Chilled Convection?

[Xeniczone]

I've came up with an idea for chilled convection. Though I've never really seen anybody do something like this other then putting there computer on top of a air duct or something.

Here is the CAD image of the device.


The thermal electric Module sits in the middle of 2 large heat sinks. Surrounded by 4x 60mm fans. this will sit in the front of the case and take up the hard drive bays (then the hard drive will move to the floppy drive bay.

The left side of the device is the hot side. The 2x 60mm fans on this side will blow air on it and down and out the bottom of the case. (My case is a NZXT. Lexa so it sits a good 1.5in to 2in of the desk.)

The cold side is the left side and it blows straight on through into the computer case.

I'm new to Thermal Electric devices, and I did read the article for beginners but doesn't mention air cooling.

I plan on using a 500watt dedicated powersupply and a ~230watt thermal electric module.

My basic question is will this work?

Here are the specs of my rig.
AMD Athlon 4400+ 2.3ghz (i'd like to o/c but it gets to hot the way it is now)
2gigs DDr2 800Mhz Memory.
nVidia 8800GTX
550watt PSU (for system)
500watt PSU (for cooler)

Here is a pic I'm going to recycle from some other forums. I circled all the fans in the system.



-Xeniczone

[evil-98]

same idea here, let me see if i can find some more pics



basicly the same as your design, peltier cooled case


now refering to your design, why not get away with a single 120mm fan?
it would be more powerful and quieter. generally it isnt powerfully enough
to cool down the case significantly, maybe drop it a degree or 2 but its
worth a try. i wanted to try it, just copper sinks are quite expensive

edit:

[Xeniczone]

Interesting stuff,

I had though of using 120 instead of 4x 60's but there comes a few advantages to having the multiple fans.

For starters an average 60 will move 22CFM of air so one side will have 2x 22CFM Fans blowing into one side of it. While a average 120 will have around 65CFM so only about half of that will be going into one side of the cooler.

but I don't know, don't see why 1x 120 wouldn't work. But there are some neat things that 4x 60mm can do that a 120 can't like turn off some of the fans to reduce air flow, (though you could always slow the 120 down but w/e. Or turn half the fans off so only the hot side is getting air while the cold side can be in standby and get even colder for no warm air is moving across it etc.

Yeah Copper heat sinks are kind of expensive but I planned on using/building large aluminum ones.

[Uncle Jimbo]

Interesting stuff,

I had though of using 120 instead of 4x 60's but there comes a few advantages to having the multiple fans.

For starters an average 60 will move 22CFM of air so one side will have 2x 22CFM Fans blowing into one side of it. While a average 120 will have around 65CFM so only about half of that will be going into one side of the cooler.

but I don't know, don't see why 1x 120 wouldn't work. But there are some neat things that 4x 60mm can do that a 120 can't like turn off some of the fans to reduce air flow, (though you could always slow the 120 down but w/e. Or turn half the fans off so only the hot side is getting air while the cold side can be in standby and get even colder for no warm air is moving across it etc.

Yeah Copper heat sinks are kind of expensive but I planned on using/building large aluminum ones.

My nephew built something like this during several TEC experiments. His put the TEC at the top of the case, hot exhausted up and cool went into the case as in your design. For efficiency you want to use a number of undervolted TECs rather than one big one. A couple of things he found -
1 - routing the air internally with an insulated duct directly to the CPU cooler rather than just into the case does a better job of dropping CPU temps. You limit losses from trying to cool a larger, radiating space. Pumping about 200W, he was able to achieve -5C in a 25C ambient with 60CFM into a Titan Vanessa, holding a 265 opty at 2.8G to about 10C. But he had very large high performance sinks on the hot side - .06 C/W for each TEC
2 - the air exiting the CPU cooler will still be way under ambient and can be routed around to cool other components
3 - Cooling the whole case is not very effective. Once you get the case to ambient or maybe a degree or so under, you are done - the case will efficiently radiate whatever you pump in at that point.


He was using undervolted 12730s but based on analysis done on this forum over the last few days, you can get better efficiency with other TECs. But with air cooling you won't achieve the high CoP of WC designs.

