Wesley's Tool-Box

• Installation and Benchmarking

Now it was time for the real test; to hook the air conditioner up to my system and observe temperature differences. System specification is as follows:

CPU:	AMD T-bird 1GHz @ 1.12GHz
M/B:	Soltek SL-75KAV (KT133A Chipset)
Heatsink:	Zalman CNPS3100-Gold
RAM:	256MB x 1 Corsair PC-150 SDRAM
Video:	LEOTEC GeForce2 MX 32MB
HDD:	IBM DeskStar 75GXP 30GB
OS:	Windows 2000 SP2
Else:	Classified ^_^

Tests were done in a stable ambient temperature environment for 40 minutes and MBM5 was used for temperature detection and logging. CPU was at 100% load during all times of the test using distributed.net client.

• Conclusions

Providing cold air to the intake fan of the system, the air conditioner should be lowering the system components overall. The result shows that it does indeed do that, as it lowered the CPU temperature by approximately 3C, which correlates to the dry run results. Although not shown here, the GF2 MX's chipset core experienced nearly identical temperature drop also. I believe the project has been a success! I think I can last this summer without much worries now.

• Designing and Making Housing

The housing was to be made in clear acrylic material like the previous project. To accurately create a housing for the air conditioner, I used a drafting program to create parts and virtually assemble them to see if there were any errors before actual work. My skills at using these programs improved since the last project mostly due to the fact that I'm taking classes about them in the university. Seeing that there weren't any problems to be found, I went back to my home town (Ulsan) to ask for the skilled help of the person who made the case for my Portable Athlon. Without much fiddling, he was able to exactly reproduce my draft into reality. Coincidentally, while this was being made, on the other side of the town, Confederations Cup 2001 preliminaries were being held at newly built Munsu Soccer Stadium. Korea won over Mexico 1-0 that night.

• Dry Run

The final product meatures 15.0 x 15.5 x 17.0 (cm). This fits in a small totebag, and it is quite possibly yet another 'portable' unit. With everything in order, I set out to find optimal operation point of this unit. Too fast an airflow, the cold side would not be able to put out cold enough air because of insufficient cooling time; too slow and the hot side would overheat. I found out that using 5V for the fan was better than 12V in this sense. Temperature drop against ambient was measured between 2.5C and 3.5C during the tests. Not spectacular, but when a few degrees of drop in temperature matters a lot in overclocking, it certainly didn't look so bad at all.

 

• Proto-Assembly and Testing

With key parts now in hand, I could now put them together for some preliminary testing. I was not yet certain if this could actually work, and the lab people did voice some concern about the possible failure, so I needed to see some solid proof. So I put the TEC in between the heatsinks and powered up.

The test went rather satisfactorily. The heatsinks distictly heated up or cooled down, and air nearby was affected by it. Unfortunately, it did not have a housing to put them together yet, and it sat around like this for a week.

• It Could Just Cool Air

I certainly wouldn't try the experiment again on my T-bird Athlon in my Portable Athlon unit for two main reasons, and that has to do with the TEC I have in question. Its power rating is 72W. A TEC can only pump heat up to about half of the power rating effectively, meaning the cold side has to have a load of less than 36W. A T-bird easily tops 50W in maximum power consumption, and that's one of the reason why it's not fit for use here. The other reason is that I have 150W power supply unit in the system, and it is already pretty much maxed out. You don't see a T-bird running on a 150W PSU every day anyhow. It has nowhere near the 72W power left to power the TEC.

Therefore, I decided to use the TEC to cool the air itself, in a fashion not so differently from an air conditioner. Thus was born the TEC air conditioning project. The parts I needed now: two massive heatsinks, some fans to blow air over the heatsinks, and a power supply to power both the TEC and the fans. I bought yet another 150W PSU to use for this project as you can see below. I did not want those big, conventional units, and I didn't want a long one like last time. So I got a microATX version. If you see closely, I attached a small switch on the motherboard connector so I could turn it on and off.

• Massive Flower Power!

As for heatsinks and fan, I needed not look further than Zalman Tech's heatsink laboratory. The people there are quite eager to help me out, so I asked if I could have some heatsinks for the air conditioning project. They told me that I could not use a conventional flower heatsink they make, because the end of the fins does not cover the 40x40(mm) surface area of the TEC fully and would not effectively cool it. They suggested making a modified cooler that would cover all that. Of course, it would need far more fins than usual. Two to three times more, in fact.

In the end, two massive heatsinks resulted, one in copper, one in aluminium. The fins spread out in near half-circle; the copper one has 135 fins, and aluminium, 145. The only downside to this is that the fins are too tightly packed together and might block good airflow. Nevertheless, it seemed good enough for the project, not to mention it's very eye-catching.

• Summer + Overclock = Bad

Seasonal changes can be an annoyance to overclockers. Winter can provide exceptional overclocking environment, then summer comes up and you can't overclock so much anymore. Korea has vastly distinct seasonal characteristics, and while it may be easily -10C in winter, 30C or higher is common in a summer day. The day I'm writing this article was no exception.

It is quite important to keep room temperature(a.k.a. ambient temperature) low in cooling and overclocking, because a normal cooler is directly dependent on it. It uses air to exchange heat, so it cannot cool down to something below the air's temperature. However, the room temperature is easily affected by the season, especially if you're living in a place where there is no direct means of temperature control. So a hotter season will have detrimental effects on an overclocked computer.

One may resort to opening the computer case and direct a huge fan towards it, but alas, that is out of option for me, as I have this Portable Athlon which cannot be opened while in use. I might be able to cool myself down, but it would have no affect on my unit. Furthermore, I currently reside in a dorm room, so I cannot place an air conditioner. So my system has to take the summer heat directly... unless I do something about it.

• TEC to the Rescue

Then it came to my mind that I had a TEC(Thermo-Electric Cooler) lying around in my toolbox. TEC is a small block of ceramic which has two types of semiconductors sandwiched in it so that it transfers heat from one side to the other side very quickly when electricity is supplied to it. This heat transfer effect is known as 'Peltier effect', and so you may have seen the TEC commonly referred to as a 'Peltier'. I had once tried to use it on an overclocking experiment, but it failed due to insufficient cooling on the hot side. The CPU involved in the experiment almost burned out in the process, and the TEC, along with the project, was abandoned. You can see the photo from that experiment below.

If that experiment taught me anything, modern day CPUs are too hot for a TEC-on-cooler to cool sufficiently. So now what?