CPU Fans - Which, when, and why

[socketa]

Done lots of searching, but haven't found much useful info on this:
I put a fan/heatsink back onto a good MS-7191 socket 939 motherboard that i brought 2nd hand some time ago, and it just twitches back and forth (which makes me wonder how (or if) it was working on that board in the first place).
So i had four similar sized AMD fans that i tried, and here are the results (including the one just mentioned):
Black Red Blue 0.45A : Just twitches back and forth (and sets of BIOS No fan alarm)
Black Red Blue 0.24A : Twitches once (and sets of BIOS No fan alarm)
Black Red Blue 0.08A : Works OK ~2812RPM
Black Red Yellow 0.1A : Works OK ~3245RPM
Black Red Yellow Blue 0.3A: Works OK ~3125RPM

So, which fan(s) would meet this particular board's specification (the user manual doesn't state this)?
Interesting that the "0.3A" fan works, but the "0.24A" doesn't.
And, on the 3-pin fans, does the yellow wire being in place of the blue wire mean anything significant?

[stj]

they could be faulty or it could be a board fault not providing enough current.

[socketa]

is it possible that they are not faulty, and that the specs of this board wont supply such fan current?
And after some more searching i just found this: "Some motherboards can only supply up to 2A or so on each fan connectors" - https://linustechtips.com/main/topic...amps-for-fans/

[R_J]

Does this picture help? Black is ground, Red is supply, Blue is tach/Sense. The Yellow on your fan would be for a board that has a PWM to control the fan's speed. Your board does not support the yellow pwm control so the fan will run at a constant speed.
this may help https://allpinouts.org/pinouts/conne...cpu-4-pin-fan/

[Per Hansson]

It is more common in my experience that the RPM/tach/sense cable is Yellow.
But as can be seen in the picture above it can be blue too.
There are standards for wiring colors but almost nobody cares

[eccerr0r]

Measure the voltage of the middle pin (+12V).

I've seen some motherboards try to reduce noise of 3-pin fans by reducing voltage. Doing this is not a good idea because some fans will not start spinning at the lower voltages. Check bios if you have some control, make sure all fans are set to "full speed" so those fans that require full voltage have the power needed to spin.

The 4-pin fans are the proper way to reduce noise, that 4th pin PWM controls the fan speed.

[momaka]

Originally Posted by eccerr0r

I've seen some motherboards try to reduce noise of 3-pin fans by reducing voltage. Doing this is not a good idea because some fans will not start spinning at the lower voltages. Check bios if you have some control, make sure all fans are set to "full speed" so those fans that require full voltage have the power needed to spin.

^ This.

I have the same exact motherboard, and indeed the BIOS has an option called "Smart Fan" that is enabled by default. This option reduces the speed of the CPU fan to reduce noise, and only increases the speed once a certain temperature threshold is reached (I think 40C is the default.)

But I've had the same experience as you: leaving that option (Smart Fan) default causes some fans to turn extremely slowly or not at all, triggering the CPU fan alarm.

To solve this, go into BIOS, and under the PC Health Status tab, set that option to "Full Speed". This will make the CPU fan turn full speed, which can be OK or extremely loud, depending on the fan you use.

Speaking of which, let's cover that next:

Originally Posted by socketa

So i had four similar sized AMD fans that i tried, and here are the results (including the one just mentioned):
Black Red Blue 0.45A : Just twitches back and forth (and sets of BIOS No fan alarm)
Black Red Blue 0.24A : Twitches once (and sets of BIOS No fan alarm)
Black Red Blue 0.08A : Works OK ~2812RPM
Black Red Yellow 0.1A : Works OK ~3245RPM
Black Red Yellow Blue 0.3A: Works OK ~3125RPM

So, which fan(s) would meet this particular board's specification (the user manual doesn't state this)?
Interesting that the "0.3A" fan works, but the "0.24A" doesn't.

So it's worth noting here that for AMD CPU fans, depending on who made the heatsink (Taisol, Cooler Master, AVC, and etc.), the fans would also vary quite a bit in speed and performance.

I find Taisol usually used Delta "screamer" fans that were extremely loud on full 12V power and always needed "tweaking" (whether through the motherboard or external controllers or resistors mods) to bring the noise to a more sane level.

