The Effectiveness of Air Ducts in CPU Cooling

Summary: This article demonstrates how an external air duct can produce CPU temperatures which are as little as 7C above room temperature, and -1C to +9C above case temperature. This is done by cutting a hole in the side panel of an ATX case and blowing air directly over the heatsink, via a short duct. An average aluminum heatsink is used.

Contents:

Goal
Hypotheses
Story
Procedure
Results
Evaluation of Hypotheses
Notes
Questions
Update July 24
How to tell if there is a pressure imbalance in your case
Photos

External Air Duct Project

Goal:

Draw air from outside the case and expel it directly onto the CPU heatsink, producing a cooling effect.

Hypotheses:

  • Air outside the case is colder than air inside the case (Tcase).
  • External air expelled over the heatsink will produce lower temperatures.
  • Providing the most direct route from outside the case to the heatsink will allow for greater airflow.
  • Gradual compression (tapering) of airflow from 80mm to 60mm will maintain air velocity (CFM).
  • Expelling air from duct at a distance from heatsink will allow a greater airflow than extending the duct directly into the heatsink.

Story

It was a hot and muggy June day up here in Toronto when I decided to do
something drastic to my PC – to bring air from outside the case directly over the heatsink.

See, my Chrome Orb had been cooling my Duron 800 quite well until the
heat wave hit, and suddenly Motherboard Monitor was reporting 46 degrees Celsius CPU temperature. Having read the articles about how AMD motherboards could grossly miscalculate the temperature, I was worried that the temp could be much higher than MBM reported.

Procedure

  1. Cut a hole in side panel of case corresponding to the CPU area of the
    motherboard.

  2. Mount an 80 mm fan over the hole.
  3. Use vinyl grommets and long machine screws to attach the 80mm to the outside of the case – grommets minimize vibration noise.
  4. Build a cardboard duct that tapers the 80mm airflow into a 60mm stream
    directly over the heatsink.

  5. Place a large bare passive heatsink on the CPU.

I should tell you that I don’t have a Dremel, so I was forced to cut a hole in the metal with a wood bit, and then pry the thing apart with pliers! I did manage to make a respectable circular hole, and I even drilled screw holes for all four corners. Simply cutting an 80mms square in the side panel is NOT the way to go, as you will have no way to mount the fan. I figured out the proper placement and size of the hole with a ruler and a pencil – it’s not very difficult.

Note that the cardboard duct gradually tapers from 80mm to 60mm in size, thus concentrating the air stream. Unlike small plastic fan adapters, this duct appears to allow the fan to work close to its full CFM rating.

Also note that the duct leaves a few inches clearance from the heatsink, so the air rushes over the sink with maximum velocity. If I had built the duct to actually fit over the sink (rather than hover a few inches away), the airflow would be greatly reduced.

Results of New Heatsink and Ducting

Ambient temperature (the temperature in my bedroom) for all tests was around 28C (82F). This is unseasonably hot for June in Toronto.

Initial results with Chrome Orb (thermal pad):

  • Case 36C / CPU 46C at 800Mhz, default voltage IDLE (no load test done)

Final Results with ductwork and Evercool ND3 heatsink with generic thermal
grease:

  • Case 34C / CPU 35C at 800Mhz default voltage IDLE
  • Case 35C / CPU 39C at 800Mhz default voltage LOAD
  • Case 35C / CPU 39C at 900Mhz max voltage IDLE
  • Case 35C / CPU 44C at 900Mhz max voltage LOAD

The results speak for themselves: only 1C above Tcase (case temp) when idle at 800MHz. As you can see, even at the max voltage supplied by the K7Tpro2a (1.92v according to MBM) at 900Mhz, the CPU is still cooler under load than at idle with the Chrome Orb.

Evaluation of Hypotheses

TRUE: Air outside the case is colder than air inside the case (Tcase)
TRUE: External air expelled over the heatsink will produce lower temperatures.
TRUE: Providing the most direct route from outside the case to the heatsink will
allow for greater airflow.
TRUE: Gradual compression (tapering) of airflow from 80mm to 60mm will maintain
air velocity (CFM).
TRUE: Expelling air from duct at a distance from heatsink will allow a greater
airflow than extending the duct directly into the heatsink.