[Xeniczone]

Thanks for all the info.

It does make more sence to pipe all the cold air to the individual components that need it.

I will work on some more CAD programs and see what I can get.

[Vinas]

Pumping about 200W, he was able to achieve -5C in a 25C ambient with 60CFM into a Titan Vanessa, holding a 265 opty at 2.8G to about 10C.

He was moving air at -5C? That would be one hell of an insulation project. The condensation was a major problem I presume? Or is that just the temp of the cold side of the TEC?

[THE JEW (RaVeN)]

Any pics Jimbo? It would be interesting to see.

[Uncle Jimbo]

He was moving air at -5C? That would be one hell of an insulation project. The condensation was a major problem I presume? Or is that just the temp of the cold side of the TEC?

Condensation is not quite the right word - the cold side sink developed a layer of frost. The box around the cold side sink (which acted as a plenum) was insulated with 1/8 neoprene foam. The -5C temp was recorded in that plenum. Outlet was an 80mm flex tube which went to a plenum mounted around the Titan Vanessa which was just a ghetto cardboard rig. There was a fair amount of condensation in that box. The airflow was maintained by the fan on the Vanessa. The air coming out of the back of that plenum was at about 5C and had tubes that ran to the air intake for the GPU, with smaller tubes running to exhaust over the RAM and other internal components. The whole rig was kind of modeled on whole house air conditioning setup. The big problems were frost buildup at the cold side and condensation in the plenum around the CPU cooler.

He was running a pair of 12730s, each with a .06 C/W sink, using a modified PWM control to maintain the cold side at -10C, so the TEC operating points moved around a little, but the TEC was at about 9.5V drawing about 18A. Hot side was about 38C in a 25C ambient. Those parameters mean each of the 2 TECs should have been pulling 60W or so.

I think he shot some pictures while it was running - I'll check and see. It was just a test so he didn't keep it together, but there is another version coming up which will use a larger number of TECs running lower power to try for improved CoP.

[Kai Robinson]

I was thinking of something like this as a general 'cold air unit', as my aircooled TEC idea has been shot down in flames...

I've got a big Thermaltake Silent Boost K8 which is enough chunky copper to get started

[Jorlen]

Ambient, when cooling something below it you need to insulate. Otherwise you are wasting energy and cooling power cooling things that dont need it.
If your goal is to maintain subambient airtemperature I see no reason in taking in ambient air and cooling it below ambient when you can reuse sub-ambient air already inside the case. You will need to seal that volume of air off and insulate it. If you chose to enclose the whole case or just a smaller box around the CPU is your choice.
But as long as air colder then ambient is leaving your case it should go back to the coldside sink. Easiest is way is to have two separate bodies of gas, outside and inside.
Condensation can be handled since you wont be adding moisture constantly to the sealed case, as long as the seal is decent you will dry the air out after a while and buildup will halt.

[Uncle Jimbo]

Ambient, when cooling something below it you need to insulate. Otherwise you are wasting energy and cooling power cooling things that dont need it.
If your goal is to maintain subambient airtemperature I see no reason in taking in ambient air and cooling it below ambient when you can reuse sub-ambient air already inside the case. You will need to seal that volume of air off and insulate it. If you chose to enclose the whole case or just a smaller box around the CPU is your choice.
But as long as air colder then ambient is leaving your case it should go back to the coldside sink. Easiest is way is to have two separate bodies of gas, outside and inside.
Condensation can be handled since you wont be adding moisture constantly to the sealed case, as long as the seal is decent you will dry the air out after a while and buildup will halt.

Interesting - that is another project my nephew is working on. He insulating the whole case with 3/8 neoprene and then recirculating the cold air. He built an external box of wood (it's a pretty good insulator) with the TEC in it, routing the air in and out with flex tubing. I know he has some photos - I'll get them and post. As with WC, if the TEC power exceeds the load, it should get colder and colder, but cooling everything in the box is a lot more heat.

That is also a lot more case modding work than just cooling the CPU though, not sure how many people would want to undertake it. I'll let you know how it goes.