Meanwhile, AVC and some Cooler Master fans often came with a built-in NTC thermistor that would decrease the speed of the fan when the ambient temperature was low. Although a lot more quiet, I don't like these fans for two reasons:
1) The NTC thermistor circuit was sometimes implemented such that it affected the RPM/tacho output, and the fan would show speeds "all over the place" or none at all.
2) With motherboards like the MS-7191 that has fan control enabled by default, the NTC thermistor circuit in these fans would sometimes lower the voltage going to the fan too much and cause the fan not to start at all and/or twitch, like you observed (and I had the same exact experience with my MS-7191 board.) For these fans, set the board BIOS option to "Full Speed".

The good news about the MS-7191 board is that it does allow fan speed to be controlled through software as well. For a very long time, I used a little utility called SpeedFan (link) to manually set the CPU fan speed on my MS-7191 motherboard, because I changed from an AVC heatsink to a TaiSol heatink that had an extremely loud and powerful Delta fan (6500 RPM at full speed!!! - a serious screamer for a 70 mm fan.) Thus, I set my BIOS to have the CPU fan run at "full speed", and then when the OS booted, I used the SpeedFan software to decrease the CPU fan speed to a level that was quiet but still fast enough to keep the CPU cool under most circumstances (occasionally, I'd bump up the speed when gaming in the summer, for example.)

On a slight tangent, the MS-7191 motherboard also allows for CPU voltage and clock multiplier to be changed through software. This is useful, because many socket 939 CPUs can be under-volted to save power and run cooler. Depending on the CPU you got, you can shave 10-15 Watts from its maximum power draw, which on the stock AMD heatsinks will show significant results in terms of temperature drop at full load. I find CrystalCPUID (link) to work very well for controlling CPU V_core and multiplier on s754 and s939 CPUs.

Once you find what is the lowest voltage that your CPU can take at full speed without crashing (you can use OCCT to stress-test the CPU), take that voltage and use the next notch up.

For my Athlon 64 FX-57, IIRC I was able to run it mostly-stable down to 1.3V at full 2.8 GHz speed, instead of the stock 1.4V V_core. So I opted for 1.325V, IIRC, and the chip is perfectly stable at that. Doing this, my power consumption @ full CPU load dropped by about 12-13 Watts less, and so my CPU runs about 5C cooler. I've tested my CPU at this voltage long enough that I even set the CrystalCPUID shortcut with that voltage preset to run at Windows startup. (NOTE: if you are using Windows Vista or 7, you'll have to set the CPU to run at 100% speed in both the minimum and maximum power states in the Power settings in Control Panel. Otherwise, the OS will revert your settings when it throttles the CPU.)

Last but not least... and this is even more on a tangent... but you may want to recap that motherboard - at least those OST caps around the RAM sockets and the Northbridge chip. Mine had Teapo SM caps, and they were all failed. Recap thread here:
https://www.badcaps.net/forum/showthread.php?t=63069

Originally Posted by socketa

And, on the 3-pin fans, does the yellow wire being in place of the blue wire mean anything significant?

Nope.
In either case, it is used only for RPM/tacho monitoring.

Originally Posted by Per Hansson

It is more common in my experience that the RPM/tach/sense cable is Yellow.
But as can be seen in the picture above it can be blue too.
There are standards for wiring colors but almost nobody cares

Well, this board is back from the days when PWM fans were still not very common (at least in consumer stuff.) Thus, some fans came with a blue wire for the RPM/tacho, and others with yellow. So same function, but different color wire.

If it's a 3-pin fan, generally one can assume the fan's 3rd wire is for RPM/tacho. Very rare to see 3-pin fans that are PWM with no tacho. The PS3 slim fan IS one that fall in this category, though, in case anyone cares.

