1. This forum section is a read-only archive which contains old newsgroup posts. If you wish to post a query, please do so in one of our main forum sections (here). This way you will get a faster, better response from the members on Motherboard Point.

E4300 / 680i overclocking

Discussion in 'Overclocking' started by Phil Weldon, Apr 17, 2007.

  1. Phil Weldon

    Phil Weldon Guest

    With this platform

    Windows XP Pro SP2
    Intel E4300
    EVGA 680i SLI
    EVGA 8800 XTS 320 MByte
    2 GByte PC 8500 memory
    Seagate 500 GByte SATA II hard drive

    I am now using Orthos CPU Burn-In, recommended by 'Fishface'.
    nTune does not work at all for me, even after updating nTune and flashing
    the system BIOS to P27.

    Some numbers using Orthos CPU Burn-In (small FFTs, both CPUs), ten hour run:
    Case side open, no case fans

    CPU: 2.4 GHz
    FSB: 1066 MHz
    Memory: 1066 MHz
    CPU voltage: 1.325 VDC
    Memory voltage: 1.8 VDC
    CPU temperature: 54 C
    System temperature: 39 C
    GPU temperature: 54 C

    *****
    (as above, but 10 minute Orthos run)
    CPU: 2.7 GHz
    FSB: 1200 MHz
    Memory: 1200 MHz
    CPU voltage: 1.4000 VDC
    Memory 1.8 VDC
    CPU temperature: 62 C
    System temperature: 39C

    *****
    (10 minute Orthos run)
    CPU: 2.85 GHz
    FSB: 1267 MHz
    Memory: 1056 MHz
    CPU voltage: 1.325 VDC
    Memory voltage: 1.8 VDC
    CPU temperature: 63 C
    System temperature: 38 C
    GPU temperature: 53 C

    *****
    (ten minute Orthos run)
    CPU 3.0 GHz
    CPU voltage: 1.325 VDC
    ### spontaneous reboot as Windows loads

    *****
    CPU: 3.0 GHz
    CPU voltage: 1.350 VDC

    ### Orthos halts on error

    *****
    CPU: 3.0 GHz
    CPU voltage: 1.375 VDC
    ten minute Orthos run
    Ambient temperature: 23 C
    CPU temperature: 69 C
    System temperature: 39 C
    GPU temperature: 53 C

    *****
    CPU: 3.150 GHz
    FSB: 1400 MHz

    CPU voltage: 1.375 VDC
    ### spontaneous reboot as Windows loads

    CPU voltage: 1.3875 VDC
    ### spontaneous reboot as Windows loads

    CPU voltage: 1.4000 VDC
    ### Orthos halts on error

    CPU voltage: 1.4125
    ### Orthos halts on error

    CPU voltage: 1.4250
    ### Orthos halts on error

    CPU voltage: 1.4325
    ### Orthos halts on error

    *****
    Now, for the sweet spot, so far:

    CPU: 2.70 GHz (50% overclock)
    FSB: 1200 MHz
    Memory: 1200 MHz
    CPU voltage: 1.2500 VDC
    Memory voltage: 2.3 VDC
    Ambient temperature: 23 C

    (one hour Orthos small FFTs, both CPUs)
    (Task Manager reports both CPUs at 100%)
    CPU temperature: 57 C
    System temperature: 38 C
    GPU temperature: 53 C

    (one hour nTune Stability test [CPU, Memory, PCI-E, GPU selected])
    (Task Manager reports CPU 0 at 100%, CPU 1 at 100% [80% of the time])
    CPU temperature: 42 C
    System temperature: 38 C
    GPU temperature: 61 C

    Time for another pass through the manual, and another search through Google
    about problems with nTune.
    (nTune allows the adjustment of parameter LOWER than the settings in the
    BIOS, but not to HIGHER settings / automatic tuning does not work at all,
    even after updating nTune and flashing the BIOS to P27)