[Xeniczone]

Ambient, when cooling something below it you need to insulate. Otherwise you are wasting energy and cooling power cooling things that dont need it.
If your goal is to maintain subambient airtemperature I see no reason in taking in ambient air and cooling it below ambient when you can reuse sub-ambient air already inside the case. You will need to seal that volume of air off and insulate it. If you chose to enclose the whole case or just a smaller box around the CPU is your choice.
But as long as air colder then ambient is leaving your case it should go back to the coldside sink. Easiest is way is to have two separate bodies of gas, outside and inside.
Condensation can be handled since you wont be adding moisture constantly to the sealed case, as long as the seal is decent you will dry the air out after a while and buildup will halt.

Though I do have plans to build a different computer using this idea from what your saying with the insulation and stuff I would need a whole new case set up as this one isn't fit for insulation, and because of this I think I will do a liquid cooling system instead.


If the case was well insulated and the case was releasing cold air then couldn't the cold air be routed back to the front of the case and used to help cool the hot side and cool the cold side even more?

Maybe even more complex but use 2 inputs on the front and back of the case. Both inputs with shut off valves. If the air from the back of the case is greater then ambient then the rear of the case is open and the front of the case is open to allow normal case air flow. If the exhaust is cooler then the ambient then the rear of the case and the front of the case are closed so that the air from the back of the case will follow a exterior tube and flow back into the front of the case.

[Jorlen]

Either you have overcapacity in your cooling system, which means you should use the cold air from inside the case to either cool the hotside (generally a bad idea) or you recool that air again passing it through the coldside sink again. Or you dont have that much capacity and air leaving case is warmer then ambient.
Your suggestion is that during idle you might have overcapacity and during load you dont. So you use valves to switch strategies "in flight". To me it seems that building a system that can switch from reusing cold air to not reusing is quiete complex, and I frankly dont see it having being that effective in terms of cooling vs complexity of build. And when you think about it, air is short lived heat buffer, all that switching will accomplish is very low idle temps. Load temps will work the same as having no switch. And usually load temps are what matters, right?

I dont know what budget you might have. If you state your budget and if you wanna cool the whole case or just the CPU, we can narrow down you options for you.

[Vinas]

@ Uncle Jimbo -- that's a nice setup... Very interesting way to hook things up for sure. One thing though, if he were to reuse the warm exhaust air I would think there would be less moisture in the air since the cold side already condensed the moisture out. Either that or he could use the setup as a dehumidifier

If you get some pics I'd be very interested in seeing them. This kind of thing sparks my curiosity for sure!

[Xeniczone]

Generally I'm on a low buget and I have managed to theorize a cooling system for around 200 bucks.

But I think I can spend as much as needed really. probably a max would be 350 though I really could find better ways to spend that money w/e

I'm after CPU/GPU. The CPU is 109F under load, and the graphics card is around 157 under idle. Not sure what it is under load.

[Jorlen]

Then comes a second question, what is an acceptable noise level to you?
Fans are probably the cheapest way for you to add cooling power to your system.

If you wanna use peltiers, you need to get lucky and find a lot of really cheap heatsinks. You are more likely to find 20 stock coolers for almost nothing that allows you to use 10 cheap TECs. Then to find 2 really really good heatsinks to work on 1 big TECs.

My suggestion is to mount 10 TECs and 20 heatsinks so that the hot sinks are outside the case and the cold ones are inside. Through the top of your case. Find a PSU that can run the tecs and fans. Run the TECs at about 6-8V. THe more noise you can stand from the fans the more voltage you can run the TECs at.
As I said, if you get lucky with finding 20 really cheap heatsinks this will work really well for your budget. The TECs are about $50 on ebay including shipping. You can most likely use an old AT/ATX PSU to power the TECs, since they dont need more then 100-250W.

[Xeniczone]

Then comes a second question, what is an acceptable noise level to you?
Fans are probably the cheapest way for you to add cooling power to your system.

