• Is CPU fan startup controlled by CPU or some other circuit?

    From Woozy Song@3:633/10 to All on Sunday, November 01, 2026 13:55:15
    I got a PC to check. CPU fan spins up, stops then starts again as usual. Nothing else happens - no video display, keyboard and mouse don't light
    up. No BIOS beeps. Removed RAM to trigger a fault, it still doesn't
    beep. Swapped graphics card, CMOS battery and power supply, still
    nothing. So mainboard is totally kaput.

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Paul@3:633/10 to All on Sunday, November 01, 2026 02:52:08
    On Sun, 11/1/2026 1:55 AM, Woozy Song wrote:
    I got a PC to check. CPU fan spins up, stops then starts again as usual. Nothing else happens - no video display, keyboard and mouse don't light up.

    No BIOS beeps. Removed RAM to trigger a fault, it still doesn't beep. Swapped graphics card, CMOS battery and power supply, still nothing.
    So mainboard is totally kaput.

    Check PSU voltages with a multimeter.
    Your +5VSB may be overloaded, and the PC cannot start
    without it. Make sure all loads are disconnected from
    USB and PS/2 ports first (even if this leads to a
    no-keyboard beep).

    Check the CMOS battery voltage with your multimeter.
    Should be 3V for a mid-life battery. They can be
    close to 3.3V when brand new. The value must be greater
    than 2.4V (2.0V for RTC plus 0.4V for the Vf of the diode
    in the battery path). A 100 ohm resistor is used to limit current
    flow. One poster managed to make the CR2032 battery
    explode, but, that took "skill" and that is unlikely
    to happen to you. They don't normally do that, and
    you really have to work at abusing the battery, for
    it to do that.

    The PSU has a PWR_OK, and again, the motherboard
    will not start if the PWR_OK is not asserted by the PSU.

    When these things happen, you are fault isolating
    to multiple things. It could be more than one power
    supply failure. It could be the motherboard shorted
    and causing the power supply to shut down.

    Some boards have four white LEDs which are like
    "staging LEDs". They advance as the boot process
    gets closer. And to advance, that is the BIOS firmware
    that under program control, drives the process forward.

    But the board won't come out of RESET unless all
    the indicators are good. The motherboard does not
    "measure" 3.3V, 5V, 12V for the purpose of not
    starting. It relies on the PSU to do the measuring
    and assert PWR_OK when the voltage is in the
    window for each. The +5VSB is used by the motherboard
    logic and must be present. The -12V these days, operates
    a serial port chip, such as a 75232 (it runs of -12V and
    +12V to make RS232 signal levels).

    Intel used to host the site with the power supply specs.
    Archive.org now has copies.

    ATX12V_PSDG_2_2_public_br2.pdf

    https://web.archive.org/web/20070112183223if_/http://www.formfactors.org:80/developer/specs/ATX12V_PSDG_2_2_public_br2.pdf

    On page 37 is the main connector. Pin8 (gray) is PWR_OK.
    Pin9 (Purple) is +5VSB. Both should be 5.0V when the
    power supply is running. The +5VSB is only rated for
    2.5 amps to 3 amps, which is why it is relatively easy
    to overload it.
    (ATX plugged in)
    +5VSB -- reg -- 3VSB ----------------- diode --> ------+
    |
    BAT45(3legs) +----- Southbridge (well)
    SOT23 | RTC and 256 byte CMOS RAM block
    CR2032 --- 1000 ohm ---diode --> ------+
    resistor

    On some boards, if you don't follow the instructions when
    clear_CMOS with the jumper, the BAT45 is damaged bad enough
    it gets burned. That is an ORing diode device, and it
    switches to the CMOS battery when the line powered 3VSB is
    not present. If the diodes are burned open-circuit, the
    SuperI/O hardware monitor may have a connection to the
    wire on the right, and it prevents the board from starting
    if that node is at zero volts. Not all SuperI/O have this
    feature, but some do. In some cases, you can visually trace
    this circuit near the CR2032 battery socket. The three-legged
    device may have "045" printed on it. It is a SOT23. The
    diodes are Schottky, for their lower forward voltage at
    relatively low operating currents (10uA).