    Phil Weldon
     
    Phil Weldon, Apr 17, 2007
    #1
    1. Advertisements

  2. Phil Weldon

    Ed Medlin Guest

    Good info, thanks. I will probably not use nTune as it kind of reminds me of
    the Asus AI software that just wouldn't work for me. Are you going to try
    watercooling at some time? Just looking at some of your results that just
    might make a difference in getting to 3+ghz. If you use the Peltier too, who
    knows? About all my 'stuff' shipped yesterday (Mon) and the rest today and
    tomorrow and now I am antsy....:). My major splurge on this build
    (something I wanted but didn't really need) is the Samsung S203-2408 24"
    monitor......:). That was my gift from Mama.......:). She is almost as
    into this build as I am, which is a really rare thing indeed. She will get
    the old system.....:).

    Ed
     
    Ed Medlin, Apr 17, 2007
    #2
    1. Advertisements

  3. Phil Weldon

    Jack R Guest

    Phil, good info, thanks!
    Question, what maximum CPU temp under stress and at idle do you find
    acceptable?
    (BTW, I'm a hardware engineer, with comparable experience, in time, as
    you...starting with Fortran and punch cards)
    Keep up the good work,
    Jack R

     
    Jack R, Apr 17, 2007
    #3
  4. Phil Weldon

    Phil Weldon Guest

    'Jack R' wrote, in part:
    | Phil, good info, thanks!
    | Question, what maximum CPU temp under stress and at idle do you find
    | acceptable?
    | (BTW, I'm a hardware engineer, with comparable experience, in time, as
    | you...starting with Fortran and punch cards)
    | Keep up the good work,
    _____

    Question: "what Maximum CPU temp under stress and at idle do you find
    acceptable?"

    Short answer: Anything that works!
    Longer answer:
    The idle temperature is useful as a diagnostic to check the cooling system,
    but otherwise has no impact. I consider that when you overclock Intel CPUs
    you usually get a generous overclock out of the box. When you start pushing
    past that, then reducing the operating temperature and raising the CPU core
    voltage trade the built-in head room of those two parameters to gain more
    operating frequency headroom. My grasp of the trade off is that higher
    voltages help the transistors switch more quickly (and improve the rise and
    fall times.) Reducing the operating temperature reduces resistance
    (slightly in the room temperature range) and helps in the same way raising
    the core voltage does. Perhaps you can expand on or correct my
    understanding.

    What I LIKE to see is an operating temperature in the low 50s C for
    conventional cooling when the CPU is under maximum stress.

    * Intel CPUs have an on-CPU-die thermal diode for reading the CPU
    temperature, as well as a diode that will halt operation and let the CPU
    cool off (this diode will internally halt operation and allow the CPU to
    cool off; it also outputs a signal 'THERMTRIP#'. In six or seven years of
    participating in this newsgroup I've never seen evidence that any Intel
    Pentium class CPU has been destroyed by heat. (The Pentium 60 and 66 may
    not have had this protection.)

    One thing I've done to evaluate the value of extra cooling is inexpensive
    and easy to do:
    (a.) buy a can of 'Freeze-down' component cooler, ~ $6 US at RadioShack
    (any of several different brands will do, a large can is better)
    (b.) remove the CPU heatsink
    (c.) begin spraying the component cooler on the bare CPU, continue
    spraying throughout the test [a second person is a VERY big help in this]
    (d.) boot up and set higher numbers in the BIOS, then load the OS and
    run at moderate CPU stress, then maximum.
    (e.) repeat until you hit a wall
    The best speed you get is what you can expect with sub-zero Celsius CPU
    cooling.

    Component cooler is also a good way to find the location of the System
    thermistor on the motherboard. While the system is operating, keep an eye
    on the System temperature as you spray small areas of the motherboard
    (perhaps 1.5 inches by 1.5 inches [5 X 5 cm]). The System temperature will
    drop quickly and sharply when you are in the right neighborhood. Then
    concentrate on a smaller and smaller areas until you can pinpoint the
    thermistor (it may be at a silk-screened notation like 'TH 1'.