If you wanna use peltiers, you need to get lucky and find a lot of really cheap heatsinks. You are more likely to find 20 stock coolers for almost nothing that allows you to use 10 cheap TECs. Then to find 2 really really good heatsinks to work on 1 big TECs.

My suggestion is to mount 10 TECs and 20 heatsinks so that the hot sinks are outside the case and the cold ones are inside. Through the top of your case. Find a PSU that can run the tecs and fans. Run the TECs at about 6-8V. THe more noise you can stand from the fans the more voltage you can run the TECs at.
As I said, if you get lucky with finding 20 really cheap heatsinks this will work really well for your budget. The TECs are about $50 on ebay including shipping. You can most likely use an old AT/ATX PSU to power the TECs, since they dont need more then 100-250W.

I could care less for noise. Though I am looking into liquid cooling.

Heat sinks I could probably get on ebay.
Same for the Tec's.
I have an 500 watt Power Supply in my old Computer I don't use anymore.

Would running 10x ~120watt TEC at 5 volts be good. If not I do have a Regulator that could run them at 7.5 volts?

Because if I run them at 5v then it will keep a load on both rails the 5v rail for the TEC and 12v rail for the fans.


Another idea to throw at you, Could I keep them all in the bottom of the case and run a wind tunnel through the hot side and run more fans to blow from the bottom to the top.

[Jorlen]

You can put them anywhere you like. I dont know if this still holds but I remember that you need to load the 5V line on computer PSUs to get full power on the 12V rails. You can connect the TECs in series of 2 and get 6V per TEC using the 12V line. Depending on the capacity of your heatsinks 6V might not improve cooling noticably. But it might still be a good idea to spread the load over both 5v and 12V, since thats the way they are meant to be used.
I like this idea. Its like a noisy version of my project.

[Uncle Jimbo]

You can put them anywhere you like. I dont know if this still holds but I remember that you need to load the 5V line on computer PSUs to get full power on the 12V rails. You can connect the TECs in series of 2 and get 6V per TEC using the 12V line. Depending on the capacity of your heatsinks 6V might not improve cooling noticably. But it might still be a good idea to spread the load over both 5v and 12V, since thats the way they are meant to be used.
I like this idea. Its like a noisy version of my project.

Most modern supplies don't require loading the 5V line, but if you want to use just the 12V, use a 33 Ohm 5 W resistor between 5V and ground - it will waste less than a watt and assure stable operation. If you use the 5V line, you don't need to worry about that.

There are a number of inexpensive TECs on eBay, mostly 12709s. Those are usually listed as 136W units. Those are at the low end for what you want to do but if cost is a factor, give them a try. I have seen them as low as $4 each. Those rated at 168W are 12710s. 5V is a high efficiency point for any 127 TEC, and is also commonly available at high current on any ATX supply.

I saw a current offer for 20 of these for $50 plus $20 shipping, which is pretty cheap. At 5V those will burn just over 10W per unit and pump a little over 15W across 25C, for a total heat load of 25W per unit.

Most really cheap CPU / fan heatsinks have performance of around .5C/W, some a little better. That gives you a rise of about 12C above ambient. If you move enough air through the wind tunnel, you should see a 10C or more drop below ambient on the cold side - if you can move enough heat. 12 of the commonly available TECs will move 200W. Your 500W supply may have plenty of juice to drive the load if it is older - newer ones shift more power to the 12V lines. You need 2A per unit, or 24A for 12 TECs, and older 500W PSUs typically have 35A to 40A available on 5V. Newer ones may have only 20A, so look at the max current available.

From personal experience, pumping 200W into an insulated case with the case airflow very limited can drop internal temps by 5C or more.

Going to 6V (running 2 in series off 12V and pairing those up) moves you up the power curve. Now you are at 2.5A for a total of 15W burned to move 25W. Now the same TECs move 300W, but your total heat load is 500W on the hot side, so your wind tunnel needs to move some serious air.

You are still iprobably OK with a 500W supply, since the draw is only 15A total. But be sure and check it - you don't want to use more than 70% of the listed max amperage, so you want to see at least 20A available.

Another advantage of this setup is that you can run these TECs off the MOLEX plugs, as long as you don't put more than 2 TECs on a connector. That makes for an easier setup.

