I'm just now working beside the computer, without the case on it and can observe the CPU fan blades perfectly well. They are motionless at all times. Even when the machine boots.

Of course, I'll solve this problem tomorrow, but tonight I want to keep using the machine. Is it very dangerous? The CPU is a Celeron clocked at 750MHz and the only appl which is putting some effort on it is the GUI. In the GUI I'm running the web browser and Gimp, a graphical desing tool... uuuf, I thing I'd rather shut it down right now!

I'd would be monitoring the CPU temperature each, say, 10m, but I lack the software. Quite a pity. Excuse me for this meaningless post but I leave it anyway in case somebody has an idea.

The fan shouldnt be totally off.

Either you've got a dead fan, a dead connector or the fan isnt connecting properly.

Thanks. Either way, I'll have it repaired right tomorrow, although only one minute could suffice to burn the CPU. If I were a wise man I'd shut down the GUI.

By the way: three cables go to the fan. Two must be power. Which leaves only one for control (for I've seen how it speeds up and down). But at the same time, the BIOS hardware monitor (BIOS menu at boot time) knows its speed. How can the same cable serve to control and sense. Is it bidirectional?

You can have supply current and (many different) signals on one wire.

Imagine 1,000 concurrent telephone conversations on one wire. No problem.

Of course! The telephone company sends a DC voltage and signal when you hook off. And they did it without electronics when the thing began.

That is called Frequency Division Muliplexing(FDM), AC (300-3000 Hz) vs DC (0 Hz). There is also space division multiplexing (different wires) and correlation division multiplexing (as in WiFi operating on the same frequency).

And you can run a Celeron 750 with heatsink but without a fan. It will stop operating but not fry with a probability of 90%. Poor people who don't have a spare fan lying around..

jlinkels

Frequency Division Multipexing and space division multiplexing, it's interesting. Of course, if not on the same wire, then on different wires, but using frequency to discriminate.

You info gives peace to my heart. I do not have spare fans lying around. I do have broken fans lying around. They are one of the components with higher failure frequency, I think.

Let me tell you I remember your nickname well, 'cause I have it in a file about audio/computer recording stuff. It begins "If I understand well, you have sound on Line In...".

Hmmm, didnt I tell you somewhere else that if you are running at 750Mhz, its either overclocked (to all hell), underclocked, or not a celeron?

There was no 750MHz celeron, 733MHz + 766MHz are the closest versions.

The GUI will be putting more load on the CPU than command line, but I'd be shutting down till you can fix the fan. A 60/70/80/92/120mm case fan fastened to the heatsink would do the job just fine, and if you used a bigger fan than the stock version would it probably run cooler and quieter than stock as well.

For a 3 pin CPU fan, the pins supply- ground, 12v power and a speed signal. BIOS/OS control over the 12v line is how the CPU fan changes speed.

Pulse-width modulation (PWM) is closer to what are you thinking, and generally has 4 pins- ground, +12V, sense, control.

You can get away with much faster than 750MHz intel P3 series fanless. I have, but not with the stock heatsink. I used a much, much bigger heatsink on an athlon 2200+ (1800MHz) downclocked to 1330MHz.

I also ran an athlon XP 2500+ fanless at slightly overclocked speeds fanless, but that was using water cooling.

Typically 3 wire CPU fans have a tachometer output to measure RPM. The speed most likely is changed by modulating the input voltage.

I would never run any modern system without a fan, unless I want to burn a hole in the mobo. For older computer you may be able to get away with no fan, but you may need to underclock.

I'd much rather run a 'modern' CPU like an iX/Core 2 XXX/Athlon II/Phenom II than an older CPU in a fanless seup.

With modern power/frequency control, it would be safer and run cooler. Even without that, modern CPUs tend to have similar or lower TDPs compared to Athlon XP/P4/etc .

Lower TDP = less heat to disapate = easier to keep cool.

I would let others test this theory on their systems. When I say old computers I mean older than P4.

In that case, been there, done that, not that hard at all. P3s/early athlons with a nasty big heatsink from a late model athlon run fanless pretty well.

I've run newer systems than that fanless, plently of times. I just havent got up to Athlon X2/Core2Duo CPUs yet.

Gimmie a year or so and I will have run a athlon X2 or higher fanless. I just need to get a Athlon AM2+ CPU for my 'spare' board.

As a matter of fact you can get away with almost anything fanless if your mechanical design was laid out for that.

Compaq desktop and servers had a long time a fanless CPU heatsink, oversized and the air sucked into the power supply was routed over this heatsink. My 1U server has no fan on the heatsink, but the air displaced by the case fans is funneled over the heatsink. That doesn't say anything.

The AMD CPU's in the 800 MHz range needed a correctly sized heatsink with fan. But the power dissipation was relatively low and these computers were able to run for some time without forced cooling. Eventually CPU operation became erratic and the machine would stop, but usually there was no damage to the CPU.

Unfortunately these CPU's had no thermal protection yet, so an incorrectly seated heatsink would cause a blown up CPU within seconds. Later CPU's have a thermal protection which shuts down the CPU when it gets too hot. Still I would not count on it.

Where it all comes down to is that a CPU generates heat (duh!) and this heat has to be transferred to the environment. Between the die and the air is a thermal resistance, expressed in K/W. (degrees Kelvin per Watt). Calculation is like Ohm's law, temperature difference is dissipation times thermal resistance.

For example, a CPU die can have a Rth of 10 K/W. When dissipating 30 Watts, the temperature difference between environment and die is 10 x 30 = 300 degrees. Whatever the environmental temperature is, this instantly blows up the CPU. When attaching a heat sink with a Rth of say 1.5K/W, temperature difference is 1.5 x 30 = 45 degrees. With an environmental temperature of 25 C, die temperature is 70 degrees, which is a familiar value.

There is also a thermal resistance between the chip internal and the outside of the die where the heat sink is attached. Internally the chips temperature is higher than on the outside of the die.

When no heat conducting compound is applied between die and heat sink, there is an additional Rth present which must be added to that of the heat sink. Without compound this can be as high as 0.5-1.0 K/W, meaning an additional temperature increase of 30 degrees. You want to avoid that, so after application of heat conducting compound this Rth drops down to 0.1-0.2 K/W.

For a given heat sink Rth is specified for still air, under certain conditions like vertical position of the fins. For heat sinks with forced air cooling like being used on CPUs Rth is specified under the condition of an airflow with a certain liters per minute.

Bottom line is that within limits a heat sink with forced air flow can be replaced with a fanless heat sink, provided the Rth is sufficiently low. In case of the AMD 800 the recommended heat sink could almost transfer sufficient heat to keep the CPU operational.

Put the same heatsink on a CPU with 72 Watts dissipation like the first AMD-64 CPU's and it would blow immediately, fan or no fan.

jlinkels

AFAIK, the third wire does not control the speed. It's for the tachometer inside the fan. I think it's most likely that your motherboard pulses the power to the fan on and off to lower the speed.

Newer fans actually have 4 wires, where the 4th wire is for speed control.

1 members found this post helpful.