Sign in with
Sign up | Sign in

Benchmark Results: Iometer 4 KB Random And Streaming Read/Write

Does Your SSD's File System Affect Performance?
By , Achim Roos


The 4 KB random write testing really serves as a poignant reminder of FAT32's age. Considering what we saw in AS SSD and CrystalDiskMark, the results don’t come as a surprise. It is obvious that the flexible cluster size of FAT32 at large drive capacities becomes an issue for write operations. The 4 KB chunks have to be accommodated into 32 KB clusters. exFAT does the job much better than FAT32, although it may also have large cluster sizes. Since its free clusters are indexed, however, finding one is a simple matter of looking up the corresponding bit in the bitmap.

Again, the results on sustained throughput resemble those seen in our earlier tests, and there are really no surprises there.

Display all 45 comments.
This thread is closed for comments
Top Comments
  • 18 Hide
    neon neophyte , April 13, 2012 5:16 AM
    I remember the crossing from Fat32 to NTFS. It was significant even back then. Ever since I have craved a new file system offering to rekindle a fading memory of youth and joy. *sniff*
  • 17 Hide
    trumpeter1994 , April 13, 2012 1:30 PM
    MarthisdilHardly no one uses Linux in a home environment, thus, ext4 and linux whiners need to stop.

    I don't run linux, but since it has such a dominant presence in the servers you connect to every day...... yes it is relevant
  • 16 Hide
    haplo602 , April 13, 2012 8:35 AM
    any other than windows/mac filesystems ? zfs ? btrfs ? ext3/4 ? jfs ? xfs ?
Other Comments
  • 9 Hide
    aznshinobi , April 13, 2012 4:14 AM
    Those SSD drives.... *drool* Wish I could afford them.
  • 18 Hide
    neon neophyte , April 13, 2012 5:16 AM
    I remember the crossing from Fat32 to NTFS. It was significant even back then. Ever since I have craved a new file system offering to rekindle a fading memory of youth and joy. *sniff*
  • 0 Hide
    hmp_goose , April 13, 2012 5:31 AM
    [misses HPFS]

    [wonders what sectors per cluster means to an SSD]
  • 1 Hide
    aicom , April 13, 2012 6:15 AM
    hmp_goose[misses HPFS][wonders what sectors per cluster means to an SSD]


    NTFS was heavily based on HPFS (when MS and IBM were both working on OS/2). It even shares the same MBR partition type code.
  • -5 Hide
    confish21 , April 13, 2012 6:26 AM
    get article ty so much!
  • 16 Hide
    haplo602 , April 13, 2012 8:35 AM
    any other than windows/mac filesystems ? zfs ? btrfs ? ext3/4 ? jfs ? xfs ?
  • 14 Hide
    lorfa , April 13, 2012 8:41 AM
    Agree with haplo. Wanted to see ext4 at least.
  • -6 Hide
    Anonymous , April 13, 2012 9:56 AM
    billafuEnjoyed the article. Sadly, I am still unable to justify spending nearly a dollar per gigabyte for an SSD when HDDs are less than a dime per gig. Maybe when that price difference is a little bit closer.

    120gb for a 120$ and HUGE performance increase and you still complain? How about you get a job.
  • -4 Hide
    Badelhas , April 13, 2012 10:14 AM
    doctorpink120gb for a 120$ and HUGE performance increase and you still complain? How about you get a job.

    Mega LOL!
  • 9 Hide
    ojas , April 13, 2012 11:17 AM
    lostmyclantoms is partner of micosoft I want some linux test =) 2012 and nothing about linux ?

    I wonder what it means when they say
    Quote:
    For this piece, we're going to go into more depth on file systems with a focus specifically on Windows users, since our rigs in Germany are all Windows-based.
  • 10 Hide
    baynham , April 13, 2012 11:40 AM
    ext4 please


  • 6 Hide
    AndrewJacksonZA , April 13, 2012 11:45 AM
    Thanks for the article. It answered some questions that I'd been pondering for a while. I'm a bit disappointed that you missed ReFS which has debuted in Windows 8/Server 8 - even though the OSes are still in beta.

    And ext3/ext4. And yes, I read that your German labs are Windows based, but still, it would've been nice. How many enthusiasts and admins that read this use ext3/4 is another question. :-)

    Thanks.
  • 12 Hide
    jclambert1 , April 13, 2012 1:12 PM
    I use linux at home regularly - in my primary laptop and file server
  • 17 Hide
    trumpeter1994 , April 13, 2012 1:30 PM
    MarthisdilHardly no one uses Linux in a home environment, thus, ext4 and linux whiners need to stop.

    I don't run linux, but since it has such a dominant presence in the servers you connect to every day...... yes it is relevant
  • 7 Hide
    Anonymous , April 13, 2012 2:26 PM
    I think these tests could also include popular Linux filesystems, such as ext4 and BTRFS, as they seem to have some optimizations for SSD-based drives... from some tests (you can find them on Phoronix), they swiftly beat NTFS/FAT filesystems...
  • 4 Hide
    Vatharian , April 13, 2012 2:31 PM
    Most of MLC-based SSDs around are used as a boot drive. On Windows there is completely no choice on which FS to install on. FAT32 is too dangerous, since it offers no protection/detection of corrupt writes. So for casual windows users article holds no meaning. People who use SSD for workstation based work, eg. video processing or databases are forced to use NTFS, because FAT32 can't handle 4GB+ files and exFAT holds no protection, since there are very little repair/recovery tools for it (especially freeware ones). Also portability suffers. Mac users are pretty much forced to use HFS+. For other uses, non-raided ones, data serving can be done on ext4 or xfs, and I think the last one would need to be thrown in, following optional ext4. Other question is how OS handles the FS. I'd want to see HFS+ partition mounted under Windows and NTFS mounted on Mac (ntfs-3g works!). I even use linux installed on same ntfs volume as my Win7). About WinRE - does all the benchmarking tools work with it? And last, but not least - what about testing these drives in raw mode and comparing this to overhead thrown in by FS?
Display more comments