Why is DRAM voltage increased to over specification?

Discussion in 'Gigabyte' started by John Doe, Feb 26, 2010.

  1. John Doe

    John Doe Guest

    My memory Kingston KVR800D2N5 should run between 1.7 and 1.9 V.
    but for some reason my GA-EP45-UD3L sets the DRAM voltage to 2.0
    V. In the BIOS, the DRAM voltage setting says "1.8 Auto" even
    though EasyTune says it is at 2 volts. So I manually changed the
    Auto entry to 1.8, and now EasyTune says the DRAM voltage is
    correctly set at 1.8. Are there any other settings that the BIOS
    will automatically set too high? That happens without any
    overclocking. It was not corrected by changing "Turbo" to
    "Standard".

    Thanks.
     
    John Doe, Feb 26, 2010
    #1
    1. Advertisements

  2. John Doe

    Paul Guest

    Motherboards do stuff like that, to try to ensure they can POST
    with a wide range of memory products. Such an approach would not
    be necessary, if the SPD on each DIMM represented conditions
    that a DDR2 DIMM would pass at JEDEC nominal 1.8V. It is the
    fact that the DIMM makers take shortcuts, that causes the
    motherboard maker to have to do stuff like that. If the
    motherboard maker doesn't do it, the result is more phone
    calls to Tech Support when a motherboard won't POST.

    The server motherboard and server DIMM industry, is more conservative,
    and you're more likely to see adherence to JEDEC conditions for
    both DIMMs and motherboards. In the desktop industry, it is a
    game of "liar's poker". Which means you have to "up the ante"
    to play. Gigabyte has decided that 2.0V is safe to use
    with any DIMM. (Your Kingston datasheet doesn't spec the
    "destruction voltage" for that DIMM. The JEDEC value of 1.8V
    is simply the recommended industry wide operating voltage.
    The 2.0V delivered, might not be doing any harm at all. We
    have no real way of knowing, since the memory industry doesn't
    toy with the idea of a "destruction voltage". They only choose
    to meet the JEDEC spec, and any room above that is gravy.)

    If you want to get technical about it, we could discuss the
    table on PDF page 23 here.

    http://download.micron.com/pdf/datasheets/dram/ddr2/1GbDDR2.pdf

    That table contains "Absolute Max" values. Vdd is limited to
    2.3 volts according to that. That could be an instantaneous failure
    value. But another operating point, might be somewhat closer to a
    "long term life" value, which might be a little bit lower than
    that. At least one semiconductor company, has some of its
    specs related to 100,000 hour operating life, so for excursions
    outside the norm, they sometimes guarantee an operating life under
    those set of conditions. In this case, of the Micron memory chip,
    it likely isn't destroyed instantly at 2.29 volts, but we don't know
    whether it will last quite as long as an identical chip operated
    at 1.8V for its entire life. At least that value of 2.3V, should
    give you some idea how much danger you were in, when the motherboard
    delivered 2.0V. A different manufacturer of memory chips, could
    have a value other than 2.3V in their table.

    *******

    As for how many voltages could be adjustable on a motherboard,
    there could be eight or nine voltages that they could adjust.
    Just to give you some idea. Thankfully, adjustable regulators
    cost more money, so not every motherboard is that adjustable.
    The fewer adjustments, the fewer mistakes they can make.

    In general, it is pretty difficult to verify whether parts
    of your motherboard are being abused by boosted voltages.
    I doubt I could find all the voltages on my motherboard, that
    claim to be adjustable, and verify them with a multimeter.
    Even if they provided a copper pad with a label next to it,
    I'd be pretty happy with that as a means to verify the
    voltages. And that wouldn't cost them any money to do.

    There was one motherboard, that had "scope points" soldered
    to the board. Which is a little metal stand, with a cup shape
    in the end of it, to make it easy to poke with a meter probe.
    That is the closest I've seen, to a motherboard designer
    caring whether a user could verify their voltages. On all the
    motherboards I have here, there isn't any hint as to what
    the various regulator circuits are connected to. Or where is
    a good place to probe them. So if you wanted to verify the
    key eight or nine voltages, well, good luck to you.

    Paul
     
    Paul, Feb 26, 2010
    #2
    1. Advertisements

  3. John Doe

    Hueyduck Guest

    John Doe a écrit :
    I wondered about that kind of behavior too.
    I had launched a thread named
    "EP35C-DS3R memory overvoltage control" on 20th of last october.

    (sorry: I don't know how to create a direct link towards a usenet article).

    For the record, in my case, I had to learn that the default setting of
    the motherboard created some errors with memeory (I had to redo the
    memtest to see that). After I lowered the voltage to the memory brand
    recommendation, there was no more errors. Not that I know.

    The EP35C-DS3R is the firt Gigavyte MB I bought, and it is also the
    first time I see such a picky attitude towards memory sticks (I had to
    change the first sticks I had bought, since they produced an amount of
    errors so high out iof incompatibility).
    Frankly, this is not a good point in favor of Gigabyte. But you have yo
    chose a MB, right?

    Huey
     
    Hueyduck, Mar 21, 2010
    #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.