    Phil Weldon

    | Phil, good info, thanks!
    | Question, what maximum CPU temp under stress and at idle do you find
    | acceptable?
    | (BTW, I'm a hardware engineer, with comparable experience, in time, as
    | you...starting with Fortran and punch cards)
    | Keep up the good work,
    | Jack R
    |
    | | > With this platform
    ..
    ..
     
    Phil Weldon, Apr 17, 2007
    #4
  5. Phil Weldon

    Jack R Guest

    Phil, thanks for your detailed reply.
    I think you are accurate with your description, although I'm no expert. I
    think raising the voltage also improves the signal-to-noise ratio. reducing
    data errors.
    Most devices are rated (commercially) at 70C, but the silicon can run at
    over 100C. The packaging thermal path is the limiting factor. I agree that
    staying in the 50'sC is conservative. I've known people to run high 60's +.
    Your trick of finding the MB sensor is a good one!
    Thanks again,
    Jack R

     
    Jack R, Apr 17, 2007
    #5
  6. Hmm... I got a little confused where you say your DDR2 1066 memory is
    running at 1200. Wouldn't you want to increase your divider to clock it
    down closer to 1066 where it won't be a bottleneck? Pardon my ignorance,
    but I'm new to C2D OCing having been with AMD previously.
     
    Phil, Non-Squid, Apr 17, 2007
    #6
  7. Phil Weldon

    Phil Weldon Guest

    _____

    The Memory clock can be locked to be independent from the CPU clock.
    Running the Memory bus at 1200 MHz and the FrontSide Bus at 1200 MHz just
    means that the two are 'locked' together. Memory at 1200 MHz runs faster
    than at 1066 MHz, so the system has better performance.

    Memory becomes a bottle neck only when its bus must be lower than the CPU
    bus. However, when the memory bus speed can be raised no more, then it
    should be 'unlocked' from the frontside bus (a CPU : memory ratio that is
    more than 1:1, say 5:4.)

    A system running with an FSB of 1333 MHz and a memory bus of 1333 MHz has
    better performance than a system with an FSB of 1333 MHz and a memory bus
    speed of 1066 MHz. What you want is the highest possible stable speed for
    EVERYTHING. But CPU speed is more important than the memory speed,
    especially with larger L2 caches.

    Phil Weldon

     
    Phil Weldon, Apr 17, 2007
    #7
  8. I think I was unclear. Is your memory certified for 1200MHz speeds even
    when marked as 1066MHz?
     
    Phil, Non-Squid, Apr 17, 2007
    #8
  9. Phil Weldon

    Thomas Guest

    His memory is certified for 800 'MHz' DDR, which would be 1600 'MHz'
    'FSB'... So Phil still has a lot of headroom left :eek:)
     
    Thomas, Apr 17, 2007
    #9
  10. Phil Weldon

    Phil Weldon Guest

    'Thomas' wrote, in reply to 'Phil, Non-Squid':
    | > I think I was unclear. Is your memory certified for 1200MHz speeds
    | > even when marked as 1066MHz?
    |
    | His memory is certified for 800 'MHz' DDR, which would be 1600 'MHz'
    | 'FSB'... So Phil still has a lot of headroom left :eek:)

    My memory is certified for PC8500, which is 8500 / 2 / 4 = ~ PC1066 or a
    memory bus speed of 1066 MHz. It is 'SLI ready memory' which means the SPD
    contains extended settings that are used by nVidia chipset motherboards. In
    the case of my Patriot memory the settings include a memory voltage of 2.3
    VDC when the memory is operated at 1066 MHz bus speed or higher. Since the
    extended specs set a memory voltage that is already 0.5 volts HIGHER than
    the memory chip manufacturer specifies, I doubt there is much more headroom,
    certainly I don't plan to raise the memory voltage above 2.3 volts.

    PC800 DDR2 memory is 800 MHz X 2 (for DDR2) X 4 (for quad pumped) = PC6400
    memory = 200 MHz memory clock.
    PC1066 DDR2 memory is 1066 MHz X 2 (for DDR2) X 4 (for quad pumped) = PC8500
    memory = 266 MHz memory clock.
    For quad pumped memory, the bus speed is 4 X the clock speed, 4 data
    transfers per clock cycle (four clock phases are used, each phase 90 degrees
    apart.) Dual data rate means 2 transfers for each of 4 clock phases. Thus
    the 4 X 2.