Going with a bigger TEC can move a lot more heat, or use fewer TECs, or both. 12715s are under $10 plus shipping, 12726's are about $12 plus shipping, and 12730's about $20 plus shipping.

Using the 12730 units at 5V will draw 8A for power in of 40W. You move 70W for that. So just a few of those move a lot of heat. But now you are over the capacity of a MOLEX connector, and cheapie heat sinks won't do the job.

The 12715s might be a sweet spot. At 5V you draw 3.75A at 30C differential, for 19W power in, and you move 35W per unit, for total heat out of 55W. 8 of those will move 280W. The total draw is 30A, which your 500W PSU might do if it is a good one. Going to 6 units drops the heat pumped to 210W, but the current draw also drops, to a total of maybe 23A.

You probably don't want to run some at 5 and some at 6 (splitting the 12V). The main power handling stage of an ATX PSU is usually common to 3.3V, 5V and 12V so you risk overheating it if you try and draw high power from all the lines.

As an aside, you can usually adjust the internal control pot for the 5V line on those supplies to 5.5V or even 6V, depending on the clamp voltage (which can also be adjusted), That gives you a way to tweak power levels without getting too exotic. You don't increase the max amperage though.

[Xeniczone]

My 500 watt power supply is fairly new so I was a little worried while reading your post but turns out it still supplies a lot of the amps to the 5v.

+3.3V - 28.0A
+5V - 38.0A
+12V - 22.0A
-5V - 0.5A
-12V - 0.8A
+5 VSB - 2.0A

I did try to find some heat sinks on ebay and didn't find much. Costing around 10 bucks for a heat sink depending on how many I would need would end up costing 200 more dollars on heat sinks alone.

I looked up car's heat exchangers cost 20 bucks, and 50 bucks new.

but I'm not sure still trying to figure out everything. Both liquid cooling and TEC cooling are looking around the same price.

Heat exchanger: 20-50
Pump: 30-50
8800GTX Block: 60-70
AMD Block: 30-40

So maybe if I got the Pump and the Heat exchanger, then made my own reserve. That will cost around 50-100 dollars. Then need to get all the TECs.

Hmm... I'm lost... Out of ideas.

[Jorlen]

One massive heatercore or even cooling radiator from a car in the outside loop. A smaller heatercore inside the case, 2 massive waterblocks that can clamp as many TECs as possible, two pumps.
Now you have your own airconditioner.
It will be more compact then a compressor based air-con system. But performing worse and producing about the same noise.
TECs have two things going for them, size and silence. Possibly also that they are easier to use in custom cooling solutions.
Focus on their strongpoints, dont make things big and noisy, because if thats acceptable you really should use a compressor.

[Uncle Jimbo]

One massive heatercore or even cooling radiator from a car in the outside loop. A smaller heatercore inside the case, 2 massive waterblocks that can clamp as many TECs as possible, two pumps.
Now you have your own air conditioner.
It will be more compact then a compressor based air-con system. But performing worse and producing about the same noise.
TECs have two things going for them, size and silence. Possibly also that they are easier to use in custom cooling solutions.
Focus on their strong points, don't make things big and noisy, because if thats acceptable you really should use a compressor.

To expand on Jorlen's comparison, if you use WC on the hot side, and operate the TEC at around 30% of Imax, you can get efficiency (CoP) of about 2 with a differential of 10C. A good phase change system has CoP of 1.3 to 1.5, and that is regarded as pretty efficient. The advantage of the TEC system is that it requires less specialized knowledge, and you can apply the cooling in a small physical package. It can also be quieter, depending on how you do the WC. Phase change will produce a bigger temperature delta at high efficiency, so it is a better choice for very low temps.

Your PSU almost certainly has something like 5.2V at anything less than max load. Using 12730 units at 5.2V with WC on the hot side will draw 10A for power in of 52W. If you hold the hot side to around 30C, you move 105W per TEC across 10C. Two of those will make a pretty nice 'mini air conditioner' in a small physical size.

