Re: Computer turns on and then just dies

s|b <[email protected]> wrote:

> On Thu, 21 Nov 2024 12:22:45 -0600, VanguardLH wrote:
> 
>> After replacing the CMOS battery, did you reset the CMOS table (to avoid
>> corruption of its settings, and restore them from EEPROM)?
> 
> I did the reset, but don't remember if that was before or after I
> replaced the battery. Will do it again.
>  
>> When you apply power, do the fans spin, especially the CPU fan?  Or do
>> they just wiggle a little, and stop?  Or not spin at all?
> 
> CPU fan, case fan, power supply fan, graphic card fan, they all spin at
> high speed, but only for a few seconds.
>  
>> Used air cans to dust out the inside, like the heatsinks on the CPU and
>> GPU?  If those are filthy, or not well secured to the CPU and GPU, or
>> the thermal paste has dried out, the heat from the chip cannot get
>> dissipated from the heatsink
> 
> I'm beginning to think this could be it. Wouldn't there be warning beeps
> though?

Thermal protection does a power shutdown.  While they measure
temperature, reaching that temperature means it is on the rise, and
there is more heat building up.  The component will get even hotter,
especially when fans are turned off (so it be smarter to keep the fans
running for a minute after the thermal shutdown, like how some electric
fans on car radiators keep spinning after turning off the car).  They
have to shut down at temperature that permits sufficient time to prevent
further buildup of heat causing damage.

In some setups, the CPU gets throttled (runs slower) to reduce its heat
output.  That makes the computer really slow, but it still starts.
However, if the temperature is too high, the thermal cutoff is immediate
to prevent burning up stuff.

Often you can use your nose as a detective.  Sniff around inside, like
the exhaust side of the PSU (that pushes air into the case), the voltage
regulator capacitors around the CPU.  Although you could get an infrared
temperature probe, just putting a finger on heatsinks and memory modules
might hint stuff is getting too hot.  Same for touching the drives.
Just remember that what is too hot for your finger isn't that hot for
electronics.  The CPU is probably rated for an *operating* temperature
up to 80C (176F), but your finger won't enjoy that temperature.  Third
degree burns happen at 66C (150F).

You won't know if new thermal paste is needed until you separate the
heatsinks from the chip, but then you'll need to repaste, anyway.  Don't
reuse thermal paste.  After separating the two, clean off the old paste,
even if it looks good, and apply new paste.  Most users apply thermal
paste way too thick.  Metal-to-metal is best for thermal transfer.  Air
is very bad.  The paste is used only to fill the microscopic gaps
between chip plate and heatsink.  Unfortunately not all chip plates and
heatsinks are perfectly flat, so the paste also accomodates larger gaps,
but thermal paste is not as good as metal-to-metal.  More paste is not
better.  Rare few users lap the plate atop the CPU and the bottom of the
heatsink to get them as flat as possible to ensure maximum
metal-to-metal contact.  

Some setups don't use paste, but thermal strips as those are more neat
to apply (less mess), don't require expertise in applying a *thin* layer
of paste, and don't ooze out under the pressure of clamping the heatsink
to the CPU.  Those liquidfy under pressure and heat: they get applied to
the heatsink which is pressed down onto the CPU which heats up, and the
tape liquidfies.  Thermal strips are not as effective as thermal paste,
but thermal paste gobbed on too thick is almost as bad as not using any.
If you do end up scraping off the old paste and applying new paste, be
careful of those containing metallic particles.  The premise is the
metal particles provide better thermal transfer, but any that oozes out
when the heatsink is clamped onto the CPU can short out foil paths and
component connections on the mobo around the CPU.

>> When powering up, also check the PSU fan spins.  If not, it also has
>> thermal protection.
> 
> Last time I checked it spinned, but I will take extra notice next time.

Most PSUs have only 1 fan: exhaust (into the interior of the case).
Some have 2 fans: intake (outside air into PSU) and exhaust (inside of
PSU into the case).  Mostly those are for redundancy: if one fan fails,
the other is sufficient.  

There are some PSUs that don't push any air inside the case.  Usually
those are bottom-mounted in the case.  They suck up air from the bottom
and push out to the side.  The outside air goes through the PSU, and
only through the PSU to cool just the PSU.  The idea is not to push
preheated air from heated components inside the PSU into the case.
Instead separate case fans are used to get cooler outside air inside the
case.

Although you may see the fans spin doesn't mean their spindles will spin
freely (minimum friction).  If a fan, especially the CPU heatsink fan,
doesn't spin fast enough, the BIOS can power down the computer.  Without
the fan spinning, there is no air passing over the heat sink to cool it.
Often the BIOS settings will show the fan speeds, but you might not be
able to navigate to that screen since the computer is powering off
pretty quick.  Take an ear swab, wood stirrer stick, or something thin
and non-conductive to flick the blades on the fan to spin it.  The fan
should not immediate stop when the stirrer stick stops pushing the
blade.  If accessible, you can also put your finger on the hub to spin
the blades attached to it.  If you feel resistance, or almost a
throbbing sensation (there is intermittent resistance), or the hub with
blades won't keep spinning a turn or two after you flick the hub, the
fan is worn out.  That could result in the fan not spinning fast enough,
the BIOS sees a slow fan, and shuts down to prevent thermal damage from
lack of air cooling.