P4P800-E Deluxe: SATA slower than PATA drives???

Discussion in 'Asus' started by Martin Hirsch, Sep 24, 2005.

  1. Hi,

    My original configuration consisted two PATA Maxtor 160gb drives on the
    promise controller in a non-raid IDE configuration. These drives at boot up
    show up in the promise bios as UDMA6 drives and are operating at the optimum
    speed.

    I then installed a new WD 360gb SATA drive on the Intel controller with a
    fresh operating system (XPsp2) and made it the bootable drive. Since this
    drive is SATA I expected equal or better performance than the original
    Maxtor PATA drive installation but it was noticeably slower in every aspect
    with just the operating system and a few programs installed.

    I went into the bios and noticed the SATA drive was auto configuring as
    UDMA5 on the Intel controller.

    I then plugged the drive into the promise SATA connector on the MB and again
    at boot up it showed up as a UDMA5 drive where the other two PATA drives
    show up as UDMA6.

    I have gone into the bios and made all the correct adjustments I believe.
    When connected to the Intel controller I set the following:

    "Onboard IDE Operate Mode" [Enhanced]
    "Enhanced Mode Support On" [S-ATA]
    "Configure SATA as RAID" [No]

    The Maxtor drives are noticeably faster at 133mb/s than the new WD SATA
    drive which I believe is running at 100mb/s.

    How can I get the new SATA drive up to speed.

    Thanks for any help!! Martin
     
    Martin Hirsch, Sep 24, 2005
    #1
    1. Advertisements



  2. Hi Martin!


    Not even that, also (a good drived) AGP is faster than PCIx.




    Best Regards,

    Daniel Mandic
     
    Daniel Mandic, Sep 24, 2005
    #2
    1. Advertisements

  3. Martin Hirsch

    deus maximus Guest

    Keep in mind that the speed rates for the two drive types are not
    the speeds of the disks but only the maximum throughput on their
    respective buses. To compare disk speeds go here
    http://www.storagereview.com/comparison.html
    You may be suprised to learn the relative speeds of each disk you
    have.
    Secondly , just because a disk is UDMA 5 or 6 will not necessarily
    determine its relative speed. The site above runs apps on each
    drive so their benchmarks do give a speed comparison.
     
    deus maximus, Sep 24, 2005
    #3
  4. Martin Hirsch

    Paul Guest

    Another surprise for Martin, would be this in the ICH5 datasheet.
    This is in the section describing the PATA interface (pg.183)

    http://developer.intel.com/design/chipsets/datashts/25251601.pdf

    "reads at the maximum rate of 100 MB/s"
    "write transfers at a maximum rate of 88.9 MB/s"

    Inside a lot of the first SATA disk drives, there was a bridge
    chip. In other words, they were still using the same PATA controller
    chips as before, and to make a SATA interface, they just slapped
    a chip onto the controller board, to bridge from SATA to PATA.
    In fact, some of those bridges, only run at 100MB/sec.

    There are all sorts of bottlenecks in the disk architecture,
    and only benchmarking and looking at burst performance and
    sustained transfer performance, will hint at exactly what you
    got when you purchased your disks.

    The best disks only sustain about 70MB/sec at the heads, so none
    of these bottlenecks is really killing sustained transfer rate. Any
    arguments here, are about burst-to-controller-cache performance.

    Also, the seek time is an important parameter too. If looking
    for files all over the disk, the answer only comes back, as
    fast as the heads can be moved to the new area on the disk.
    The biggest enabler there, is higher RPM rates, like 10K or
    15K, instead of 7200 RPM. At 15K, SCSI is the answer.

    Paul
     
    Paul, Sep 24, 2005
    #4
  5. Martin Hirsch

    Newz Guest

    You're comparing 2 different brands of HD, two different controllers
    and two different buses.
    This makes it a bit hard.

