• Re: HDTune - fast at end? [1/1]

    From Paul@21:1/5 to J. P. Gilliver on Wed May 3 20:56:40 2023
    XPost: alt.os.linux, alt.windows7.general

    On 5/3/2023 4:55 PM, J. P. Gilliver wrote:
    In message <u0p3bc$qnoo$1@dont-email.me> at Fri, 7 Apr 2023 08:46:36, Paul <nospam@needed.invalid> writes
    []
    While in Windows, run HDTune benchmark, and look for "bad spots" in the curve.
    []
      https://www.hdtune.com/files/hdtune_255.exe   # ten year old free version
    []
    On a hard drive, the outer circumference offers better rates than the
    hub does, which is why the benchmark curve gently declines to half-rate.
    When you see stairsteps in the bench curve, that is "zoned recording",
    and the formatting of the tracks changes from one part of the disk
    to another, on purpose. The stair steps then, are normal, and part of
    design.
    []
    I have (what I think is a) fairly conventional HD (500G, in a second-hand laptop I bought in January), which displays a fairly normal HDTune curve - flattish up to about 40%, then gentle rolloff to about half speed: but then hops up again to a high
    speed for the last 2% or so! It's consistent - two pairs of runs (I always run it twice) about 3 months apart. (The second run also shows the Access Time - the yellow "milky way" of spots - along the bottom.)

    I'll try to include the last pair but they may not attach or propagate.

    Well, you certainly got your moneys-worth.

    It's like something from a Cracker Jack box.

    One of your plots, has HDD sequential transfer rate and SSD-like access times.

    The piece on the end, *might* be consistent with a short-stroked
    drive. A short-stroked drive only uses half of the platter (the
    outer half) and the heads never touch the hub. On a short stroked
    drive, the transfer curve starts at full rate, but at the end, it
    has only declined to about 80% or so. The transfer speed is
    mostly consistent over the storage surface.

    I am lucky enough, to have acquired just one short stroke drive,
    and there is absolutely no notation in the part number, indicating
    my Cracker Jack either. I have three drives of that model, two
    normal, one is short stroked.

    Mine is a a WD 500GB 3.5" drive which is using a 1TB platter inside,
    both surfaces are certified, and they only use the first 500GB because
    I only paid for a 500GB drive. They make up a batch of 1TB drive,
    some become 500GB drives, some stay as 1TB drives. This solves the
    problem, of having no platters available any more, to make the
    500GB drives.

    But that translation, makes no sense. You would not "jump the heads to
    the middle of the disk" for the last bit of certified storage on the
    drive. What I'm saying is, if your drive was short stroked, the height
    of the material on the right, is consistent with a short stroke drive.
    But that's just a (weak) attempt to explain where the height would come
    from.

    But your Access Times of 0.3ms blows the whole charade. Something
    like that might happen, via a user adding some sort of Samsung caching software. Still pretty hard to justify or believe.

    There are two anomalies, and I cannot adequately explain either of them.

    I'm not saying it's Space Aliens that did it, but it's Space Aliens.

    If you were running RAID, had a 500GB hard drive, a 16GB Robson cache,
    maybe there would be some way to rig that. (Check and see if your
    storage is being run by the Intel RST driver.) Like first, I have to dream
    up some materials to make this work. But then the "behavior" part of
    the observation, still makes no sense. A Robson cache, I don't think
    the curves look like that, and the label for the upper left would
    not read the way it does either. There would be an artifact of
    the presence of RST for the drive name.

    Kudos on your puzzle.

    Paul

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From J. P. Gilliver@21:1/5 to Paul on Thu May 4 06:33:57 2023
    XPost: alt.os.linux, alt.windows7.general

    In message <u2uvs8$1gp0k$1@dont-email.me> at Wed, 3 May 2023 20:56:40,
    Paul <nospam@needed.invalid> writes
    On 5/3/2023 4:55 PM, J. P. Gilliver wrote:
    []
    I have (what I think is a) fairly conventional HD (500G, in a
    second-hand laptop I bought in January), which displays a fairly
    normal HDTune curve - flattish up to about 40%, then gentle rolloff to >>about half speed: but then hops up again to a high speed for the last
    2% or so! It's consistent - two pairs of runs (I always run it twice)
    about 3 months apart. (The second run also shows the Access Time - the >>yellow "milky way" of spots - along the bottom.)
    I'll try to include the last pair but they may not attach or
    propagate.

    Well, you certainly got your moneys-worth.