Using WC on the hot side, you could just drop the TECs into the bottom of the case, with a smallish heat sink on the cold side to distribute the cold air. Hose routing would be simple, just a single run to the bottom of the case, and no worries about condensation or leakage (if there is any it would just be at the bottom of the case, not on any components). You would also drop the case fan speeds way low, since you want the cold air from the TEC to stay in the case and not be pumped into the room. This would look kind of like the picture at the bottom of post 2 above, but a lot smaller.

Heat load from that setup would be a total of about 320W. You would not need to go crazy on the radiator to manage that load. You could end up with a very quiet box with internal case temperatures maybe 5C below ambient, and at a pretty low cost - you can get a Swiftech Apogee GT CPU Water Block for $45 at http://www.heatsinkfactory.com/swift...k-p-16489.html
Using your low cost rad and any decent pump, you should get good performance. The Apogee is 50mm so you will need to mount it to a hot plate to cover a 62mm TEC, but those are fairly easy to find at under $10. The 12730 TEC is about $30 including shipping. So you should be able to put that rig together for under $200 total.

If you add in-case WC to the setup, you can do even better. You would not need another radiator, since the internal WC would be cooled by the cold side of the TEC. You would be able to keep the internal WC a degree or so below ambient for a heat load up to 250W. While that is not super sub-zero OC territory, either of those designs should allow you to get a decent overclock on your 4400 while still maintaining a nice quiet case, and no worries about condensation on your expensive components.

[Uncle Jimbo]

I've came up with an idea for chilled convection. Though I've never really seen anybody do something like this other then putting there computer on top of a air duct or something.

-Xeniczone

Here's a heat sink that could make the convection idea work -
http://www.aavidthermalloy.com/produ...n/augfin.shtml
the one in the link is 12 in wide and 16 in long. with 100CFM it delivers .014C/W, and it is big enough to mount a pair of fairly quiet fans on the end.

It would pretty much cover the top of a tower case, but it is plenty big enough to mount 4 62mm TECs, or 8, or whatever you want. Talk about Xtreme...

Using 4 12730's, 2 pairs in series to get 6V each from a 12V supply, you would draw 8.4A per pair, a total of 16.8A at 12V, for total power in of 200W. You move 280W across 20C. Your rise above ambient is only 6.75C, so at a 25C ambient, the cold side would be at 11.75C, pretty darn chilly. The case air would be warmer, but moving 280W out, if the air flow is right, you should get some good cooling.

The heat moved goes up fast if the dT drops - so if there is a big sink inside, and the cold side is only 10C below ambient, you will move over 400W of heat.

Running the TECs higher, or using bigger TECs, can move some monster heat. For example, using 4 of the 437 TECs at 12V, you would draw 13A per TEC, for a total power in of 624W. You move 680W across 20C dT.

That much power would heat up even this big sink, but if you use 120CFM fans, you will see a rise of 16C over ambient, or 4C below ambient on the cold side. 680W is big cooling - 2200 BTU/Hr.

[iocedmyself]

Ok well these may be overdue, but i've just come across them so...better late then never.

Uncle Jimbo is oddly enough actually my uncle, thus making me the nephew that has been playing with tec set ups to which he reffered.

I initially got the idea of a TEC AC unit 3 or 4 years ago, shortly after seeing the $400 case marketed as being an AC case...which off the bat i thought was horribly designed even before seeing how poorly the cooling performance actually was. Evil-98 posted a pic of it in one of the first posts.

My logic was based on the simplistic explanation of "hot air rises, cool air falls" Which in my opinion made the placement of the TEC/heatsink stuffed in the front under the hard disk racks was about the worst possible location.