    First of all, is that WD HD a native SATA drive? WD so far has just
    put SATA bridge chips on their SATA drives, meaning that the drive is
    simply PATA. If you have one of these SATA-II drives, they are native
    SATA.

    Second, did WD ever embrace ATA-133? I believe Maxtor was the first to
    do that and the others never really followed (IBM-Hitachi, WD). They
    shouldn't need to either, as ATA-100 can cope with anything the drive
    throws at it. Especially when you have 2 separate channels as the
    Promise controller has.

    Third, is your WD drive actually faster than the Maxtor 160? You need
    to have a look at independent reviews. Don't expect SATA to be faster
    than PATA. Because both buses are faster than the drives themselves,
    the differences will likely be related to the drive performance.

    Fourth, are you running the Intel Application Accelerator? Depending
    on your motherboard chipset, this may give you a performance increase.

    If you really want to test what's happening, you should run a Maxtor
    on the Intel controller PATA side. That would allow you to rule out
    the controller. Not entirely though, as the SATA interface on Intel
    controllers is faster nowadays than the PATA.

    Good luck!

    Grtz,
    Nwz
     
    Newz, Sep 25, 2005
    #5
  6. My WD SATA model is WD3200JD and my PATA Maxtors are Y160PO. I don't believe
    the WD is a SATA-II drive.

    I have run the WD SATA drive on the promise controller with identical
    results. It is noticably slower.

    I'm not running the intel app accelerator. This would only benfefit drives
    on the intel controller right?? If so I will install it and put back the WD
    drive on the intel controller and see what happens.

    If I run my Maxtor on the intel controller as you suggust , wouldn't that
    hurt performance since these are 133 drives and the intel controller maxes
    out at 100 for PATA??
     
    Martin Hirsch, Sep 25, 2005
    #6
  7. Martin Hirsch

    Paul Guest

    If you run the Maxtor on the Promise controller, you are limited
    by the practical burst rate on the PCI bus. While PCI is theoretically
    133MB/sec, practically it is 100-110MB/sec. (To maximize it, you can
    crank up "PCI Latency Timer" in the BIOS. That will give you a
    better looking benchmark, but will cause other things not to work
    smoothly on your computer, like perhaps audio.)

    If you run the Maxtor on the Southbridge, you are limited by the IDE
    interface design. (The Intel datasheet claims the Southbridge does
    have good "plumbing" inside, as the interface to the Northbridge
    runs at 266MB/sec. It is just the interface to the disk that doe
    not care to embrace the ATA133 standard.)

    A native SATA drive (no bridge chip inside), plus the Southbridge,
    is the only way on that motherboard to get a burst rate
    higher than about 100MB/sec.

    Southbridge benchmark RAID0 - Raptor disks can manage 2x70MB/sec
    by spec, and the benchmark maxes at 122MB/sec write, and 100MB/sec
    read.

    http://www.abxzone.com/forums/showpost.php?p=595566

    Promise benchmark RAID0 - the poster thinks this is hot stuff ?

    http://www.abxzone.com/forums/showpost.php?p=650264

    Based on the Southbridge benchmark, it still appears that
    really good transfer rates will not be possible. Why it only makes
    it to 122MB/sec is a puzzle.

    I would say there _is_ something up with the innards of the
    Southbridge. For example, look at this 4 disk RAID0 array
    on an ICH7R Southbridge. It is flatlined at 160MB/sec for
    most of the surface of the array, implying there is a
    "plumbing" problem inside the Southbridge. That is only
    40MB/sec per disk.

    http://www.abxzone.com/forums/showthread.php?t=94560&highlight=ich7r+raid

    Maybe the only way to get flashy benchmarks, is go S939 ?
    (Haven't seen any benches for those...)

    The results here look a bit different, so perhaps this
    is some kind of benchmarking issue:

    http://www.gamepc.com/labs/view_content.asp?id=serialcompare&page=6

    Paul
     
    Paul, Sep 25, 2005
    #7
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.