    Phil Weldon


    | Phil, Non-Squid wrote:
    | > > > Hmm... I got a little confused where you say your DDR2 1066
    | > > > memory is running at 1200. Wouldn't you want to increase your
    | > > > divider to clock it down closer to 1066 where it won't be a
    | > > > bottleneck? Pardon my ignorance, but I'm new to C2D OCing having
    | > > > been with AMD previously. --
    | > > > Phil
    | >
    | > I think I was unclear. Is your memory certified for 1200MHz speeds
    | > even when marked as 1066MHz?
    |
    | His memory is certified for 800 'MHz' DDR, which would be 1600 'MHz'
    | 'FSB'... So Phil still has a lot of headroom left :eek:)
    |
    | --
    | Met vriendelijke groeten, Thomas vd Horst.
    |
    |
     
    Phil Weldon, Apr 17, 2007
    #10
  11. Phil Weldon

    Phil Weldon Guest

    'Phil, Non-Squid' wrote:
    | I think I was unclear. Is your memory certified for 1200MHz speeds even
    | when marked as 1066MHz?

    'Certified' for is not the same as 'capable'. The E4300 I use is
    'certified' for 1.8 GHz, it is 'capable' of much more. My memory is
    'certified' for 1066 MHz. At a 1200 MHz memory bus speed, my PC1066 /
    PC8500 memory is overclocked by ~ 13%. With 'SLI-ready' memory, the
    unlocked multiplier EX6800, and nVidia chipsets the whole idea of
    'certified' has gotten vague. See my reply to 'Thomas' for more
    information.

    By the way, does 'Phil, Non-Squid' mean that you are Navy, but not a
    submariner?

