Kiwi Quiet P4 Cooling

blog image

Peter Scott writes from New Zealand about his efforts to quiet a noisy P4 2.66A box. His article is a saga of experimentation and inventiveness, which seem requisite for a successful PC silencing project. Peter also shows us another variant of the HDD anti-vibration decoupling suspension technique we espouse. This article will inspire those for whom the latest “silent” gear is not available.

December 20, 2004 by Peter Scott

Peter Scott is a computer consultant who works with computers much of the day in a quiet home office environment, an eco-inspired earth house with amazing acoustics. He writes from Auckland, New Zealand about his efforts to quiet a noisy P4 2.66A box. His article is another saga of experimentation and inventiveness that seem to be requisite for a successful PC silencing project. Peter also shows us yet another variant of the HDD anti-vibration decoupling suspension technique we espouse; he would probably say it’s part and parcel of Kiwi make-do ingenuity. It’s also reminiscent of the first year of SPCR when DIY project article submissions were the norm, not the exception. I expect this article will inspire those who live in places where the latest "silent" gear is not available.

Mike Chin, Editor / Publisher of SPCR

I approached the task of quieting this noisy office PC by tackling
each of the four noise sources in turn. It turned
into quite an incremental approach, chipping a bit of noise off here and bit
there, until no one source really stood out and the overall noise level was
very tame. It took several weeks, surprisingly.

Firstly I ordered an all copper heatsink, the Thermaltake Golf 325, which
was available locally for a reasonable price. My criteria for the CPU cooler
were:

  • must be able to fit 80mm fan,
  • no more than NZ$50 and
  • be efficient enough
    to run on a slow fan.
  • It’s shown below complete with some homemade rubber fan
    mounts.

I ditched the supplied Thermaltake Smartfan
because it has a real growl of a rumble at any speed. I guess its strength would be in overclocking, etc., as it spins up to
4500 rpm. I couldn’t locate any NZ sources of either Papst or Panaflo L1 fans,
but did track down a supply of 80mm SilenX thermistor controlled fans. These
are quiet, and seem to have good build quality, but more about these later.

So with the parts arrived, I gutted the box, and set about improving the ventilation
into this particular case. Something you might know about NZers is that we
do things on a shoestring, using bits of whatever we have around. The case
has survived a number of upgrades, and, due to its vintage, was in serious need
for some more air flow. I enlarged the lower front air holes, and created
air holes at the upper rear where none previously existed. In hindsight I
should have cut these out completely and used wire grids, but I was being
cautious about RF pollution. Note the soft silicone rubber mounts, and sundry
other case holes taped over to preserve the thermal chimney effect.

As for the PSU, I couldn’t locate a NZ source of Seasonic or
Nexus gear, and thought I’d have a go at modifying the existing PSU. After taking
the top off the PSU and fitting one of the SilenX fans using the supplied
silicone mounts, I figured the thermistor should go near the heatsinks but
out of the main airflow. After firing the modified PSU up, initially it ran quietly but
the new fan quickly rev’d up due to the inherently warm
environment of a 65% (in)efficient switching PSU. 

For all the manufacturers
raves about silence, the SilenX fan certainly didn’t sound too silent to me. It was only
after reading some more that it dawned upon me that so called silent fans
aren’t really that much more quiet for a given airflow, they just
tend to run slower and hence more quietly. (See graph at  www.cpemma.co.uk)
The SilenX 80mm starts off at about 1400 rpm and tops out at around 2400 rpm.
So I sorted through my old fan collection (which comprised mostly of a bunch
of plain old garden variety case fans) and basically tested each one for a)
starting reliably at 5v, b) absence of vibration and “rumble”, and c) decent
airflow at 12V.

After concluding that the SilenX themistor wasn’t really much use, with a quick
snip it was lopped off. I determined with a multimeter that it has about
11K ohms resistance when cold, and about 1K when hot. So I soldered on a 10K resistor and
— lo! — a permanently silent, albeit low airflow fan.

Unsatisfied by this lack
of control, I found a simple fan controller circuit on the web (www.cpemma.co.uk/ef.html)
which has only 5 parts. I threw that little circuit together, and mounted
it on an empty PCI slot hatch. Our local electronic parts retailer had these
rows of header pins which when you snip a pair off and solder them onto the
circuit board, fit most fan socket connectors quite nicely. Here’s the finished
result.