[socketa]

Thanks for that very informative post.
Any idea of what fan you were running in that board that you recapped?
With smart fan off, the 0.45A Delta fan (https://commons.wikimedia.org/wiki/F...AFB0712HHB.png) still initially twitches (or runs fairly slow), but after a short while it starts spinning and fluctuates up to 950 RPM, with the alarm turning on and off, as the speed rises and falls above and below 700RPM.
By the time it's upto 800RPM, the CPU temp is 51 degrees celcius
With the 0.24 AFB0712MB Delta Fan, it just vibrates every 3 seconds and doesn't even spin (i shut the board down when it reached 51 degrees)
I checked the voltage over these periods - and it's essentially 12V all of the time
Also, plugged in four case fans (sunon 0.25A, T&T 0.24A, AVC 0.3A, AVC 0.41A) and they all worked well
So are these fans kaput, or isn't the board designed to have enough power to drive these Delta fans?
Or are these particular fans prone to this type of failure?
How does one select the appropriate fan for a particular board?

I also tried them in a 7VM400M-RZ board, and they didn't spin in that either
Incidentally i also noticed that with that board (7VM400M-RZ) the CPU fan fail warning is not activated, even if i remove the good fan from from the CPU header and can hear the BIOS "Good post beep" - i even let it boot into the OS with speakers plugged in and still no fan fail warning, And, yes, the CPU fan fail warning is enabled in the BIOS.
What would cause that? And would there be any innate protective measure to shut down the computer before it destroys the CPU?

[stj]

make an adapter and test the fans from the hadrdrive connector

[socketa]

exactly the same deal there - so that means that the fan windings are burned out?

[stj]

or the driver chip is bad.

[socketa]

is this "driver chip" a transistor?

[eccerr0r]

All BLDC motors have a driver chip...

[socketa]

OK thanks
i watched https://youtu.be/bCEiOnuODac ,
and see that the hall sensor could be faulty as well

[eccerr0r]

Most of the fans I've seen nowadays the hall sensor is integrated into the driver chip...

[eccerr0r]

I do have to say I'm intrigued at the failure mode of the fan.

By far the most common BLDC fan motor failure I've seen is... guess what? You guessed it: bearing failure.

I think I've only come across one motor that had chip failure, and that may have been due to poor chip design coupled with... guess what? You guessed it: bearing failure...

[ChaosLegionnaire]

yupp, due to the age of the equipment being from the 2004-5 s939 era, im going to take a guess and say that the oil in these sleeve bearing fans have dried up causing them to spin more slowly than rated if u have the datasheet for the fans and pay attention to the rpm sensor readings. i had that in some of my sleeve bearing fans. taking them out and adding a couple of drops of sewing machine oil into their bearings brought their rpm back to their nominal ratings.

since this is bcn, like not using a board with bad caps since that could kill the board permanently and irreversibly, i would avoid using a fan that isnt working right until u figure out and fix whats wrong with it thats causing it to spin slowly.

[socketa]

Originally Posted by eccerr0r

I do have to say I'm intrigued at the failure mode of the fan.
By far the most common BLDC fan motor failure I've seen is... guess what? You guessed it: bearing failure.
I think I've only come across one motor that had chip failure, and that may have been due to poor chip design coupled with... guess what? You guessed it: bearing failure...

yeah, to me, that sounds more probable - i blow on the two unobliging fans (while unpowered) and they spin for roughly half as long as the others (although that could be because of the stronger magnetic field?)

i'll add a drop of oil later on and see what happens

[momaka]

Originally Posted by socketa

I checked the voltage over these periods - and it's essentially 12V all of the time
Also, plugged in four case fans (sunon 0.25A, T&T 0.24A, AVC 0.3A, AVC 0.41A) and they all worked well
So are these fans kaput, or isn't the board designed to have enough power to drive these Delta fans?

More than likely you got bad fans. Or their bearings are bad if they don't spin freely.

Originally Posted by socketa

How does one select the appropriate fan for a particular board?

Back in the days before fan control (i.e. fans running full tilt at 12V), certain fans were typically made for certain heatsinks. For example, a heatsink with dense fins would usually have a smaller, higher speed, higher static pressure fan. Whereas, HSs with fewer fins with more space would have larger fans.

And nowadays with PWM control (or even just regular voltage control, like your motherboard), it's more of "what is the maximum temperature I'm willing to let my CPU run without making the PC too loud?" Some people prefer more quiet PCs and will let the temperature drift closer to max. In the case of your motherboard, I think the fan RPM is turned up only to keep the temperature below a certain threshold... which may not work well with certain heatsinks, where the fan will crank up to 100% all the time under full load, because the HS cannot keep the CPU under the desired temperature. So for those cases, it all comes down to experimentation.

