Question for Paul

Discussion in 'PC Hardware' started by Astropher, Jan 18, 2015.

  1. Astropher

    Astropher Guest

    I have a Gigabye X58A-UD3R motherboard, i7-950 based system that is
    showing signs of what appears to me to be a power supply fault.

    If I power the system down for several hours AND switch off the AC to
    the supply, when I attempt to power it back up again, pressing the
    power button does nothing and then after some period of time (perhaps
    20 minutes or more) the system powers up and runs normally. When it
    is in the dead state, there are no LEDs illuminated on the motherboard
    (except for the blue 'clear CMOS' button), the fan doesn't attempt to

    I first noticed the problem a couple of months ago, and at that time
    the power up delay was only a minute or so. It has progressively
    reached the current state.

    When the system is running, I can reboot it, power it down and
    immediately power it up again with no problems. The fault only occurs
    when the system has been powered down for some time (with the AC input
    switched off).

    The power supply is an ANTEC VP650 V2 and is just on 12 months old.
    The previous power supply (a Vantec ION 520W job) died after 3 years of

    My question is: Do you think this is a power supply issue? The machine
    runs flawlessly otherwise. if it is a power supply issue, what brand
    of supply is good these days. I bought the VANTEC because it was
    supposed to be a quality power supply.
    Astropher, Jan 18, 2015
    1. Advertisements

  2. Astropher

    Paul Guest

    You can test the power supply separately.

    The PS_ON# wire, if you connect it to GND, the power supply should
    start. You would hear the power supply fan spinning.

    To do such a test, you would want to disconnect the supply from the
    system, put some electrical load on it (a couple old hard drives on
    the Molex cables will do). Then switch on at the back of the power
    supply, then connect PS_ON# to GND. The PSU fan should spin. If
    you have a multimeter, you can make a quick check of the output

    The power supply should respond instantly, each time you ground PS_ON#.
    There should not be any 20 minute delay in such a test. This is to
    prove the power supply is not taking a 20 minute nap.


    I don't know the reason for your symptoms. I've heard of these
    symptoms before, but could offer no concrete suggestions for
    dealing with them. The problem could be with the motherboard,
    somewhere in the PS_ON# logic, maybe where a capacitor
    is involved in RESET timing or the like. Some component which
    isn't charging up in the expected time.

    The power supply, has two supply sections. The +5VSB (the one
    that is lighting your blue LED right now), has its own switching
    supply. The main rails (3.3V/5V/12V) involved a bigger chunk of stuff.
    The motherboard cannot "start" anything, unless +5VSB is running.
    The fact a LED lights up, tells me +5VSB is present. This is why
    I'm guessing this is not primarily a power supply problem. If you
    had no LEDs glowing at all, I would be less certain, and then
    doing a power supply test (both sections) would be necessary.

    But if the +5VSB section is able to light the LED, that suggests
    a motherboard problem. First the +5VSB comes on, then the motherboard
    logic drives PS_ON#, then the main supply rails are turned on and
    the fans spin. The motherboard is holding up the works.

    These problems can be debugged to the "nearest replaceable
    assembly" with a multimeter. But in terms of the logic on
    the motherboard, some of the schematics, I can't even follow
    what they're doing in there. The logic snakes all over the place,
    and it's not always arranged in an order that points to the
    organization behind it. Motherboards have things like
    "backfeed cut", which disconnects power from certain
    things (when the system is off), to prevent the wrong
    part of the motherboard from getting charged up. It could
    even be a failing backfeed, which allows +5VSB to get into
    a part of the circuit where normally +5V runs. When that
    happens, the motherboard "can't tell it is turned off". So
    the logic state is inconsistent. There is no reason
    to generate a power-on sequence, if you think you're
    already on.

    On one system, the back feeding path turned out to be the
    computer monitor. If the computer monitor was disconnected,
    pushing the front power button would bring up the computer instantly.
    Then, the user could plug in their VGA connector again
    and start work. The VGA cable at one time had a +5V pin,
    and some monitors delivered power to that pin. And it had
    something to do with something power related, flowing back
    towards the motherboard. Now, if that was my system,
    I'd have made a VGA to VGA extension cable, and removed
    the +5V pin at one end. I think a lot of modern VGA
    cables are missing that pin (so making such a cable
    is no longer necessary), so this really shouldn't
    be possible any more. But again, I mention this, so if
    you're using VGA, you can unplug the monitor and see
    if the system starts (cold) faster than 20 minutes time.