    Phil Weldon

    | Phil Weldon wrote:
    | > 'Phil, Non-Squid' wrote:
    | >> Hmm... I got a little confused where you say your DDR2 1066 memory is
    | >> running at 1200. Wouldn't you want to increase your divider to
    | >> clock it down closer to 1066 where it won't be a bottleneck? Pardon
    | >> my ignorance, but I'm new to C2D OCing having been with AMD
    | >> previously.
    | > _____
    | >
    | > The Memory clock can be locked to be independent from the CPU clock.
    | > Running the Memory bus at 1200 MHz and the FrontSide Bus at 1200 MHz
    | > just means that the two are 'locked' together. Memory at 1200 MHz
    | > runs faster than at 1066 MHz, so the system has better performance.
    | >
    | > Memory becomes a bottle neck only when its bus must be lower than the
    | > CPU bus. However, when the memory bus speed can be raised no more,
    | > then it should be 'unlocked' from the frontside bus (a CPU : memory
    | > ratio that is more than 1:1, say 5:4.)
    | >
    | > A system running with an FSB of 1333 MHz and a memory bus of 1333 MHz
    | > has better performance than a system with an FSB of 1333 MHz and a
    | > memory bus speed of 1066 MHz. What you want is the highest possible
    | > stable speed for EVERYTHING. But CPU speed is more important than
    | > the memory speed, especially with larger L2 caches.
    | >
    | > Phil Weldon
    | >
    | > | >> Phil Weldon wrote:
    | >>> With this platform
    | >>>
    | >>> Windows XP Pro SP2
    | >>> Intel E4300
    | >>> EVGA 680i SLI
    | >>> EVGA 8800 XTS 320 MByte
    | >>> 2 GByte PC 8500 memory
    | >>> Seagate 500 GByte SATA II hard drive
    | >>>
    | >>> I am now using Orthos CPU Burn-In, recommended by 'Fishface'.
    | >>> nTune does not work at all for me, even after updating nTune and
    | >>> flashing the system BIOS to P27.
    | >>>
    | >>> Some numbers using Orthos CPU Burn-In (small FFTs, both CPUs), ten
    | >>> hour run: Case side open, no case fans
    | >>>
    | >>> CPU: 2.4 GHz
    | >>> FSB: 1066 MHz
    | >>> Memory: 1066 MHz
    | >>> CPU voltage: 1.325 VDC
    | >>> Memory voltage: 1.8 VDC
    | >>> CPU temperature: 54 C
    | >>> System temperature: 39 C
    | >>> GPU temperature: 54 C
    | >>>
    | >>> *****
    | >>> (as above, but 10 minute Orthos run)
    | >>> CPU: 2.7 GHz
    | >>> FSB: 1200 MHz
    | >>> Memory: 1200 MHz
    | >>> CPU voltage: 1.4000 VDC
    | >>> Memory 1.8 VDC
    | >>> CPU temperature: 62 C
    | >>> System temperature: 39C
    | >>>
    | >>> *****
    | >>> (10 minute Orthos run)
    | >>> CPU: 2.85 GHz
    | >>> FSB: 1267 MHz
    | >>> Memory: 1056 MHz
    | >>> CPU voltage: 1.325 VDC
    | >>> Memory voltage: 1.8 VDC
    | >>> CPU temperature: 63 C
    | >>> System temperature: 38 C
    | >>> GPU temperature: 53 C
    | >>>
    | >>> *****
    | >>> (ten minute Orthos run)
    | >>> CPU 3.0 GHz
    | >>> CPU voltage: 1.325 VDC
    | >>> ### spontaneous reboot as Windows loads
    | >>>
    | >>> *****
    | >>> CPU: 3.0 GHz
    | >>> CPU voltage: 1.350 VDC
    | >>>
    | >>> ### Orthos halts on error
    | >>>
    | >>> *****
    | >>> CPU: 3.0 GHz
    | >>> CPU voltage: 1.375 VDC
    | >>> ten minute Orthos run
    | >>> Ambient temperature: 23 C
    | >>> CPU temperature: 69 C
    | >>> System temperature: 39 C
    | >>> GPU temperature: 53 C
    | >>>
    | >>> *****
    | >>> CPU: 3.150 GHz
    | >>> FSB: 1400 MHz
    | >>>
    | >>> CPU voltage: 1.375 VDC
    | >>> ### spontaneous reboot as Windows loads
    | >>>
    | >>> CPU voltage: 1.3875 VDC
    | >>> ### spontaneous reboot as Windows loads
    | >>>
    | >>> CPU voltage: 1.4000 VDC
    | >>> ### Orthos halts on error
    | >>>
    | >>> CPU voltage: 1.4125
    | >>> ### Orthos halts on error
    | >>>
    | >>> CPU voltage: 1.4250
    | >>> ### Orthos halts on error
    | >>>
    | >>> CPU voltage: 1.4325
    | >>> ### Orthos halts on error
    | >>>
    | >>> *****
    | >>> Now, for the sweet spot, so far:
    | >>>
    | >>> CPU: 2.70 GHz (50% overclock)
    | >>> FSB: 1200 MHz
    | >>> Memory: 1200 MHz
    | >>> CPU voltage: 1.2500 VDC
    | >>> Memory voltage: 2.3 VDC
    | >>> Ambient temperature: 23 C
    | >>>
    | >>> (one hour Orthos small FFTs, both CPUs)
    | >>> (Task Manager reports both CPUs at 100%)
    | >>> CPU temperature: 57 C
    | >>> System temperature: 38 C
    | >>> GPU temperature: 53 C
    | >>>
    | >>> (one hour nTune Stability test [CPU, Memory, PCI-E, GPU selected])
    | >>> (Task Manager reports CPU 0 at 100%, CPU 1 at 100% [80% of the
    | >>> time]) CPU temperature: 42 C
    | >>> System temperature: 38 C
    | >>> GPU temperature: 61 C
    | >>>
    | >>> Time for another pass through the manual, and another search through
    | >>> Google about problems with nTune.
    | >>> (nTune allows the adjustment of parameter LOWER than the settings in
    | >>> the BIOS, but not to HIGHER settings / automatic tuning does not
    | >>> work at all, even after updating nTune and flashing the BIOS to P27)
    | >>>
    | >>> Phil Weldon
    | >>
    | >> Hmm... I got a little confused where you say your DDR2 1066 memory is
    | >> running at 1200. Wouldn't you want to increase your divider to
    | >> clock it down closer to 1066 where it won't be a bottleneck? Pardon
    | >> my ignorance, but I'm new to C2D OCing having been with AMD
    | >> previously. --
    | >> Phil
    |
    | I think I was unclear. Is your memory certified for 1200MHz speeds even
    | when marked as 1066MHz?
    | --
    | Phil
    |
    |
     
