Silent computer: only wire cutters and a soldering iron. Success story

We have a desktop. We sit working - we write the code or we look at the cats, and this box roars alongside. We have wire cutters, a soldering iron, a blue electrical tape, some wires, a little heat shrinkable cambric, and basic knowledge of electrical engineering and computer design. Is it possible to pick this box somehow so that it roars at least a little quieter?


First, two simple problems from the theory. We are talking about noise, about cooling.
Why is computer noisy? Because the fans are spinning.
And how much does the computer roar, and how does it depend on the fans?
Let's solve a simple problem: here is a tryk - and the speed of the fans will fall twice. How much decibels will the noise decrease?

Fans - this is important, it is a heat dissipation, the electronics should not overheat.
Then another problem: the cooler blows the processor. CPU temperature 35 degrees Celsius. And then suddenly the processor has increased heat dissipation twice. What will be the temperature of the processor?
')
The noise of the system unit consists of the noise generated by the coolers and the noise of the hard drives.
Skip the hard drives, it can not be changed and it is not the main source of noise.
Where does the noise of coolers come from? The air flowing around the radiator - once, and the noise of the mechanics of the cooler itself: the beating in the bearings, the sounds emitted by the windings of the cooler (irregularities of the cores, magnetostriction, etc.)

Turbulence

Calculate the Reynolds number for the cooler-radiator system.

Re = (flow rate) * (characteristic size) / (kinematic viscosity)

According to the study , the performance of the cooler ranges from 20-60 cubic meters per hour, or 0.005 - 0.015 cubic meters / sec. We take 0.01 cubic meters / second as the average for the hospital coolers.
The diameter of the middle cooler is 7 cm = 0.07 m.

Then the flow rate = (performance) / (area) = 0.01 / (pi * 0.07 * 0.07 / 4) = 2.6 meters per second

The characteristic size we already have is the diameter.

Then Re = (2.6) * (0.07) / (16/1000000) = 11000

According to the same wiki , and other sources, with such a number of Re, the air flow is turbulent .

When turbulent flow in the air there are vortices. Whirlwind is an independent formation, breaking off from the fan blades, the eddies form air pressure surges. The same thing happens when the vortex formation hits a hard surface. The pressure surges from numerous eddies create the same noise sound that comes from a working cooler.

Is it possible to make the flow laminar, that is, vortex-free, so that there is no noise? Even if the air flow rate decreases four times, the current will remain turbulent, vortex (the Re number will be equal to 2500).
However, there is an intuitive assumption that reducing the air flow rate will reduce the noise produced. About this below.

Noise cooler mechanics.
All sound effects from the mechanics of the cooler are directly dependent on the strength of the current flowing through the windings. The nature of these sound effects is diverse and complex, and there is no sense to argue on the fingers. However, it is obvious that when the current flowing through the windings decreases and the cooler's rotation frequency decreases, the sound effects will decrease.

Indeed, why talk on the fingers, if there is such a study . It is dedicated to the fans, but the cooler is the fan, all conclusions will be fair. It shows only the flow rate, but from the theory of fans it follows that the air flow, and hence the flow rate, is directly proportional to the rotational speed (rpm). We halve the speed - two times less than the flow rate.

So, all fan noise is proportional to the rotational speed. Quote the study

It is known that turbulent noise is a source of quadrupole type, and its sound power is proportional to ~ u8, and vortex noise is a source of dipole type, and its sound power is ~ u6, where u is the peripheral speed. When the rotation frequency decreases, the rotation noise also decreases, which has a dipole (load noise) and monopole (drive noise) nature, and their sound power is proportional to ~ u6 and ~ u4, respectively.