I can already see the argument defending the placement of "actually hot air travels upward to cooler air in an attempt to be cooled, thus making the bottom of the case the best place for it to be as it's the coolest part of the case and you don't want the TEC performance to take a hit from hotter ambient air"

Again my logic was that despite wanting the TEC to be in the coolest possible place, having it at the bottom meant that no matter how cold the air produced, the air was already where it wanted to be, aside from the other obvious fact of the cold air immiedietly being blown past the relitively hot HD's and instantly warmed up. More over, the justification for the extra cost of the case and electric bill was that your CPU, RAM, GPU and motherboard as a whole would be benifiting from the chilly TEC air. Having it in the bottom front, blowing out air that's 10c - 20c may succeed in dropping the entire case temp by a few degrees. which of course drops the componet temps by a few degrees. Which in all likely hood refering to the average users case set up, could be attained by either installing higher CFM fans, redirecting the case airflow, or some combination of the two.

My other interest in this was the fact while water cooling may be a great cooling set-up, especially paired with a TEC under the water block its also got a lot of downsides that range in minor to "holy S**T my $4000 computer is a F***ING aquarium and i have second degree burns on face/arms/chest/crotch"

The average user doesn't have the interest or ability to install a WC rig, especially when the ones that give reliable high cooling performance can cost half the price of an entire system.

Failing to install the waterblock, or tubing properly risk cooling performance and possibility of hardware damage through overheating or water leaking.

Improperly sealing the cpu socket, gpu core, NB core, water tube connections risk leaking at worst, or condensation dmg even in properly set up WC rigs.

Overheating can occur just from failing to introduce the coolant liquid to the loop properly as air bubbles may form.

No matter how good your WC set up is, water by itself means the best you can possibly do is ambient temps. Even that is a stetch in most WC rigs unless you're willing to spend upwards of $300-$400 which take up an enormous amount of space. For that kind of money i would expect sub ambient temps of some kind.

So you can introduce a TEC to the WC loop. Allows for sub ambient temps sure, but ups the possible problems to the aquarium 2nd degree burn scenario previously mentioned.

By in large the only sure fire way to get sub ambiet temps with WC is putting the cold side of a TEC on the CPU/GPU/NB/RAM and cooling the hot side with the waterblock. This pretty much requires the minimum of a $200 WC rig for a single cpu or gpu, and upwards of $500 even when only cooling just the CPU, NB and one high end GPU each with a TEC.

Now even without a TEC in the mix, cooling just a cpu or gpu alone there is the possibility of the water heating to the point of becoming steam. Steam causes pressure to build up which could either melt the tubing in some instances or cause the seal to rupture spraying very hot water inside or outside your case. The preferance as to where being dependant on your relitive proximity to the case when it happens. This could happen due to pump failure, fan failure on the radiator, either because of power failure or componet failure mineral or any kind of gunk build up within the loop.

With a TEC, those same things are possible but with added consaquences. At best the TEC might blow up, in which case the hardware would hopefully shut down before any damage was done. Worst case one of the problems all WC rigs face would occur, causing the TEC not to get cooled which would heat up the hardware. This of course may also cause the system to shut down once it heats up to a certain point. The downside being that depending on how the TEC is powered it could remain on even with the system off. Now when the hotside isn't being cooled by the Water block, the hotside not only gets hot but the "cold" side heats up as well. In this situation with the water not circulating it will almost certainly get to the point of boiling with the power it takes to cool cpu's and even some gpu's. The question is whether or not this will happen before or after the TEC reaches a tempture capable of frying the cpu/gpu/nb. Even if the TEC burns out before it kills the chip, the water will still likely kill the hardware. But then there is the possibility that you'll hit the trifecta. The pump loses power, but is otherwise fine but the tec overheats to the point of frying the chip, leaving everything else unharmed but then staying on long enough to cause the water to boil, which cause a rupture in the tubing which fills your case with water and now frying the rest of the hardware and damaging the expensive WC rig before the TEC overheats and dies.

WC has improved in performance, price and reliability in the past 4 years, but whlie the concerns i had then are still true today, they were alot more common place back then. However unlikely those scenarios may be, they are possible.

Those basic facts being true, it made sense to me to have the TEC mounted on the top of the case. Which i released may not be the simplist of set-ups, or even a possible one for most average users case. But it allowed for the Hot side heatsink to exist outside the case, with the cold side positioned in direct line with the motherboard componets most desired to be cooled. With the "hot-sink" outside the case you're able to use much larger TEC's then the incase config, or use numerous smaller tecs. Hell, even use numerous large TECs.

