Is higher DDR speed than FSB speed ok?

Discussion in 'Gigabyte' started by Don, Nov 23, 2008.

  1. Don

    Don Guest

    My motherboard has a 533 mhz fsb bus and the
    recommended DDR speed is 333 MHz (PC2700).
    Will I lose on performance or damage anything if
    I use a DDR 400 MHz (PC3200)?
     
    Don, Nov 23, 2008
    #1
    1. Advertisements

  2. Don

    Paul Guest

    It depends on the motherboard and its chipset, as
    to what speed it might offer. If you're telling me,
    that the chipset is limited to DDR333, then a
    DDR400 stick can run at 400, 333, 266, and 200.
    In other words, a PC3200 is a "universal donor".
    It is backward compatible. If the divider choices
    in the chipset are limited, the BIOS will run
    it at 333. Problem solved. It means the PC3200
    stick runs exactly as if it was a PC2700 stick.

    For further info on your options, I need

    1) Motherboard make and model number
    2) Number of sticks you plan to run
    or the total amount of memory you're
    shooting for.
    3) If you have existing memory, and are
    seeking to add some more memory, indicate
    what is currently installed. I may end
    up recommending you throw some of it
    away etc.

    HTH,
    Paul
     
    Paul, Nov 23, 2008
    #2
    1. Advertisements

  3. Don

    Don Guest

    Hi Paul. Does this mean that one can plug the
    highest speed DDR card disregarding the mobo
    FSB without taking a hit in performance (ie- the
    atual computer working speed)?
    For some strange reason higher speed cards are
    now sometimes available at a lower street price
    than lower speed ones.
    thanks .... :)
     
    Don, Nov 24, 2008
    #3
  4. Don

    Paul Guest

    When you run a stick at a different speed, the
    bandwidth changes, but the latency stays more
    or less the same.

    Take some CAS3 at DDR400. DDR400 has an input
    clock of 200MHz, and the clock period would be
    5ns. CAS3 times 5 nanoseconds is 15 nanoseconds.
    CAS is the delay for the delivery of the first
    data, when it is asked for.

    When you operate the stick at DDR333 in that case,
    you can use a latency number the same as the
    original spec. That means you still have a 15
    nanosecond limitation on delivery. DDR333 means an
    input clock of 166MHz, and the inverse of that, the
    clock period, is 6 nanoseconds. By happy
    coincidence, if we divide 15/6, we get 2.5
    cycles. So we could use a CAS latency setting
    as tight as 2.5 ticks. That doesn't mean
    the latency has changed, because the delay until
    the data is ready, is still 2.5*6=15ns, so
    nothing has changed in real terms.

    You can use CPUZ, and the SPD tab in the
    screen display, to examine the SPD table
    stored on the DIMM. And if the SPD contains
    timing info for more than one clock speed,
    you should be able to see the CAS value
    change with frequency. (I.e. Attempting
    to keep the latency a constant, getting
    no worse.)

    Operating at DDR333 means less theoretical
    bandwidth from the stick. So that is a
    tangible result. If we were running WinRAR
    and compressing some files, it would make
    a detectable difference. For many other
    applications, if wouldn't matter as much.

    I use a rough "1/3rd" rule for performance.
    Just to make up a number, if I divide 333/400,
    that is about 83%, or 17% less bandwidth. The 1/3rd
    rule says the compute performance changes by
    17%/3 = about 5 to 6 percent or so. Now, if
    the memory subsystem happened to be offering
    more bandwidth than the FSB, then in fact
    the difference might be different than that.
    The 1/3rd rule applies to real measured
    bandwidth numbers, so you'd have to benchmark
    the two cases, and get the real bandwidth
    numbers. (I use the memtest86+ screen for that,
    not because of a superior test method, but because
    it is consistent from test to test and quick
    to execute.) So saying those numbers were
    roughly right, then the stick running at DDR333
    might cause the computer to lose 5% performance,
    if the memory was the bottleneck. That little
    rule is a guesstimate, and is only roughly true
    for small bandwidth deviations. It the bandwidth
    is cut in half, you might notice the guess
    at the impact isn't so good. The guess applies
    to an average program - a program like Photoshop
    might see more of a loss, than other programs.
    WinRAR and Photoshop should be a couple of the
    programs that are more sensitive to memory
    performance.

    I'm sure an economist could tell you why the
    slower ones are more expensive, but I'm not
    going to try. I didn't get a very good mark in
    my Economics 100 course :)

    HTH,
    Paul
     
    Paul, Nov 25, 2008
    #4
  5. Don

    geoff Guest

    If you're telling me,
    Exactly, it is only a rating or spec. The question is like asking, my car
    can go 120 mph, is it ok if I drive it 70 mph on the highway?

    --g
     
    geoff, Nov 27, 2008
    #5
    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.