A Preliminary Study on Air Exhaust within Confined Spaces

esplin2966

Limp Gawd
Joined
Jan 18, 2015
Messages
215
Introduction:
I am currently at the prototype stage of my PC Case Design. While negotiating with manufacturers, I decided to spend my spare time running various temperature experiments to better prepare me for the prototype testing. The present study attempts to characterize how the positioning between intake and exhaust holes impact cooling performance within confined spaces.

While the testing process gave insight to the performance of certain intake-exhaust configurations, some of the test results are puzzling. I hope that by posting my test results online, we can arrive at a plausible explanation through constructive discussion. Do not hesitate to suggest conjectures for me to test.


Test Setup and Methodology:

Test Configuration:

CPU: Intel i5-4590S
Cooler: Jonsbo HP-400 ZONE
Motherboard: Asrock Z97e-itx/ac
Memory: G.Skill Ripjaws X Series 8GB (2 x 4GB) DDR3-1600 Memory
Storage: Transcend MTS400 256GB M.2-2242 Solid State Drive
Case: Mini-Box M350 HTPC Case
Power Supply: picoPSU-16​0-XT + 192​W Adapter ​Power Kit

This test configuration is chosen for convenience. It is my main PC. The interior configuration is shown below:

EiZebE0.jpg


The CPU cooler is setup to intake air. A duct made from cardboard is placed around the intake fan to prevent hot air recirculation.

Temperature Assessment Methodology:

To characterize thermal performance, I subject the CPU to 10 minutes of load and record the temperature of each processor core in 1 second intervals. I then assess the maximum temperature, total average temperature, average temperature of the second half of data, total standard deviation, and standard deviation of the second half of data. I also show a plot of temperature vs time for each core. The load is applied by running a customized setting of Prime95, shown below:

fTzrHwM.png


The program I use to record the temperatures is RealTemp. After every 10 minutes of load testing, I let the computer sit idle by keeping nothing but the desktop open on the computer for 10 minutes before starting the next load test.

Methodology:

On this computer case, there are 5 faces that contain exhaust holes, labeled in the 2 pictures below:

BU4iilP.jpg

hkXTENu.jpg


To control exhaust, I use tape to cover up the holes on different faces. A total of 5 configurations were tested:

Configuration 1 - All Exhaust:

Faces 1 and 5 are taped. Faces 2, 3, and 4 are open. This configuration guarantees that cold air intakes only from the top face and that hot air only escapes from horizontal holes.

Configuration 2 - Near Exhaust:

Faces 1, 3, 4, and 5 are taped. Face 2 is open. This configuration guarantees that cold air intakes only from the top face and hot air only escapes from the horizontal hole closest to the intake.

Configuration 3 - Far Exhaust:

Faces 1, 2, 4, and 5 are taped. Face 3 is open. This configuration guarantees that cold air intakes only from the top face and hot air only escapes from the horizontal hole farthest from the intake.

Configuration 4 - Obstructed Exhaust:

Faces 1, 2, 3, and 5 are taped. Face 4 is open. This configuration guarantees that cold air intakes only from the top face and hot air only escapes from the horizontal hole at the front. In this case, the hot air needs to flow past the two RAM sticks to reach the exhaust.

Configuration 5 - No Exhaust:

Faces 1, 2, 3, 4, and 5 are taped. No face is open. This configuration guarantees that cold air intakes only from the top face and hot air can only escape from small holes in the I/O shield.

The load test is applied in 3 sets, with each set taking place on a different day. In every set, each configuration is tested one after the other. The order of configurations tested is different for each set.


Results:

Set 1:
Configuration 1:
a6Dok9T.png

h2NwmHh.png


Configuration 2:
vpjfCRZ.png

oRBMD4k.png


Configuration 3:
WQ8GIPm.png

VwKBUWT.png


Configuration 4:
gGmsjSM.png

LUQGTiD.png


Configuration 5:
CP3yCH1.png

2BOO6M2.png
Set 2:
Configuration 1:
pIrTgA1.png

iqqabQz.png


Configuration 2:
dcnMNwR.png

ohSQb9a.png


Configuration 3:
PNaVBau.png

7dbcjbs.png


Configuration 4:
ozENIQZ.png

nSdkWft.png


Configuration 5:
mMn9nRd.png

XlGlomX.png
Set 3:
Configuration 1:
9IcPlrH.png

MQJAgwy.png


Configuration 2:
eKim1i3.png

nbBRfSV.png


Configuration 3:
WhHWFhE.png

CVktJpT.png


Configuration 4:
GqqeOKr.png

X5mxnbF.png


Configuration 5:
bE0gOA4.png

EVC9ROF.png
Discussion and Conclusion:

For comparison, we evaluate how much better each configuration performed when compared to the average performance of all configurations.

