MSI P6N SLI Platinum Overclocking

Discussion in 'MSI' started by JD, Jun 16, 2009.

  1. JD

    JD Guest

    I have never overclocked a board before but.....

    Games are running kinda slow and my proc and board are getting to be a
    few years old... I would like to squeeze another year out of this
    motherboard and proc so if someone can tell me how to adjust it would
    be appreciated.

    I have the following

    MSI P6N SLI Platinum
    4 case fans so plenty of cooling
    E6600 Proc

    Just bought a ge force 260 card for it - whatever I do to the MB /
    Proc I just dont want it to fry my new video card...

    Looking for some settings that are moderately better - I dont need
    something that is going to shoot my house to the moon.

    I have search for settings, but for the most part it looks like people
    pushing the settings to the limits.

    If anyone can help me it would be appreciated.

    Thanks

    JD
     
    JD, Jun 16, 2009
    #1
    1. Advertisements

  2. JD

    Paul Guest

    OK, here are 439 overclocking results for E6600.

    http://www.hwbot.org/listResults.do...filterUser=true&filterBlocked=true&limit=1000

    There seem to be a few 3600MHz overclocks, about 50% above
    stock. The verification link for one of them is here.

    http://valid.canardpc.com/show_oc.php?id=376057

    The CPU input clock is raised to 400Mhz. That is
    FSB1600. 400MHz x 9 = 3.6GHz. The RAM uses
    the 4:5 divider, so the memory clock is 500Mhz or
    DDR2-1000. There could be other dividers, that
    don't push the memory as hard.

    The memory clocks up at the same time as the processor.
    So if you want to raise the CPU clock, and keep the
    memory relatively the same, you need a memory divider
    to drop the memory a bit to compensate.

    The normal practice when overclocking, is to raise a
    clock a little bit at a time. The idea being, you're
    changing the clock only enough to help you spot trends.
    You boot into Windows and run Prime95 stress test.

    http://majorgeeks.com/Prime95_d4363.html

    Say you raise it 5MHz at a time. You start at 266Mhz
    (the nominal value), then try 271MHz. Boot and try
    Prime95 for 10 minutes. If there are no errors,
    you shut down, enter the BIOS, and try 276MHz.

    Eventually, you reach a point, where Prime95 errors
    out in less than ten minutes. You're still able to
    boot into Windows, because the idle desktop is
    less demanding than Prime95 is. So the Prime95
    is like a "canary in a coal mine". It warns
    about impending doom.

    Once you get the warning like that, the next step,
    is to increase Vcore a bit. Now, I don't know right
    off hand, what the nominal Vcore is. You can check it
    with the Speedfan hardware monitor output. The value
    read, doesn't correspond to the value set in the BIOS,
    as the Vcore circuit tends to "droop" under load,
    and "overshoot" about 0.050V above normal
    when lightly loaded.

    http://www.almico.com/speedfan438.exe

    Anyway, you apply a little voltage boost to Vcore.
    Say, try 0.050V more perhaps. Rerun the Prime95 test.
    Does Prime95 pass ten minutes ? OK, then you're
    back to frequency increases.

    If you plot your results, the curve begins to
    have a slope. Pretty soon, you may see a couple of
    possible shapes to the curve. I have an older
    processor here, and it "hit a wall". I could
    increase Vcore all I wanted, and I couldn't get
    one more Hertz out of it. But other processors,
    will give you an extra 400MHz for every 0.1V or
    whatever. Your job then, is to judge the effects
    this is having. If the processor runs too hot,
    then on a hot summer day, your overclock is likely
    to fail in the middle of a game. So you want
    an overclock, which can be handled well by the setup.

    There can be other limits. For example, with 45nm
    processors, you can easily apply more voltage than
    is safe. So the results may suggest a little more
    voltage would get you a bit further. But above
    a certain level, the processor may only last for
    a short time. I believe for 65nm processors, it
    isn't quite as bad, and they run out of steam
    before the Vcore gets out of hand. (As an overclocker,
    part of your research, is finding other people who
    have ruined processors, and learn from them.)

    My main purpose in describing the above, is to
    point out, that you don't just go 266 --> 400
    in one shot. Because that'll black screen for
    sure.

    Another word of warning, is to understand what
    other clocks are affected by what you do.
    For example, at one time, the PCI Express and
    SATA cable clocks, also varied with some of the
    other clock settings. You could cause some other
    subsystem in the computer to fail, while the
    overclocking experiment was in progress. In
    some cases, the hard drive was corrupted when
    Windows boots. Effectively, your copy of Windows
    gets bricked. (Moral of that story, do a backup
    *before* you try this.)

    Another alternative, and one I've used in the
    past, is to boot a Linux LiveCD. There is a version
    of Prime95 for Linux that you can get from mersenne.org.
    Since the Linux boot CD cannot get corrupted, and since
    it runs all the disk drives read-only, the odds are
    better that anything stored on the system won't be
    affected.

    Overclocking the motherboard I own, is a bit more
    difficult than yours. I have to jump from FSB800 to
    FSB1066, with no fine clock divisions. My BIOS is
    broken. I adjust my Vcore using a boost resistor
    (so the voltage takes a relatively large step as well).
    My system will accept the 33% overclock, but isn't stable
    in games. So I cannot regularly use that option. You,
    on the other hand, should have plenty of fine tuning
    settings, to pick a value of overclock which can
    survive Prime95 for four hours, with some
    3D gaming thrown in at the same time.

    There are other burn-in utilities currently making
    the rounds, but I don't keep track of them. Something
    called Linx ? And perhaps something from Intel. But the
    objective remains the same - to pick conditions which
    are good enough, that the computer never malfunctions
    in stressful usage.

    Also, until you understand the interactions of the
    controls, I recommend reviewing the results with
    CPUZ in Windows. For example, if you increase the
    CPU input clock by 5MHz, use CPUZ in Windows and
    see how much the clocks have changed. And what things
    are tied together.

    http://www.cpuid.com/cpuz.php

    One thing that may not be handled well, is the BIOS
    ability to adjust memory timings. Say, for example,
    before you turn any knobs, the memory is DDR2-800
    with 5-5-5-15 timing. You turn the knob, and later
    check the memory, and it is running DDR2-1000 with
    the same 5-5-5-15 timing. This is not right, and the
    reason it is not right, is because the timing should
    scale with the frequency. 1000/800 = 1.25. So
    the timings should probably be 7-7-7-21 (because you
    round up for safety). So be wary of what settings
    the BIOS is not correctly adjusting, while making
    your changes. If you set the memory to 7-7-7-21,
    while it was still DDR2-800, then you'd have
    "headroom" for increasing the RAM frequency.
    Or, you could drop the memory divider, such that
    the frequency doesn't rise quite as high.

    Some motherboards now, are exceedingly complicated.
    And some of the controls actually work, if you can
    find some results with all those controls listed.

    You can get a ton of results here, and I use an
    external search engine to search the forums. You
    could easily end up reading a couple thousand posts,
    to find enough adjustments for your board or chipset.

    http://www.xtremesystems.org/forums/

    On occasion, Anandtech has a survey of some of the
    settings for "strap", Northbridge voltage, termination
    voltages and the like, for a certain chipset. But generally,
    you cannot expect every chipset to have nice "recipe-like"
    scripts of stuff to do. So there is a learning process
    involved, and poor documentation to boot.

    Paul
     
    Paul, Jun 16, 2009
    #2
    1. Advertisements

  3. JD

    JD Guest

    Your response was more than I expected - thanks
     
    JD, Jun 16, 2009
    #3
    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.