It was another year and a half, or 2 years from the observation to when i actually made any attempt myself to test anything.

The first attempt was a combination of dirty and creative. Uncle jimbo was impressed with my idea, and decided to give me a hand with it as he has exponentially more technical knowledge of TEC and thermaldynamics than I. Once i ordered three of those 62mm E-bay tecs, the actual designing and testing began.

Unfortunatly we had not yet discovered the .06w TR melcor heatsinks designed specifically for TEC cooling But it resulted in some fun and informitive testing none-the-less.

This first pic would be a shot of the "hot-sink" mounted on the top of an ultra aluminous case. The copper bar is 5/8'' thick x 3.25'' x 12'' if i recall correctly. Bolted to it are 3 cheap-o aluminum cpu heatsinks each with a 90cfm 80mm TT tornando fan. The copper bar was lapped down to a shine over a 5-6 hour period. A hole was cut in the top of the case just barely big enough for the TEC to pass through, 64mm x 64mm at most.



The second pic here is a shot of the interior "cold-sink" which is 5/8th x 3.25 x 4 with a TT Blue-orb bolted to it using a modified version of one of it's included motherboard mounting brackets. Even at the distance in the pic you can clearly see the frost build up on the copper plate and condensation on the fins of the Blue Orb.



The third pic is just a close up of the Blue Orb to show the frost/condensation build up. After running for 10 minutes or so in a closed case the droplets on the tips of the fins actually froze. Which isn't to shabby for a custom air cooled TEC.



Another shot showcasing the frosty buildup.



This last pic is just a shot of the interior case temp. This was achieved with the case fans remaining on, but at the lowest possible speed and no insulation. Ambient temp was about 84F or roughly 26c so 19.6c puts the interior case temp at about 68F. 16F drop is pretty decent.



Despite the good performance, once Uncle jimbo found the Melcore sinks and the numbers got even better i started playing around with another alternitive - ducting. Staying true to my original observation of it being best to have the important componets cooled most even if it means the entire case temp isn't droped as much i took the "cold-sink" side of the melcor unit and ran a duct to the CPU. In doing this i dropped the CPU temp down to 8c. Granted this was a mobile barton so on a 50w cpu when you're moving 150w or so it's pretty easy to get a drop.

But in ducting the cold side to the cpu i realized that you really don't even need to have the TEC mounted in your computer case. In fact doing that will cause you to take a performance hit in cooling. Better just to have it mounted in a seperate insulated case ducted to a case fan that is set up to be an intake (which i've found reversing the normal case airflow is much better at cooling CPU/RAM/NB) then duct inside the case from the fan over the CPU heatsink (or to the GPU/NB etc intake).

At first uncle jimbo and i liked the idea of using a mini-atx case to mount the TEC in. Would have a place for the TEC psu, a PWM controller and Temp monitor. This proved to be better on paper, the cheap cases were too small, too difficult to insulate and the right ones were to expensive to be worth it. Then we thought of making or having metal cases made to a custom spec. Again good on paper, too expensive in practice.

I actually got the idea after seeing a guy who had built a really beautiful custom comp case out of Oak wood. Really annoyed me that i had overlooked something so simple. Good insulator, cheap, easy to work with. After 9months of being stuck, i'd found the answer.

The next problem was with the ducting. The plastic flex tubing you get in OC kits is too expensive and flimsy to be worth it. I've had the ridge of the acordian tube break off at the thick plastic mount juncture on 7 kits now. Plus it can't be insultated.

Right now i'm devising more traditional AC like ducting using foamboard. Pretty decent insulator in itself, nice and cheap, and rectangular ducts are easier to mount and enclose the cpu. For a final build i've come to like the idea of having UV reactive acrylic insulated with neoprene. In fact if i hadn't gotten sucked into this rant i probably would have been done with it by now =D. At any rate i've managed to get my ghetto test rig as uncle jimbo mentioned earlier on the opty, cooling my 9850 Phenom to 15c with the box heatsink.

So i shall get back to that and try and get some pics up in a few hours