'slightly' surprising if other SSD controllers didn't follow the route towards the ~560-70 max bandwidth of 6Gb/s SATA...

    This is why i've always differentiated between the C300s & a 'proper' 6Gb/s SATA/SAS implementation.]


    & then, completely separately, you have OCZ's HSDL interface (yes the ibis already exists, but it'll be the v2 ones towards the end of next year when things will become really interesting), which (currently) has the potential for 2GB/s r/w speeds per SSD - with a, 'apparently' relatively cheap, 16x controller allowing 4 of them to run at these speeds on a pcie 2.0 controller.

    [HSDL is currently implemented by using 4 lanes of pcie 2.0 per SSD -> 4xSATA raid controller --> SAS cable ---> SSD (atm with an onboard raid controller) - though the idea is that there will be more elegant solutions developed rather than using available tech to keep OCZ's development costs down.]

    This tech though, is easily scalable to provide each SSD with up to 4GB/s r/w speeds using 4 lanes when pcie 3.0 arrives.

    The big barrier to HSDL initially though will be cost but, as with anything, an improved version of the ibis tech will quickly drop into the higher end of the mainstream once there's the better versions out there.


    Yeah, so a 340MB/s read & 70/140/215MB/s (depending on the model) write speeds per SSD are anything other than "future-proof" in the relatively short term.

    [edit]

    Oh, & of course a pair of SFs (or at least the higher end year & a half plus old 120GB indilinxes (as the 30/60GB ones were slower - can't be bothered to look them up atm) vs the 256 C300) in R0 will outperform a C300 already - so the C300 isn't even future-proof now (unless you've got enough cash to spend on a decent 6Gb/s raid controller - though, again, the GC on the C300 isn't great).
     
    Last edited: 30 Nov 2010
  6. Pot Of Jam

    Pot Of Jam What's a Dremel?

    Joined:
    6 Sep 2010
    Posts:
    58
    Likes Received:
    0
    Ok so there more future proof then a sata2 drive.

    2 types of hardware prototype and obsolete.
     
  7. PocketDemon

    PocketDemon Modder

    Joined:
    3 Jul 2010
    Posts:
    2,107
    Likes Received:
    139
    You're clearly using some new definition of the term "future-proof" that i've not heard of...

    Well, (a) if something can already be bettered by alt technology atm (ie the pair of SFs, the ibis & the various pcie options) (b) will be superseded within ~4 months by vastly quicker technology using the same basic tech (ie when 'proper' 6Gb/s SSDs come onto the market) & (c) the something it's using is going to become obsolete for the fastest tech (HSDL not being backwards compatible with SATA/SAS afaik), then it's neither "the future" now (re a), nor, quite obviously, "future-proof" (re b & c).

    i'm sure it would be technically possible to make a floppy drive connect via 6Gb/s SATA, but that's not suddenly going to make it a "future-proof" high speed format, is it???
     
    Last edited: 1 Dec 2010
  8. Cleggmeister

    Cleggmeister Of reasonable knowledge...

    Joined:
    12 Oct 2009
    Posts:
    1,140
    Likes Received:
    22
  9. Frohicky1

    Frohicky1 Awaits his moosey fate . . .

    Joined:
    16 Apr 2009
    Posts:
    1,661
    Likes Received:
    9
    Hmmm, maybe its worth waiting till next year . . . I know theres always something new around the corner, but a decent Sata6gb/s would be tasty, and an excuse to get a new mobo . . . :)

    What's the current situation with raid0 and trim? Has that been resolved?
     
  10. PocketDemon

    PocketDemon Modder

    Joined:
    3 Jul 2010
    Posts:
    2,107
    Likes Received:
    139
    No - but if you ignore bittech's overreaction & faulty testing methodology of trim vs GC (ie no one fills & empties their SSD 10 times in a row in normal usage) then you'll find there's no real issue unless you're writing vast amounts of data on a daily basis (see here for the discussion i had with them)...


    Of course, some SSDs have much better GC than others...

    ...of the fairly recent controllers, some of the later indilinx ones have the most aggressive - the SFs being in the middle - & the C300 having one of the slower GC implementations...

    ...though you can use the overnight log-off method (S1 or S3 for indilinx, S1 for SFs &, i believe S1 & S3 for the C300s) if speeds start to suffer d.t. heavy r-e-w cycles.
     

Share This Page