Storage Ssd speed...

Discussion in 'Hardware' started by TobletDanillio, 10 Feb 2012.

  1. TobletDanillio

    TobletDanillio Minimodder

    Joined:
    7 Jan 2012
    Posts:
    379
    Likes Received:
    5
    I want to find out the top speed of a ssd what program can I use to find this out?

    Thanks for your help guys
     
  2. towelie

    towelie How do I Internet!!

    Joined:
    1 Sep 2011
    Posts:
    399
    Likes Received:
    10
  3. BennieboyUK

    BennieboyUK CPC Folder of the Month Sep 2011

    Joined:
    24 Nov 2010
    Posts:
    1,710
    Likes Received:
    113
  4. PocketDemon

    PocketDemon What's a Dremel?

    Joined:
    3 Jul 2010
    Posts:
    2,107
    Likes Received:
    139
    Well you should always see the 'top speed' from ATTO...

    ...but remember that these purely artificial b/ms (ATTO, CDM, AS SSD, io meter, etc) provides almost completely useless data unless you have a very specific i/o type that relies upon exactly what they're testing.

    ie - whilst AS SSD can provide both large sequential & both QD1 & QD64 4K r/ws, this provides no information as to how good or bad a SSD will be compared to another when used as a consumer OS/apps/games drive since it in no way represents the i/o types used.

    The only possible exceptions are (a) for showing off with image caps on forums & (b) as a single run to make sure that you've set everything up properly - comparing the speeds to ones using the same b/m, SSD & disk controller from a reliable web site.
     
  5. TobletDanillio

    TobletDanillio Minimodder

    Joined:
    7 Jan 2012
    Posts:
    379
    Likes Received:
    5
    I didn't :( at work we have got a crucial m4 and I wanted to know how fast it is on our unit that has sata 2 not 3 its only a atom board too with a max of 2 GB ram.
    Our pcs are slow lol and putting windows 7 on them is just a joke.

    spec
    Atom n270 (1.6GHz)
    DDR2 533MHz 1 or 2 GB
    160GB laptop driver or 2GB dom for XP embedded
    And now they are trying to get the crucial m4 but we have had a problem with our demo unit where it blue screened with our customer :( lol but I just want to know how much better it is now with a SSD over the HDD


    I know but I'm not sure what test I can do to find out real life speeds.
     
  6. PocketDemon

    PocketDemon What's a Dremel?

    Joined:
    3 Jul 2010
    Posts:
    2,107
    Likes Received:
    139
    Tbh, it's not worth attempting to... ...well, whilst there are programs like PC Vantage, they aren't hugely meaningful so...


    Well, unless you've a really specific usage in mind which relies upon a certain data transfer size, i'd just look at Anand's average (since this uses a typical balance of both reads & writes for a consumer OS, apps, etc usage) light & heavy b/ms for whatever SSDs you're thinking of...

    ...the advantages being that, unlike many other sites, all of the current SSDs have been tested on the same machine with the same drivers & they've been created by anaylising real i/os - so the whole range from sub-4K to massively sequential, which is what you get irl.


    The only slight disadvantage is that they're not constantly kept up to date as f/ws with minor tweaks are released - though you can judge trends by comparing Anand's AS SSD/atto/etc results with ones stuck on forums or whereever for the same SSD on a very similar setup.

    (well there are more than enough people who like to post their b/ms at the drop of a hat so it's not difficult to see how they're progressing)


    Personally, i think that the heavy b/m is far more important - as it's when you really push the usage that you want the speed (much as you'd buy a top end gfx card for playing new games at a high res - not for solitaire)...

    ...but if you're only ever doing some basic browsing & the odd bit of word processing & whatnot then it's the light one that'll matter.
     
  7. TobletDanillio

    TobletDanillio Minimodder

    Joined:
    7 Jan 2012
    Posts:
    379
    Likes Received:
    5
    I just need to find out also how long this SSD will last.
    I know its hard to tell with this SSD as crucial have said that it will last 32TB (64GB model) of written data but that is 20GB a day over 5 years but I want to know what the write cycle i mean I have worked it out but its something like 500 and that seem very low. OCZ SSD with the same chips have 3000 which seems better to me.

    I just want to know how long Crucial m4 64GB will last.