For example, let's say we want to quantify how well configuration 1 performed relative to other configurations for the metric of maximum temperature in set 1. We first averaged the maximum temperature of all configurations separately for each core. We then take the maximum temperature of configuration 1 and subtract the averaged maximum temperature of all configurations for each core. Finally, we average the resulting values across all cores.

This process is repeated for all configurations and all performance metrics for each set, resulting in the 3 tables shown below:

gEtpvQP.png


AnmL6fj.png


eLX8oYD.png


The 3 sets are then averaged to produce the table below:

VbXa5Fp.png


From the table above, we can immediately observe that configuration 5 performed the worst in terms of maximum temperature, total average temperature, and average temperature of the second half.

We can also immediately observe that configurations 3 and 4 performed the best in terms of maximum temperature, total average temperature, and average temperature of the second half.

The performance of configurations 1 and 2 lie in the middle, with configuration 1 performing better than configuration 2 in terms of maximum temperature, total average temperature, and average temperature of the second half.

No configurations had consistently higher/lower standard deviation than others.

From these observations, we can arrive at a few conclusions:

1) Having very little exhaust holes lead to worse cooling performances.

2) Above a certain amount, having more exhaust holes does not necessarily produce better performance. In fact, it is the placement of exhaust holes that makes a difference. To be more specific, having exhaust holes near the fan but nowhere else produces poor thermal performance. Having exhaust holes far away from the fan but nowhere else produces good thermal performance.

Unfortunately, I cannot find an intuitive explanation for conclusion 2). I welcome the reader of this study to pose conjectures that may explain my results.

Thank you for reading!
 
did you account for changes in ambient temperature? Did the room AC kick on ever during this test or allow the room to warm? Just possible sources of confounding since the temp differences were rather small anyways.
 
EMPIRICAL DATA!

begone heretic, forsooth, we should burn you at the stake
 
What is it about the second conclusion that you don't find intuitive?

Above a certain amount, having more exhaust holes does not necessarily produce better performance.

The fan only generates so much flow pressure. Once the number of holes is sufficiently large that boundary layer flow at the hole edges is not the major pressure constraint on the velocity of flow, the velocity of the air exiting the holes drops which further reduces the boundary layer constraint on flow velocity.

Below a particular flow velocity for a particular system combination (overall component location will really begin to matter), you'll start losing efficiency in heat transfer as the air has been heated by previous components in the flow path, reducing the temperature differential and the amount of heat absorbed from the further components farther down the flow path over time.

In fact, it is the placement of exhaust holes that makes a difference. To be more specific, having exhaust holes near the fan but nowhere else produces poor thermal performance. Having exhaust holes far away from the fan but nowhere else produces good thermal performance.

The air being pulled into the fan has more intermingling with the overall room temperature air when exhausted further away from the fan, reducing the temperature of the input airflow. If the holes are close enough to the fan, it is more like having a closed loop. Less intermingling occurs with the surrounding air, once again reducing the temperature differential between the air flowing over components and the temperature of the components.

To get a really good picture of what is happening, you should at least have temperature measurements of the air immediately before flowing into the fan, the temperature of the air at several exhaust points, and the general room temperature.
 
did you account for changes in ambient temperature? Did the room AC kick on ever during this test or allow the room to warm? Just possible sources of confounding since the temp differences were rather small anyways.

The AC does kick in during the tests, but that's why I did the 3 sets of tests in different orders each time to try and negate this source of error. You're right that the temp differences are small, but the difference between the various configurations are fairly consistently present.
 
I agree with nessus, the reason why observation 2 is correct is the same why your duct increases cooling performance. With holes farther away from the intake you make sure that as little air as possible is recycled, and if it is, it has time to cool down on its way to the intake.
 
Thank you for sharing your work and your experiment.

To get a really good picture of what is happening, you should at least have temperature measurements of the air immediately before flowing into the fan, the temperature of the air at several exhaust points, and the general room temperature.

I agree with others that far exhaust is preventing recirculation and thereby improving thermal performance. The largest factor in heatsink performance is temperature differential.

I am also curious why you are looking only at cpu cooling performance? I think it would be smart to also watch motherboard and component temperature as well as internal case temperature.

Interesting stuff! Thanks for putting it out there.
 
I agree with nessus, the reason why observation 2 is correct is the same why your duct increases cooling performance. With holes farther away from the intake you make sure that as little air as possible is recycled, and if it is, it has time to cool down on its way to the intake.

I am running some follow up tests to check this right now. Thanks!

Thank you for sharing your work and your experiment.



I agree with others that far exhaust is preventing recirculation and thereby improving thermal performance. The largest factor in heatsink performance is temperature differential.

I am also curious why you are looking only at cpu cooling performance? I think it would be smart to also watch motherboard and component temperature as well as internal case temperature.

Interesting stuff! Thanks for putting it out there.

To be completely honest, I only looked at CPU cooling performance and not the cooling performance of the other internal components because I overlooked it. :(
 
Back
Top