The fan voltage varies between 5V and 11.3V, and it will cope with
40W of fans! The cable pair at left is the incoming 12V power which comes
from a Molex adapter.

So now with manual control of both case and PSU fans, I decided on another
approach for the CPU fan. Again, after removing the thermistor I replaced it with a 20K 10-turn variable resistor. This approach seemed to me to have
fewer "moving" parts. You could also use a normal potentiometer and
mount it on the PCI hatch as well.  By the way by shorting out the
thermistor I discovered that the SilenX will run up to 3500rpm with good
airflow if you ever needed it. It also runs very slowly with the thermistor
cable open circuit, maybe about 500 rpm. I doubt it would start reliably.

Initially I fitted the supplied steel fan grill on the CPU cooler fan out of sheer ignorance,
but when I read about the noise that obstructions make especially on the intake
side, I tried not using it and it made a large noise reduction.  With tidy
cables the grill isn’t even necessary, and that’s what made me finally realize that
the mesh of holes I’d drilled for the case fans also weren’t that great a solution.

For fan speed and temperature measurement, people talk about Motherboard Monitor 5,
but after discovering that a) its stopped development and b) doesn’t support
my motherboard, I found Speedfan ( www.almico.com/speedfan.php )

So after I had tamed the fans, the fan noise could be got really really
quiet. I accepted that some parts would run hotter. This was particularly true for the PSU, but I justified it on the grounds that it is designed to
cope with 300W and I was only using maybe 180W. However after seeing the heat
that was constantly pouring out the back of my PSU, (a cheap replacement
I got a year or two back), I ordered a 300W Acbel product that had a
specified noise level below 23dB but also had active power factor correction,
an 80% efficiency, and a lifetime warranty. Not bad for NZ$50. It runs much
cooler, although I haven’t yet got the courage up to take the lid off and void
that great warranty. It has some sort of fan controller inside anyhow.

My PC is mainly an office machine, pretty much in idle
mode mostly. On occasion I get it to do something
chewy like video encoding or large file copies. So while I could turn all
the fans right down, this tended to mean the temperatures tended to rocket
to over 50 degrees when under load. It felt important to me to keep that CPU
from overheating, so I devised a compromise. By setting the CPU
fan to 2300 rpm, and the other fans at 1400 rpm, the CPU idles at around 29-33°C, and gets up to about 45 °C after
a 1GB filecopy at 100% CPU utilization. I found that with no intake
grill, rubber isolation mounts for the SilenX fan (very well balanced and has no rumble) on the CPU heatsink, I could run this fan quite fast
and not affect the overall noise level.


Here’s the Speedfan graph of my day
to day office type activity

(Those temperatures are in degrees Celsius). As a side topic for the
sake of experimentation I was curious how good the CPU cooler was without
the fan. This is what happened when I shut the fan down:


CPU temperature with the cooler fan turned off.

For the first couple of minutes it did nothing, then started
rising steadily. After 10 minutes I chickened out and restarted the fan. The
upward trend didn’t look healthy, with no obvious equilibrium
temperature this side of the melting point of silicon. But I wanted to know,
and now I do.

With the fans taken care of for now, it became glaringly obvious
that the hard drive, which I had never noticed before in the shade of
that nasty stock Intel cooler, was now the loudest component.
So I uncharacteristically splashed out on a new drive, an 80GB Samsung. I
um-ed and ah-ed over the Samsung v. Seagate dilemma and gave in to the Samsung
with its superior quietness and three year warranty. I’ve not had
that positive an experience with Seagate’s warranty handling in the past.

I got this thing home (it was the only part that didn’t have to come
by aeroplane) and fired it up with a fresh Windows reinstall. Yes, the drive is real quiet alright. But what it lacks in noise it makes up for
in quite a fearsome vibration. I don’t know whether this is standard for all
such 7200 rpm drives, for this model or just a quirk of this particular unit,
but left in the drive tray, or on my wooden floor, or on top of the case, it
instantly transferred this low frequency hum into the room. As it happened
I had been shopping around for some acoustic padding on the off chance that
I might want to line the case. All I could find
was this neoprene like stuff from the plastics store ( they call it yoga mat)
which is about 4 mm thick, and quite elastic. I put a couple of layers of
this under the drive and the vibration even penetrated that somewhat. So then
I made a sort of sling, popped the drive in it, and that finally put the horrid
hum to rest.