I think it's OK to let most s939 CPUs go up to 51-53C. 55C under absolute worst, but I don't usually let them go that high.

On that note, s939 CPUs lend themselves very nicely to Crystal CPUID software, which will allow you to change multiplier and voltage settings from the OS. Most 939 CPUs can be under-volted substantially before running into errors. A run-of-the-mill Athlon 64 3200+ can usually be under-volted to 1.25 to 1.3V on the core, instead of the stock 1.4-1.45V. This will typically shave around 10-12 Watts of the total power draw and easily decrease temperatures by 5C on most stock heatsinks. You can even create a custom CPU "profile" that varies the voltage and multiplier up and down with load for extra power savings (and lower temperatures.)

Originally Posted by stj

make an adapter and test the fans from the hadrdrive connector

+1

Or connect them directly to a DC wall adapter (8 to 12V will work, though some fans are OK even with as little as 5V.)

Originally Posted by eccerr0r

I do have to say I'm intrigued at the failure mode of the fan.

By far the most common BLDC fan motor failure I've seen is... guess what? You guessed it: bearing failure.

I think I've only come across one motor that had chip failure, and that may have been due to poor chip design coupled with... guess what? You guessed it: bearing failure...

Same here.

I rarely see fan electronics going bad. Usually when they do, it's because of a bad bearing, as you mentioned (be it sleeve or ball.)

Though for small video card fans, it's a different matter - I've seen a few of these cook their windings when ran at full rated voltage, even when their bearings didn't go bad. And also saw one Delta (IIRC) fan from an 2nd gen PS3, where the MOSFETs for one side of the stator coils went bad, but not the other, making the fan sometimes work if it stopped on the non-dead side (and do nothing if it stopped on the "dead" side.) Got that one fixed with a new pair of MOSFETs, though.

Originally Posted by ChaosLegionnaire

yupp, due to the age of the equipment being from the 2004-5 s939 era, im going to take a guess and say that the oil in these sleeve bearing fans have dried up causing them to spin more slowly than rated if u have the datasheet for the fans and pay attention to the rpm sensor readings.

The Delta fans that the O/P mentioned should be BB fans, though (or at least 1-ball, 1-sleeve type.) Not that I haven't seen those develop bad bearings either. But generally, even when they did, I've never seen them fail. Most quality fans like Delta, Nidec, Panaflo, and etc. typically have a stall fail-safe feature, where if the fan is stalled, it will stop and try to restart by itself, without burning out its coils.

Originally Posted by ChaosLegionnaire

i had that in some of my sleeve bearing fans. taking them out and adding a couple of drops of sewing machine oil into their bearings brought their rpm back to their nominal ratings.

Yes.

Though, it's better to do a full dismantle and bearing clean on sleeve bearing fans. I usually do it with a drill bit that is slightly smaller than the bearing diameter, and then rub (NOT SPIN) the drill bit in and out of the bearing. Then clean with paper towel and IPA. And then finally add oil and assemble the fans. With this method, I've had repaired fans run longer than they did from the factory (especially those that come packed with cheap grease.)

[socketa]

i opened up those two problematic delta fans
Of course, i initially took the sticker off of the wrong side (murphy's law) - thinking that access to the bearings would more likey be the opposite side of where the wires go in.
Yes, they are dual ball bearing fans
One of them had dust bunnies inside, and i gave both of them an oil with hair clippers oil
After giving them a spin by hand, they sound more noisey now!, so it does sound like (at least) a bearing issue.
Plugging them into 12V supply: one fan spins slowly and pulsing (about every 2 seconds) and has a rumbling feel to it.
The other fan just twitches and will do a revolution if i turn the fan to either two of the four (opposite) positions, that the fan can rest at. If it lands back on one of those positions it will go again, untill it lands on either of the other two positions, where it will just twitch.
Anyway, i'm done with these fans now - i have other ones that work OK (and i guess that the BIOS CPU temperature indicate if they are suitable), so no real problem.
Thanks for the suggestions