Notes:

In respect to the hypothesis, there was no control on results, as the
heatsink was changed. I did not test the cooling effect of a Chrome Orb acting as a passive heatsink (its own fan turned off) for the ductwork.

I did not scientifically test the CFM of the fan. However by feeling the airflow against my hand, I judged the ductwork to not greatly reduce airflow.

There is also the possibility of the airflow cooling the area underneath the CPU and lowering the the socket-thermistor temperature.

Questions:

  1. How much better would the Chrome Orb have been if I tried removing the
    pink thermal pad and applied grease?

  2. Is there a pressure imbalance in the case? Note that I removed the grills on my power supply to allow the PSU fan to work at maximum strength, so the PSU does expel a similar amount of air as the external fan pushes in.
  3. If I used a bare PAL 6035 or PEP 66 heatsink would I get a lower temperature?
  4. How would results change if I changed the distance of the duct opening from the heatsink?
  5. My Microstar K7T-Pro2a doesn’t overclock well. Both a Duron 700 and 750 could not go past 927MHz (103×9) at maximum voltage without generating Prime95 errors. This very same Duron 750 ran at 975 MHz (150×6.5) on the Abit KT7a of another Overclockers.com reader. Is there something wrong with my motherboard? It’s otherwise stable.

Update July 24/01

I recently received two low noise 26cfm fans from
Pham Computer. I replaced the CPU fan and the PSU fan with these low noise fans. My system is so quiet now that all I can hear is the hard drive and the sound of air rushing through the power supply. Note that I removed all grills (both internal and external grills) from the power supply, so I’m getting the full 26cfm of air flow.

Today with the temperature in my room at 34C (90F), my CPU at 700MHz, 1.67v is idling at 41C with a case temp of 42C (case temp is actually 1C higher than CPU temp. That’s a CPU temp of only 7C above ambient.

How to tell if there is an air pressure
imbalance in your case:

This only works if both fans are identical, and therefore work at the same rpm. As you know, when air pressure changes the rpm of the fans will change (due to back-pressure and whatnot). When both fans are running at the same speed, you shouldn’t hear anything odd. However, if one of the fans increases or decreases its rpm you should hear a repetitive "whir-whir-whir-whir" sound.

As the rpm of a fan changes, so does the pitch of the sound that it makes. In effect, when these two identical fans are running at different rpm,
their motor noises are out of tune with each other. I have a Bachelor of Music degree, and in my field this "whir-whir-whir" sound is called "beating".

The number of beats that you hear in one second indicates in Hz or cps (rotations/cycles per second) difference in speed between the two fans. Musicians use this beating sound to tune their instruments; for example two saxophone players will play the same note, and adjust their instruments until there is no beating heard.

Now, at first when I installed the two 26cfm fans, I was hearing about 6 beats (‘whir-whir-whir-whir-whir-whir’) per second. So one fan was running at 360 rpm faster or slower than the other (6 beats per second * 60 seconds = rpm).

This indicated to me that there was a positive pressure inside the case, so I removed some additional grills from the power supply (to allow greater airflow) and the beating stopped. Both fans are now moving the same amount of air, and there is standard pressure inside the case.

Photos:

Idle vs load at 800MHz 1.72 volts (1.65 BIOS setting).

Idle vs load at 900MHz 1.92 volts (*default* BIOS setting – my Duron has the voltage bridges penciled to maximum).

The Evercool ND3 heatsink, which I used with the ductwork. I removed the fan, of course. This heatsink is identical in height to the Chrome Orb.

Closeup of the 80mm fan pushing air into the duct:

Here’s the case under my desk with the fan stuck on the side. The position of the fan corresponds to the CPU socket on the motherboard.

Here’s the cardboard duct shown head-on The fan is outside the case, with the power wire fed in:

Here’s the cardboard duct viewed from the side. It is attached with metal
tape (the type used to seal air ducts). Remember that the fan is on the outside of the case.

Here you can see the bare ND3 heatsink:

Proof of high temperatures. Looks like about 82F in my room:

Rick Talbot
Notorious Networks

Tags:

Leave a Reply

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

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>