    Paul, Jan 19, 2015
    1. Advertisements

  3. Astropher

    VanguardLH Guest

    Seems a CMOS battery replacement is a cheap diagnosis step.
    VanguardLH, Jan 19, 2015
  4. Astropher

    Paul Guest

    It's a possibility. But not all motherboards are stopped
    cold by a completely flat CMOS battery. Only some designs
    seem to be susceptible. I have systems here, with flat
    batteries (my in-storage collection), that still start.

    The scenario for this would be:

    1) Battery gets loaded by the unplugged interval the
    machine experiences when not in usage.

    2) When the power comes back, +5VSB through the 3VSB rail,
    charges the battery through the ORing diode (reverse
    leakage current of less than 1 microamp). The battery
    is not supposed to charge through the diodes, and the
    diodes are there to prevent charging. The diode spec
    should be less than 1 microamp leakage at max temperature,
    in this application (the max current allowed to flow into a
    CR2032 non-rechargeable battery).

    3) Once battery terminal isn't exactly at 0.0V any more,
    system starts. This only applies to machines where the
    thing is killed by a completely flat battery. And I don't
    see a pattern as to which systems are vulnerable and which

    This flat battery thing could be a feature of the SuperI/O, but
    I've not see a comment about such a response in a SuperI/O datasheet,
    and I have a collection of around a half-dozen of those.

    Paul, Jan 19, 2015
  5. Astropher

    VanguardLH Guest

    It was the OP saying the computer wouldn't power up for awhile by BOTH
    powering down (presumably using the Power switch) AND switching off the
    A/C to the supply (by a switch on the back of the PSU or a power strip
    into which the PSU was connected).

    When an ATX computer is powered down using the Power switch, the PSU
    still provides +5VSB to the motherboard. This is used to also provide
    the +3V to the RTC chip to keep it alive while the computer is "powered
    down" (but the PSU is still alive by providing +5VSB if the A/C to the
    PSU is still live). There'll be no +5VSB hence no +3V to RTC is the A/C
    power is disconnected from the PSU.

    With a dead battery, the defaults from the EEPROM copy of the BIOS
    settings has to get copied into the CMOS RAM inside the RTC. I'm not
    sure this qualifies as a true reset of the CMOS table to ensure all BIOS
    defaults gets copied into CMOS. As I recall but from old RTC designs,
    there was a capacitor used to maintain the +3V needed to retain the
    settings in the CMOS RAM. This was to allow the user time to remove a
    CMOS battery while the PSU had no A/C to put in a replacement coin cell
    battery without the CMOS getting [partially] cleared.

    The user said it took a long time after disconnecting the A/C source
    before the problem cropped up. Well, I've had old systems that only
    gave me 18 minutes and up to 22 hours after removing the CMOS battery
    before the CMOS was unable to retain [some of] the settings. With a
    reset after putting in a new CMOS battery, I would consider the state of
    the CMOS table to be corrupt so who knows what effect it may have.

    So I figure for the cost of a CR2023 coin cell battery, remove the old
    battery, put in the new battery, and short the CMOS reset jumper on the
    mobo that this could be eliminated as the source of the problem. If the
    user had customized the CMOS copy of the BIOS settings or selected a
    different settings scheme (e.g., optimal, overclocked, etc) then he
    would have to reconfigure those settings after resetting the CMOS table
    or, at least, in the BIOS settings have it force a copy of BIOS settings
    into CMOS.
    VanguardLH, Jan 19, 2015
  6. Astropher

    Paul Guest

    You only replace the CR2032 with the power off. So there is no
    power on the system at that point. You could use the CLR_RTC jumper
    to drain any remaining current. (The reason for unplugging, is the battery
    casing is a conductor all over, and it tends to hop out of the socket while
    you're working on it, and it could get into all sorts of mischief if it
    shorts the wrong terminals together. Better to just power everything
    off before beginning a battery swap.)