    The SuperI/O can be driving the fan. Or, on modern boards
    you have a PWM (pulse width modulation 5V logic signal to
    the fan to control it). The fan should not really stop
    if it follows the Intel spec. But some brands are not
    compliant with the spec for four pin fans, and that's OK.

    It may be that the BIOS advanced part way though the
    POST code. If you had a Port 80 card with the two or more
    Hex Digits, those can show the advancement of the BIOS
    through the POST subroutines. A value of 0x00 or 0xFF indicates
    the Port 80 card has received no write operations. Other
    values are progress indicators (similar to the four white LEDs
    purpose). Some $400 motherboards have the two digit thing
    on the surface of the motherboard, but you won't be
    repairing too many PCs with those built in. At one time,
    the Port 80 card was for the PCI #1 slot. Now, there is
    a version made for the PCI Express #1 slot (x16 wide) and
    the Port 80 card would be an x1 wide single lane card.
    And you would have to plug it in, in place of the video card.
    Those cards aren't too common, but you may find some on
    Alibaba. I doubt my computer shop carries those.
    (They're short on a few things they should stock.)

    I like to test a motherboard (when it is brand new),
    outside the computer case. I set it on some insulating cardboard
    while I work on it. That also gives me a chance to do
    a visual inspection (look for burn marks, leaking
    capacitors and so on).

    Paul

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From VanguardLH@3:633/10 to All on Sunday, November 01, 2026 10:42:45
    Woozy Song <suzyw0ng@outlook.com> wrote:

    I got a PC to check. CPU fan spins up, stops then starts again as usual. Nothing else happens - no video display, keyboard and mouse don't light
    up. No BIOS beeps. Removed RAM to trigger a fault, it still doesn't
    beep. Swapped graphics card, CMOS battery and power supply, still
    nothing. So mainboard is totally kaput.

    Is this a prefab, or home built? If home built, did the jobber forget
    the thermal paste between CPU and heatsink? If prefab, how old?
    Thermal paste dries out over time, and some makers use thermal tape
    instead of paste.

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)
  • From Woozy Song@3:633/10 to All on Monday, November 02, 2026 08:14:11
    Paul wrote:
    On Sun, 11/1/2026 1:55 AM, Woozy Song wrote:
    I got a PC to check. CPU fan spins up, stops then starts again as usual.
    Nothing else happens - no video display, keyboard and mouse don't light up.

    No BIOS beeps. Removed RAM to trigger a fault, it still doesn't beep.
    Swapped graphics card, CMOS battery and power supply, still nothing.
    So mainboard is totally kaput.

    Check PSU voltages with a multimeter.
    Your +5VSB may be overloaded, and the PC cannot start
    without it. Make sure all loads are disconnected from
    USB and PS/2 ports first (even if this leads to a
    no-keyboard beep).

    Check the CMOS battery voltage with your multimeter.
    Should be 3V for a mid-life battery. They can be
    close to 3.3V when brand new. The value must be greater
    than 2.4V (2.0V for RTC plus 0.4V for the Vf of the diode
    in the battery path). A 100 ohm resistor is used to limit current
    flow. One poster managed to make the CR2032 battery
    explode, but, that took "skill" and that is unlikely
    to happen to you. They don't normally do that, and
    you really have to work at abusing the battery, for
    it to do that.

    The PSU has a PWR_OK, and again, the motherboard
    will not start if the PWR_OK is not asserted by the PSU.

    When these things happen, you are fault isolating
    to multiple things. It could be more than one power
    supply failure. It could be the motherboard shorted
    and causing the power supply to shut down.

