Quiet PC Project

[lifeguard1999]

At some point in cooling down a CPU, having a higher CFM does no good in getting more o/c speed out of a processor. The same is true for cooling down the case. Instead all that excess CFM does is generate noise. I know where that line between excess CFM and noise is. Here is my strategy for o/c a 700E to 933 and still have a quiet case.

Goal: Reduce fan noise of o/c system. Hard drive/CD noise is acceptable since they are not that loud anyway.

Secondary Goal: Don't spend a lot of money.

Fact: For every ~3dBA, the noise doubles.
Fact: A whisper is about 23 dBA.

1. To reduce the noise of the power supply fan, I took an Enermax Thermal Control Fan and swapped it with the PS fan. This fan has the following specs:

Temperature and Fan speed / estimated airflow and noise
85C - 3100 RPM / 40CFM - 34dBA
65C - 2600 RPM / 33CFM - 29dBA
45C - 2100 RPM / 27CFM - 23dBA
25C - 1800 RPM / 22CFM - 18dBA

Now if you really think about it, you will see one of my problems. Being thermally controlled, the temperature has to climb a good ammount to increase the airflow. The airflow of the case is NOT going to be 25C. In fact, the core CPU temperature at idle is 25C. Sticking the thermal probe on the core gives a temperature less than 25C. Even if the CPU temperature rises to 85C (and it won't), the fan will never reach those speeds since the temperature it measures is not the core. To solve the problem, I placed the fan on the clock generator. Now it spins at 2100 RPM. The fan is quiet! The temperature of the case (mid-tower) is still low enough so that I cannot detect an increase.

The other problem is that ASUS Probe has an alarm that goes off when the fan is less than 2200 RPM. Even at 45C, the fan still spins less than that. Oh well.

2. Looking at overclockers.com, it seems that for a PEP66, anything over 33 CFM is simply excess airflow and generates a lot of noise. I am using an Alpha P3125, which is for slot 1 CPUs and is essentially twice a PEP66. Assuming that 33 CFM provides enough cooling for the P3125 heatsink, and by having two fans, I can run the fans slower and quieter and still get the appropriate airflow.

To do this, I took a fan adapter for the P3125 and placed 80mm fans on it. The 80mm fans move about the same CFM and are quieter than the 60 mm fans. Two high output 60mm ADDA fans move 56.6 CFM at 40 dBA. Two quiet 80mm ADDA fans move 52.4 CFM at 25.5 dBA.

Yes, I do realize that 2 fans in parallel do not move twice the air. So really the two 60mm fans are moving somewhat less than 56 CFM, and the 80mm fans are moving less than 52 CFM. But the calculations are close enough.

3. Using a variable resistor (also called a Rheostat), one can vary the voltage to the fan. This is better than the 7V trick, in my opinion, because it gives more options in fan speed. Lowering the voltage a bit will lower the RPM, CFM, and noise.

Since I am moving 52.4 CFM, and all I need is 33 CFM, I can slow down the already quiet fans and make them quieter while still meeting my cooling needs.

The truth is, they were so quiet already that it did not really gain me anything. Yet again, ASUS Probe did not like fans rotating less than 2200 RPM.

You can buy a kit from CaseEtc for $35 to do this. I simply went to Radio Shack. I did not have it front mounted, as I did not see that I would be adjusting the knobs all that much since it did not gain me anything. Plus my kids would just play with the knobs.

4. After closing up that case, I found that there still was a high pitched whine. I double checked the fans and none of them were producing the noise. After a bit, I found the source of the noise - my video card. I have a Radeon 32MB DDR. I am guessing that it produces about 30 dBA of noise. Since the rest of my system is running at about 26 dBA, this is significant. A Rheosat should solve that problem, but I have yet to put it on there. While I could run it without the fan (I think), again the gain in having less noise is not worth it.

The Cost?

$10 Enermax Fan
$16 Adda 80mm fans
$04 Shipping
$10 Three Rheostats from Radio Shack

Total: $40 for a ~26 dBA system.

Now I could have made this system quieter by using 3 Papst fans (8412NGML). One in the PS (27 CFM/19 dBA), 2 on the heatsink(54 CFM/22 dBA), and I end up with about a 23 dBA system. The problem with Papst fans is their cost and finding a place to buy them. The cost to benefit ration works against them.