So, the power of turbulent noise is proportional to the eighth (!) Degree of rotation frequency. That is, the sound pressure level of noise is proportional to the fourth power of the frequency (level = root of power = root of 8 degree = 4 degree).
Wow!
Here is the volume scale (numbers in dB)

Hearing threshold 0
Ticking wristwatches 10
Whisper 20
The sound of the wall clock 30
Muffled conversation 40
Quiet street 50
Ordinary conversation 60
Noisy street 70

Let's take on this scale the volume level of the system unit is 65 dB - the average between normal conversation and a noisy street.
Reducing the fan speed by half will reduce the noise from turbulent air flow by 2 ^ 4 = 16 times, or 24 dB, that is, the volume level will be equal to the muffled conversation. Good effect!

Thermodynamic calculation
We wrote above - slow down twice. Why two? Why not at twenty? Or turn off the fans altogether, there will be no noise? After all, some air flow is required for cooling, here we reduce it - and what happens, for example, with a processor?

Let's solve the problem, which is at the beginning of the post.

So, the processor temperature = 35 degrees, the processor heatsink is regularly blown by the cooler, the flow rate does not change.
The processor has increased the heat twice. What will be the temperature of the processor with a constant flow of air?
Consider the process of blowing. Imagine that the “cooler-radiator-processor” system is a black box that transfers heat to the air flow. The amount of heat is proportional to the temperature difference between the processor and incoming air, that is, (processor temperature is the temperature of the air entering the cooler).
Inlet air temperature = temperature inside the system unit. Take it equal to 30 degrees.
Total temperature difference between the processor and the incoming air = 35 - 30 = 5 degrees. If the processor doubles the heat generation, the temperature difference will also double, that is, it will become equal to 5 * 2 = 10 degrees.
And the temperature of the processor will be equal to the air temperature + difference = 30 + 10 = 40 degrees (instead of 70, which immediately come to mind)

And what will happen to the CPU temperature if the cooling flow rate is halved?
According to the Handbook of Physics Kuhling K, the amount of heat carried away from the surface of the “air-hard wall” is proportional to the temperature difference and the root of the air flow velocity. If you reduce the air flow rate by two times, then to divert the same amount of heat, the temperature difference will increase by a factor of 2, that is, about 1.4 times.

The final formula:

Tprocessor = Air + (Tprocessor initial - Air) * root ((initial rotational speed / total rotational speed))

Perform a couple of real calculations from practice.

Tprocessor initial = 42 degrees
Air = 20 degrees (system unit open)
Rotation speed reduced by 2 times (from 2000 to 1000 rpm)
Tprocessor = 20 + (42-20) * 1.4 = 51 degrees

More conditions
Tprocessor initial = 60 degrees (old hot Athlon)
Air = 20 degrees (system unit open)
Rotation speed reduced 2.5 times (from 2000 to 800 rpm)
Tprocessor = 20 + (42-20) * 1.4 = 76 degrees

We see that even on a hot processor (temperature difference = 40 degrees) and a 2.5-fold reduction in rotational speed (= 31 dB noise reduction), the processor temperature remains within acceptable limits with a good margin.

Here on the knee test results on the processor Core i5 series 3, it is clear that even on the stress test, the temperature increases by several degrees as the speed decreases 2.5 times.



Total: theoretically, reducing the speed of rotation of the ventilators is 2-2.5 times safe for electronics, which they cool

How to reduce the fan speed? Reduce the supply voltage.
And how does the rotation speed depend on the supply voltage? Approximately linear, see the plate above.

We begin to upgrade the system unit
First of all, we have at least two fans - one in the power supply and one on the processor. Maybe another one on the video card, on older models - also on the motherboard.
What do we do:
1) turn off all the fans from the power + 12V
2) connect them together and connect to power + 5V
3) we test

1) So, we get and disassemble the BP.

The fan is just gently biting off the board (there may or may not be a connector, it does not matter). Solder the long wires to the wires coming from the cooler, bring them outside

We bite off the processor processor all the wires, do the same operation. The CPU cooler can have three or four wires, respectively. ground, + 12V, rotation sensor and (optionally) speed control. We are only interested in + 12V, the ground and everything else can be left directly in the connector

But you can bite off the ground and + 12V, this is better (this is Pentium4)

If there is a cooler on the video card - the same thing: bite and lengthen the wires

Another video card, this is a GeForce 8600 gts


Attention!
1) make sure the solder joints are reliable, pull them to strength
2) in any case, do not use electrical tape for insulation. Only shrink crib, electrical tape - only for temporary mechanical fixation of wires
3) proceed from the principle that everything that can be closed - necessarily closes, everything that can fall off - will fall off
We can reduce the rate of blowing radiators, but if you suddenly have a node on the move will remain without blowing (soldering will fall off for example), you will know about it when this node fails. Do it securely.