    Some boards have four white LEDs which are like
    "staging LEDs". They advance as the boot process
    gets closer. And to advance, that is the BIOS firmware
    that under program control, drives the process forward.

    But the board won't come out of RESET unless all
    the indicators are good. The motherboard does not
    "measure" 3.3V, 5V, 12V for the purpose of not
    starting. It relies on the PSU to do the measuring
    and assert PWR_OK when the voltage is in the
    window for each. The +5VSB is used by the motherboard
    logic and must be present. The -12V these days, operates
    a serial port chip, such as a 75232 (it runs of -12V and
    +12V to make RS232 signal levels).

    Intel used to host the site with the power supply specs.
    Archive.org now has copies.

    ATX12V_PSDG_2_2_public_br2.pdf

    https://web.archive.org/web/20070112183223if_/http://www.formfactors.org:80/developer/specs/ATX12V_PSDG_2_2_public_br2.pdf

    On page 37 is the main connector. Pin8 (gray) is PWR_OK.
    Pin9 (Purple) is +5VSB. Both should be 5.0V when the
    power supply is running. The +5VSB is only rated for
    2.5 amps to 3 amps, which is why it is relatively easy
    to overload it.
    (ATX plugged in)
    +5VSB -- reg -- 3VSB ----------------- diode --> ------+
    |
    BAT45(3legs) +----- Southbridge (well)
    SOT23 | RTC and 256 byte CMOS RAM block
    CR2032 --- 1000 ohm ---diode --> ------+
    resistor

    On some boards, if you don't follow the instructions when
    clear_CMOS with the jumper, the BAT45 is damaged bad enough
    it gets burned. That is an ORing diode device, and it
    switches to the CMOS battery when the line powered 3VSB is
    not present. If the diodes are burned open-circuit, the
    SuperI/O hardware monitor may have a connection to the
    wire on the right, and it prevents the board from starting
    if that node is at zero volts. Not all SuperI/O have this
    feature, but some do. In some cases, you can visually trace
    this circuit near the CR2032 battery socket. The three-legged
    device may have "045" printed on it. It is a SOT23. The
    diodes are Schottky, for their lower forward voltage at
    relatively low operating currents (10uA).

    The SuperI/O can be driving the fan. Or, on modern boards
    you have a PWM (pulse width modulation 5V logic signal to
    the fan to control it). The fan should not really stop
    if it follows the Intel spec. But some brands are not
    compliant with the spec for four pin fans, and that's OK.

    It may be that the BIOS advanced part way though the
    POST code. If you had a Port 80 card with the two or more
    Hex Digits, those can show the advancement of the BIOS
    through the POST subroutines. A value of 0x00 or 0xFF indicates
    the Port 80 card has received no write operations. Other
    values are progress indicators (similar to the four white LEDs
    purpose). Some $400 motherboards have the two digit thing
    on the surface of the motherboard, but you won't be
    repairing too many PCs with those built in. At one time,
    the Port 80 card was for the PCI #1 slot. Now, there is
    a version made for the PCI Express #1 slot (x16 wide) and
    the Port 80 card would be an x1 wide single lane card.
    And you would have to plug it in, in place of the video card.
    Those cards aren't too common, but you may find some on
    Alibaba. I doubt my computer shop carries those.
    (They're short on a few things they should stock.)

    I like to test a motherboard (when it is brand new),
    outside the computer case. I set it on some insulating cardboard
    while I work on it. That also gives me a chance to do
    a visual inspection (look for burn marks, leaking
    capacitors and so on).

    Paul

    The CMOS battery was only 2 volts, I repaced it with a good one.
    The mainboard (H97-DH3) has no indicator LEDs on it, not even one next
    to RAM slots to show there is power on. I took CPU out, it worked in
    another mainboard. Gave up on, as mainboard would be 11 years old, not
    really worth saving.

    --- PyGate Linux v1.5.19
    * Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)