    Phil Weldon, Apr 17, 2007
    #11
  12. Man... I was used to old-school 2xFSB with my Athlon XP... thanks for the
    clarification
     
    Phil, Non-Squid, Apr 18, 2007
    #12
  13. Do you think you could go higher on the CPU by dialing down the memory more?
    Others seem to be getting up to around 3.2 on air cooling.
    Not quite. Look up the motorcycle definition. I don't have any connections
    to the Navy.
     
    Phil, Non-Squid, Apr 18, 2007
    #13
  14. Phil Weldon

    Fishface Guest

    Fishface, Apr 18, 2007
    #14
  15. Phil Weldon

    Ed Medlin Guest

    Ahaaa.....another old fart...........:) Basic-8, Cobal and Fortran........
    I actually worked more with hardware with the old RCA Data Systems Service
    from '71-early 80s until the buyout from GE. I was already deep in the real
    estate by then so just went that route but kept up with everything as much
    as I could. Degree was in Computer Science/Accounting.

    Ed
     
    Ed Medlin, Apr 18, 2007
    #15
  16. Phil Weldon

    Ed Medlin Guest

    Ed Medlin, Apr 18, 2007
    #16
  17. Phil Weldon

    Ed Medlin Guest

    I went with the (2x1ghz) Mushkin eXtreme Perfomance DDR2 1066/PC8500 rated
    at 2.2-2.35v. I also would not go above manufacturer's rated voltages on the
    memory. All I have read about it says it can do 1200mhz ok in most cases and
    sometimes even a bit higher. That may be due to the 2.35vdc rating. Oh
    well.....it isn't here yet so I really don't know as of now......:-(.

    Ed
     
    Ed Medlin, Apr 18, 2007
    #17
  18. Phil Weldon

    Fishface Guest

    The $300 pair? I guess I'm the only one here overclocking for value.
    All right, that's it. I'm going to have to pull two sticks out, lower my
    timings and push my $100 pair for all they're worth!
     
    Fishface, Apr 18, 2007
    #18
  19. Phil Weldon

    Fishface Guest

    Yeah, I'm confused too. On my Biostar 965PT with i965 chipset and
    Phoenix AwardBios, you set the the "System Memory Frequency" to 533
    (essentially this controls the divider) and raise the "CPU Clock" speed
    from 266 to overclock synchronously. The Multiplier is adjustable through
    the "CPU Clock Ratio" setting. When raising the "CPU Clock" speed,
    there is a non-adjustable number which displays "System Memory
    Frequency." With my "CPU Clock Ratio" set to 7, my "CPU Clock"
    speed to 440, with "System Memory Frequency" at 533 (1:1), CPU-Z
    shows:

    –CPU Tab–

    Core Speed
    3077.5 MHz

    Multiplier
    x7

    Bus Speed
    439.7 MHz

    Rated FSB
    1758.6

    –Memory Tab–

    Frequency
    439.7

    FSB:DRAM
    1:1

    So, if I had to guess from Phil's "FSB 1200," "Memory 1200" and 2.7 GHz, I
    would guess "CPU clock" equivalent is 300 MHz, given the default multiplier
    of 9. But then again, he *is* feeding the memory a lot of voltage...
     
    Fishface, Apr 18, 2007
    #19
  20. I bought 2GB Corsair 800 for $116 from ZZF. So you're not alone...
     
    Phil, Non-Squid, Apr 18, 2007
    #20
    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.