2) We connect all the wires from the coolers, observing the polarity. The result will be two wires - plus and minus the entire computer cooling system (all coolers connected together).
Next, the problem will arise - how to connect to the power connector. You can drop out and use pins from old CD / DVD drives or faulty hard drives (it's better, safer - you won't pull it out)

or pins from motherboards. But then such a connection must be fixed mechanically, at least for the first time with tape, such pins are held much weaker

3) we check everything carefully, pull the wires - so that suddenly nothing will fly out. Preparing for inclusion

First start
Be prepared to quickly inspect all fans immediately after switching on.
Turn on!
Quickly inspect and check the fans - they all have to spin. If suddenly there is some kind of stopping, we turn it off immediately (we will consider this situation below)
If all fans spin - approx. The power supply fan can be checked by attaching a piece of paper to the power supply unit - it should blow lightly.
OK
Your BIOS may start to curse at you - they say that the power supply fan or low speed is not spinning or is not regulated. Go to the BIOS, put the CPU FAN to ignore
OK
Everything! Enjoy the silence!
Run whatever the CPU temperature indicates - CoreTemp , RealTemp etc. Monitor the temperature for the first few days. Look at the fans.

And enjoy the silence!

How quiet is the computer after such a rework?
Quite quiet. Only the hard drive is heard.

Are there any risks of such a rework?
Not. You can always switch the power of all the fans back to + 12V, get back exactly the same thing as it was

What to do in a situation when one of the fans does not spin when turned on?
Perhaps he does not have enough torque to start (although this situation has never been observed). In this case, make the power supply of all fans not 5V, but 7V. How?
We connected + fans to + 5V to supply power + 5V, and the ground of the fans to the ground. Connect + fans to + 12V, and ground to + 5V. In this case, the fans will be 12-5 = 7V. A small noise will appear, however it is much quieter than it was.

How many computers were converted in this way?
Since 2008, the author of these lines has personally upgraded several dozens of various types: from Pentium4 to Core i5. The entire line of Intel families between the above processors: Core2Duo, Celeron G ***, Pentium G ***, etc. The machines on AMD processors are smaller, but still - from the old ones with a 939 socket to the new ones.

Were there any gaming or overclocked computers?
No, and here it is worth considering whether such upgrades are needed for these computers.

Were there any problems?
Associated with modernization (overheating, failure of cooled units) - no

How long does this upgrade take?
About two to three hours

How do computers upgraded in this way behave in the summer?
As usual. No glitches fixed

Any add. maintenance measures?
Perhaps it makes sense to blow dust and clean the computer more often. By the way, dust accumulates in such a computer less (apparently, due to the smaller amount of air)

And what is the know-how?
Not with anything. The method described above has long been known to electronics engineers and is actively used. Auto publishing just tried to bring a theoretical basis for such a modernization and showed examples of practical implementation.

Is it possible to do something similar on laptops?
In no case! Laptop electronics work in very difficult thermal conditions, and the cooling system is different from the desktop.

UPD2 And finally. Yet, before doing anything with the cooling system, do an analysis.
1) record the temperature of the motherboard with any program (it can be considered equal to the temperature inside the system unit) and the temperature of the processor
2) run a stress test for the processor (available in many programs)
3) Record the processor temperature 10-15 minutes after running the stress test
4) perform the calculation of the temperature of the processor when the rotational speed decreases (according to the formula in the article)
If the end result is acceptable, you can redo it. If you get a temperature close to the limit for the CPU, refrain from these alterations.

(But this situation is more likely related to the gaming / overclocking machines item)