    I'm very pleased with the machine: I paid £80 for it - it's a 15+", with
    the 500G and 4G, though I presume the reseller probably inserted those.
    (It can take 8G, but I wanted 7-32.) It seems very responsive. The make
    [of the laptop] is "stone" (yes, with a lower case S), which I'd never
    heard of. The only bad point is this weird loss of connection (but not
    for YouTube and Google!) after a few hours, but that wasn't there when I
    first got it (in January IIRR), so is something I've done.

    It's like something from a Cracker Jack box.

    One of your plots, has HDD sequential transfer rate and SSD-like access times.

    I nearly always do two runs one after the other, so I'm guessing there's
    some sort of buffering - or similar - that the elderly HDTune isn't
    aware of: the pair of runs I did also have the very low access time on
    the second one, though not _quite_ as drastically so (a _few_ of the
    yellow "milky way" dots are still where they "should" be, and it shows
    1.4 ms rather than 0.3).

    The piece on the end, *might* be consistent with a short-stroked
    drive. A short-stroked drive only uses half of the platter (the
    outer half) and the heads never touch the hub. On a short stroked
    drive, the transfer curve starts at full rate, but at the end, it
    has only declined to about 80% or so. The transfer speed is
    mostly consistent over the storage surface.

    But on mine, it does decline to about 50%, or would if the odd jaggy
    wasn't there at the end! So I don't _think_ it's short-stroked.

    I am lucky enough, to have acquired just one short stroke drive,
    and there is absolutely no notation in the part number, indicating
    my Cracker Jack either. I have three drives of that model, two
    normal, one is short stroked.

    Mine is a a WD 500GB 3.5" drive which is using a 1TB platter inside,
    both surfaces are certified, and they only use the first 500GB because
    I only paid for a 500GB drive. They make up a batch of 1TB drive,
    some become 500GB drives, some stay as 1TB drives. This solves the
    problem, of having no platters available any more, to make the
    500GB drives.

    I'd have thought they'd make all drives with 1TB platters into 1 TB
    drives, with those sold as 500G not being so sold to satisfy a demand
    for 500G, but because on testing they found a big fault in the second
    half. Rather like - _many_ decades ago, before PCs I think - you used to sometimes get two versions of memory chips (or it might even have been
    EPROMs) of a given capacity, one where one of the "enable" lines had to
    be high and one where it had to be low, and it was fairly obvious that
    these were chips made to twice the capacity but had had a fault found on testing in one half or the other. (Obviously for HDs, if the fault was
    in the _first_ half they'd scrap them.)

    But that translation, makes no sense. You would not "jump the heads to
    the middle of the disk" for the last bit of certified storage on the
    drive. What I'm saying is, if your drive was short stroked, the height
    of the material on the right, is consistent with a short stroke drive.
    But that's just a (weak) attempt to explain where the height would come
    from.

    But your Access Times of 0.3ms blows the whole charade. Something
    like that might happen, via a user adding some sort of Samsung caching >software. Still pretty hard to justify or believe.

    There are two anomalies, and I cannot adequately explain either of them.

    I'm not saying it's Space Aliens that did it, but it's Space Aliens.

    As I say, I suspect some sort of caching (or similar) hardware, either
    in the drive or the PC, that is fooling HDTune on the second of two
    successive runs. (It takes about 7 minutes, with HDTune left at its
    default settings for block size etcetera.)

    If you were running RAID, had a 500GB hard drive, a 16GB Robson cache,
    maybe there would be some way to rig that. (Check and see if your
    storage is being run by the Intel RST driver.) Like first, I have to dream
    up some materials to make this work. But then the "behavior" part of
    the observation, still makes no sense. A Robson cache, I don't think
    the curves look like that, and the label for the upper left would
    not read the way it does either. There would be an artifact of
    the presence of RST for the drive name.

    Certainly not (knowingly!) running any sort of RAID. (I'm pretty sure
    there's only one HD - certainly the flap on the bottom of the machine [I checked when buying it that there was a suitable access flap - I didn't
    want one of these machines where you have to take the whole bottom cover
    off to get at e. g. HD or RAM - is the normal size.)

    Kudos on your puzzle.

    I won't worry about it if it gives me no trouble! And I'm backing up now
    I've got my external drive back (unfortunately it was loaned to someone,
    to get some of my data back, when the connection funny happened, so I
    don't have a backup from before that).

    Paul


    John
    --
    J. P. Gilliver

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)