100th post.

 

[Hanky]

Very interesting post!

In fact I'm also trying to find a good balance between cooling and silence. First of all the case is very important. Some cases are just "louder" than others (that means: higher noise when both have the exact same components). So you better look out for case reviews which include this aspect (very few do it...), to choose your case. My case, for example, has some kind of "sliding door" at front which silences the system enormously when "hiding" the 5,25" drives.
Then you can use some sound-absorbing material, for example cork which I use.

Then to the fans...in fact I'm not so sure whether your &quot;fact&quot; >>For every ~3dBA, the noise doubles<< is true, or if it's only true for the dB-scale. I've to admit I still don't understand the dB(A) scale completely, which somehow takes physiological effects into account. So, I also don't know how to add two dB(A) values since I'm not sure about it's mathematical nature (dB is logarithmic as everyone knows). That means I doubt your resulting dB(A) values ( >>Since the rest of my system is running at about 26 dBA<< ), but I can't disprove them either. Anyway, what really matters is the subjective noise and I truly believe a more or less silent PC is important unless you're only booting your OS to check whether the newest o/c speed works...

Oh btw: I only use Papst fans as my case fans (have a 8412NGM intake, and 8412NGML&amp;8412NGL outtakes right now). They're definitely great!

 

[lotust]

some good facts here, I myself am gettin so sick of all my fan niose. and all the god dam dust stuck in the fan filters after only on week.

my next case ill be doin soon will have some quiet features. lifeguard1999 where did you find all these FACTS? Ide like to read more about this. have a look at my niose maker.

I think im gonna get me one of thoes radio shace niose testerz. I think it would come in handy.

 

[lifeguard1999]

Hanky, where the *Beep* did you get your Papst fans?

Now as to Fact #1:

The equation used is: SL = 10log(I/Io) where I is the intensity of the sound waves measured in watt/m^2, SL is the sound level expressed in dB, and Io is the threshold of hearing or 10^-12 watt/m^2.

For SL = 30dB, you can solve for I and find it to be: I = 1E-9

A quick check is the rule-of-thumb that 3dB equals a doubling of sound intensity. Say you had two, 30dB fans. Then the sound level is [ 10*log(2.0E-9/1E-12) ] = 33.0103 dB. The difference between the two is 3.0103dB. Hence a doubling occurs every 3 dB.

Now as to Fact #2

A whisper is about 23 dB.

Whisper at 1 meter

So a whisper at 0.5 meter is about 23 dB.

As to the math.....

Now my fans are: 22.5 dB, 22.5 dB, and 23 dB. Since two are the same, we know that we can add 3 dB to it for their sum. That gives me 25.5 dB. Now, solve for I @ 25.5dB and I @ 23 dB.

Solve for I @ 25.5dB : I = 3.548E-10
Solve for I @ 23.0dB : I = 2E-10

Add them up (5.548E-10), recalculate SL [ 10*log(5.548E-10/1E-12) ] and you arrive at: 27.45 dB. It looks like I messed up my math in the earlier post. Oh well. Anyway, this assumes that they are in phase, and are not cancelling one another.

 

[lotust]

lifeguard1999
yer killing me with all thoes #s Aaaaaaaaaa


can I just see a picture of yor case? he he

 

[GiZzO]

Its cause hanky lives in europe..you can get Papst fans out of gum ball machines there!

 

[Hanky]

lifeguard1999,

I would never doubt the FACT &quot;A difference of 3 dB for the relative sound level S means twice the sound intensity I&quot;, because you can easily prove it:

S_1:=10lg(I/I_0) dB
S_2:=10lg(2I/I_0) dB
=> S_2-S_1 = 10lg(2) dB = 10ln(2)/ln(10) dB = 3.0103 dB (approx.)

But we're not talking about dB's here but db(A)! There we have additional correction-terms which take the frequency-dependent sensitivity of the human ear into account. For example means x dB(A) a rated sound level of x dB according to IEC-curve A (there are more, maybe also B,C,...). But that means to me if we have let's say two sound levels S_1=x dB and S_2=y dB with x dB<>y dB and while there is a difference for the dB-scale it could be, it's no difference for the human ear meaning x dB(A) is equivalent to y dB(A)...or not?
So I have to doubt whether you can handle dB(A) just like dB...