HDD anti-vibration sling.

So it seemed clear to me that the only solution to that hum with
the materials I had was to suspend the drive. I did wonder about sorbothane but it’s hard
to get here and expensive. After a bit of tinkering
I came up with something else.


A better HDD anti-vibration solution.

Its a solution that compensated some for the loss of cooling from the case
contact, kept the drive earthed and yet also suspended securely enough to
prevent damage. I wanted the drive down low out of the worst of the case heat, near the cool incoming air and the enlarged intake
holes. I added copper heatsinks fashioned out of some 0.5mm sheet I had lying
around, and made the brackets out of stainless strap about 1mm thick by 25mm
wide.  Its amazing how the use of the 25mm wide yoga mat material constrains
the back and forward motion almost completely, as opposed to the rubber bungy
cord approach seen elsewhere. The blue yoga mat stuff also has some sort of
internal reinforcing so while its elastic in its depth, the length and width
are not, hence it suspends well without giving.

Now that I’ve finished this thing I am left with the little doubt that maybe its
massive overkill, and I shouldn’t I be wondering whether a drive ought
to vibrate so? It must have quite an unbalanced rotor to do that. Anyhow I’ve
heard others lament the Samsung vibration and so the cradle stays. I figure
the added heatsink increases the surface area of the drive three fold, and
in practice the drive temp seems to settle at around the case temperature.

Here’s a drawing: (dimensions in cm). The metal “washers” that detain the 25mm wide
strips of yoga mat are just 1cm offcuts of the bracket material. The thing
is secured using standard case screws tapped into the bed of the case.

And the finished system.

Now with all the pieces put together, the main noise is a whooshy
sort of sound that is only audible when my office is quiet. As my office is
actually rather often quiet I can still hear those fans and still dream of
total silence. I read up on some recording studio acoustics and they recommend
placing soft padding on nearby hard surfaces to help absorb the whoosh of
the fans. I found some polyester house insulation and stuck some on the wall
behind the box and on the under side of my desk where the box sits. It did
help some, but didn’t exactly help heat dissipation either.

And while I didn’t notice it in the beginning the drive does have a high frequency squeal, I’d say around 18KHz
or so. Most people I know wouldn’t hear it all; I’ve got really sensitive hearing. 

Lastly it’s worth mentioning that once the case lid went on and the
machine was buried under the desk, all the temps went back up a few degrees. Here’s a plot showing opening the case about 85%
of the way along the plot. The motherboard temp dropped a degree and the CPU temp
dropped five degrees. I guess in my chain of heating, the CPU gets prewarmed
air from the HDD and across the motherboard. 

Any reservations? Now that the machine is so quiet, I’m starting to miss those
familiar audible signs of my productivity. Its a bit eery really, and I especially
miss that familiar somehow comforting seek noise of the hard drive.

In future I’d do a couple more things:

  • replace the case fan outlets with wire grills
  • drill a whole lot more holes in the front of the case to let more air in.
  • make a mark 2 fan speed controller, which has two speeds controlled by an adjustable temperature sensor threshold.
  • look at a more upright, heatpipe served CPU cooler with a more direct heat path out of the case,
  • look at creating two completely separate airflow paths for the HDD / PSU and
    the motherboard and CPU.
  • * * *

    Complete Parts List (prices in NZ$)
    ———————————————————————-
    Case: unknown plain vanilla mid tower ATX case (recycled) $0
    MoB: P4V533-MX (second hand) w/CPU…
    CPU: Intel P4 2.66A Northwood (second hand) $280
    VGA: on board
    RAM: ADATA 256MHz DDR400 $70
    HDD: Samsung Spinpoint SP0802N $85
    PSU: Acbel 300W API3PC49 $50
    Optical: Liteon Combo SOHC-5232K $70
    CPU cooler: Thermaltake Golf 325 $50
    CPU fan: SilenX 80mm fan $26
    Case fan: Titan 80mm $8
    Speed control: DIY $8
    HDD suspension: DIY $5
    ———————————————————————
    Total NZ$652, say about US$400

    * * *

    Discuss this article in the SPCR forums.

    Leave a Comment

    Your email address will not be published. Required fields are marked *