    The capacitor just after the ORing diode, where it meets the
    three volt input on the Southbridge, that capacitor is there
    for noise reasons. It's a bypass capacitor. On the two reference
    schematics from Intel pictured here, one uses a 0.1uF and the
    other uses a 1.0uF. Those aren't exactly large capacitors.
    Motherboard makers use 100uF for USB power hold-up (inrush
    buffering), so there are larger caps they could use
    if they wanted. There isn't a conscious effort here to
    hold up the battery terminal while the battery is out
    of the socket. It really depends on how fast you are at
    changing the battery, and that battery flops around like a fish
    when you're working on one of those.

    The CMOS is protected by a crude checksum, and
    the BIOS will re-initialize the CMOS if the checksum
    is wrong. But it's not a particularly strong scheme.

    And I don't know if there are any "CMOS well" register bits,
    that record that the chip has come up from a dead state.
    That would be a better way to do automated reinitialization.

    The BIOS exit menu has a "restore defaults" option,
    if you want to use it upon entering the BIOS. That's
    a quick way to request the CMOS memory bits be reloaded.

    The OP has been turning his PC off at the back. And that
    gives an opportunity for a battery on its last legs
    to be wobbling around the point at which the motherboard
    might not be able to start. The evidence to date, is
    motherboards that don't start, the battery seems to be
    at 0.0V (completely flat). If there is any sort of
    voltage left on it, the board still starts. It only
    takes three weeks of discharging the battery by the CMOS,
    to "walk down" the knee of the battery discharge curve and
    hit 0.0V. So if your theory is correct, the delay at
    startup a month from now, will change from 20 minutes to
    infinity :) Because then the battery will be really flat.

    One other thing to note about that picture above. There
    is no monitoring connection that comes straight from the
    battery. I have a suspicion that some designs (not Intel),
    they pick up a connection right from the battery and run
    it over to the SuperIO. And that might be the difference
    between designs that don't start, and ones that do.

    Paul, Jan 19, 2015
  7. Astropher

    Astropher Guest

    Thanks very much for the suggestions Paul. I will try yours and
    Vanguard's suggestions the next time I power the machine down fully.

    I normally run the machine 24/7 except for when I plan to leave the
    premises for longer than an hour during the storm season (which is
    now). We get vicious electrical storms here in the sub-tropical and
    tropical areas of Eastern Australia during the summer and they dvelop
    Astropher, Jan 19, 2015
  8. Astropher

    Flasherly Guest

    Can also pull the plug completely from the wall, plug it back in, turn
    back ON the power-supply's toggle. Don't allow, of course, provide it
    with the requisite 20min.

    Then hold down [IN] the case's PWR ON for 10 to 15 seconds, (for
    effecting an continuous open condition on the PWR ON MB jumper

    See if it comes up then. I've run into something similar with PS
    logic (both ways turning off, or on, when holding in switch/logic
    exhibits that "extended" effect).

    Might look to see if your BIOS setting for PWR ON/Resume upon
    restoration of POWER (due to brown/blackouts) will function. Both
    ways, to include disabling it.
    Flasherly, Jan 19, 2015
  9. Astropher

    Flasherly Guest

    'scuse. That should be closed/shorted contacts
    Flasherly, Jan 19, 2015
  10. Astropher

    Astropher Guest

    It appears that it was the CMOS battery. I replaced the battery a
    couple of days ago, and today the machine was powered down with the AC
    disconnected for 12 hours. It started up again properly without delay.

    Astropher, Jan 22, 2015
  11. Astropher

    Paul Guest

    Weird. And thanks for posting back. +1 for VanguardLH.
    And another item for the "big book of busted hardware" :)
    Where we learn by remote observations.

    Paul, Jan 22, 2015
  12. Astropher

    Astropher Guest

    It is pretty selfish not to follow through by posting back results.

    By strange coincidence, this week a friend emailed about how his
    computer was taking a long time to power up. It exhibited pretty much
    the same symptoms as mine. I know he has a relatively new power supply
    and It is also a Gigabyte motherboard. I will post the resuts when he
    eventually gets around to replacing his battery.
    Astropher, Jan 23, 2015
    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.