[tlug] Noise and Power (was Re: Bitsa Woes)

[jep200404@example.com]

On Wed, 14 Sep 2011 00:25:53 +0900, CL <az.4tlug@example.com> wrote:

> This past weekend, I decided to create a new torrenting machine out of 
> the bits left over from my recent desktop machine overhaul which has 
> provided me four more cores with which to produce meaningless economic 
> calculations and multiply simple programming errors at over twice my 
> previous speed.  But, instead of a quiet data monster, I have a roaring 
> dragon sitting in the corner and it is probably the noisiest box I have 
> ever built in all of my years of hammering hardware together.  It can be 
> heard in all rooms and both floors of the house.  Color me unhappy.

That's why I saved a Dell GX260 for server use. It uses about 
35 Watts. It has a big fan which is on rubber mounts and turns 
slowly because 35 Watts of heat does not require air to get 
rid of. Sitting within arm's length, I can hear it, but some 
other people have difficulty hearing it. 

For things that runs continuously, I consider the cost of power. 

At ¥1/kWh, one continuous Watt costs ¥8.76/year. 
At ¥25/kWh, one continuous Watt costs ¥219/year. 

At ¥25/kWh the electricity for a computer that continuously 
draws 100 Watts is ¥21900/year. 

Things that use less power also need less cooling. 
I have measured a SheevaPlug as pulling 3 Watts 
and it has no fans, so it is _silent_. 
You can recover the cost of a SheevaPlug in less than a 
year because it uses so much less power than your noisy beast. 
At ¥25/kWh the electricity for a computer that continuously 
draws 3 Watts is ¥26/year. 

There are plenty of other silent low power computers. 
Check out the Beagle Boards. They use more power than a
SheevaPlug, but are remarkably powerful. 

Also consider old laptops. 

Also consider something like the Asus Transformer. 
It even comes with its own UPS. 

> The main culprits are the fans in the video card and the CPU fan, both 
> supposedly rated at a max 30Db.  I'd guess the real noise output is 
> close to 80Db (but all of my noise measuring experience is motorcycle 
> exhaust noise measurements for SAE and those produce a much lower bass 
> note than a PC fan).

Generally, big slow fans are less noisy than small fast fans. 
Use bigger heatsinks with big slow fans to reduce the noise. 
Use really big heatsinks or big remote heatsinks (with heat pipes 
or liquid cooling) without fans for _silent_ cooling. 

> The power supply is being asked to power:  The mobo, 8Gb of DDR2 memory, 
> and 2-core CPU, two 120mm case fans, CPU fan, video card and fan, 3 
> internal SATA HDDs (2.0, 1.5, and 1.0Tb), two ISA CD/DVD R/Ws, a Blu-Ray 
> player, and five remote USB SATA HDDs totaling 6.0Tb (hey, it's a 
> torrenting and movie machine). 

My wild guess is that your box is pulling between 75 Watts and 150 Watts 
(if those USB drives are powered only through the USB bus). 

> It may not be up to the task, but what is?
> 
> I used a 400W PS because I had one, PC Depot is 21 km away, and I live 
> in Ibaraki-ken where no one sane drives on Sunday because of all of the 
> lost tourists and half-blind local baa-chans and jii-chans running loose 
> and piling into Jusco. 

The peak power draw (not average) is what you have to get 
a power supply for. You 400 Watt power supply is almost 
certainly more than sufficient. That it works is a clue. 

> A friend who is an EE suggested that the brand new 400W power supply 
> (rated at max 27Db and probably operating well below that) I'd had in 
> the closet is probably the culprit and says things will quiet down "a 
> whole lot" if I install a power supply with a higher output -- he 
> suggests a minimum of 650W.  Does this sound less counter intuitive to 
> the collective wisdom than it does to me?

_Some_ better quality power supplies are quieter, usually by 
using bigger slower fans that only go as fast as they need to. 
The better quality power supplies usually cost more and often 
have a higher max power rating. 

> Does it seem logical that more watts will equal many fewer Db? 

Power is a poor proxy for noise. 

> If so, how many more watts seem logical?  I tried adding 
> the peak rated wattage of all devices, but that didn't work right.  Is 
> there some rule of thumb or seat of pants way to come up with a required 
> power number?

Compare the noise of a Lexus with the noise of a Tercel at 
either similar speed or power. Compare the noise of a Tercel 
with a 1970s Land Cruiser at either similar speed or power.