    I know it will take 11 minutes 22 seconds to full the drive at full speed and to do the 32TB would take 93.5 hours at full speed (3.9 days)

    but that Sata 3 not 2 which is what we are using
     
  8. PocketDemon

    PocketDemon What's a Dremel?

    Joined:
    3 Jul 2010
    Posts:
    2,107
    Likes Received:
    139
    Why the devil would you want fill the SSD & then empty it & then fill it &... a foolish number of times in straight succession?


    Now, as a bit of background, Crucial's 32TB stat will be based on (i) a normal usage pattern, (ii) the effect of the OP & (iii) the min spec for the nand; so -

    (i) when most consumers use their SSDs, x amount will be the OS/installed programs/games & the remainder will be day to day data which will get replaced & there's the shifting around for wear levelling & block combining & whatnot - ie a mix of (largely) static & dynamic data, a mix of file sizes & the need for the SSD to move data about itself.

    (ii) & the stat will also take into account the use of the OP area - which, whilst not a specific area of nand, helps to reduce the write amplification.

    (iii) whilst the nand used in the M4 is rated for 3000 cycles, this rating is simply the point where a small fraction of the nand has typically failed - you could, by chance, have shonky nand that fails quickly or (whilst the better quality nand should be binned out for use as eMLC/HET) 'super nand' that lasts for 10s of thousands of cycles, but on average it will last for >3,000 cycles.

    - & as such the stat is simply a guide figure assuming these circumstances are followed.


    Now, what you're proposing as a 'test' (ie filling & deleting & filling...) doesn't fall within how Crucial will have calculated things on the basis of (i) & (ii) at all.

    Well, writing a file which is exactly the size of the SSD (which is what you'd need to do to be getting the max speed), then secure erasing, then writing... ...'should' have a write amplification of 1 on any non-SF SSD (the SFs compress data so it will be at least slightly lower - unless it's highly compressed data to start with, this will average around 0.6).

    [NB There will be no wear levelling - as each run through the test you'll be writing first to the nand with the lowest count - or block combining - as it's a single file.]

    This means that, using this testing methodology & average 3,000 cycle nand, you could fill the SSD >3,000 times over.

    But any OP area will not get written to on each run so this will increase the no of times.

    [NB i can't instantly recall what OP's used for the M4.]


    Conversely, if, rather than filling & emptying of the entire SSD, you instead tried to use a consumer SSD solely for a huge no of random 4K writes, they get hit far more significantly - using Anand's worst case data, a 256GB M4 could end up with a write amplification of 11.9.

    So, assuming this were identical for the 64GB M4 &, again, there was average 3,000 cycle nand inside, this could lower your no of actual writes to >252 times...

    ...again, plus the effect of the OP area.

    [NB you'd have to adjust your timing for the slower write speeds with 4Ks based upon the QD used.]


    Going back to Crucial's stat of 32TB... ...the formatted capacity of the 64GB M4 is 59.6GB in Windows, & 32TB is 32768GB, so they are assuming that, in normal usage, it can be written to ~550 times...

    ...which, based on a min cycle count of 3000, is a write amplification of ~5.45.

    [NB this seems somewhat on the high side, but it's solely calculated on Crucial's figures so who am i to argue?]


    Moving on to your comparison with a SF, assuming you've got identical spec nand in there (which is pretty common - though there are 6Gb/s consumer SFs with 5000 cycle nand), using a normal usage pattern - ie you're not -

    (i) writing nothing but highly compressed files
    (ii) or going to the some extreme like writing nothing but random 4Ks

    - they have an average write amplification of <0.6 (based on Anand's testing rather than SF's claim of 0.5).


    This has nothing at all to do with the nand rating (as it's at least semi-identical, if not actually identical, stuff), but instead d.t. the SFs compressing data as they write to wear out less nand & having much better GC than all non-SFs (the Samsung 830 being the nearest 2nd place).


    So, obviously assuming that there's no inherent issue with either drive that caused it to fail prematurely, based solely on known data, a normal consumer usage & not increasing the OP from spec#, the nand in a SF will allow ~9x more writes.

    [#NB as said many times previously, if you increase the OP by underpartitioning then you increase the longevity - see, for example, how increasing the OP (from a starting point of 37.5%) by an extra 20% affects the 4K longevity in the "Product Spotlight" data on the right of this info for the intel 710.]
     
    Last edited: 11 Feb 2012

Share This Page