Coolit's USB Beverage Cooler

[treatmentx]

I had an idea....

I already own this
http://www.justechn.com/reviews/articles/2006/01/31/usb_beverage_chiller.php

then custom drilled into one of these
http://cgi.ebay.com/KOOLANCE-CPU-300-H06-CPU-Cooler-Cooling-Water-Block_W0QQitemZ160133172540QQihZ006QQcategoryZ5105 6QQssPageNameZWDVWQQrdZ1QQcmdZViewItem

Cheap TEC into the loop...

thoughts?

edit: someone kinda had a similar idea ..... albeit this is making the ultimate drink cooler... whereas I'm thinking of cooling CPU's.... 2.5W should be ok without worry of condensation? >=)
http://semprongamerrules.110mb.com/semprongamerrules_gamer/index.php?option=com_content&task=view&id=14&Itemid=26

[billb]

I had an idea....

"Are flames supposed to shoot out of it like that?"

Get real!

[treatmentx]

lol is it that bad? I don't see why not. take another copper plate, place against TEC plate, it's already got a housing, and a finned sink/fan on the other end.

[Otter]

Yes, it's that bad a idea. :) A 6W TEC won't be able to handle 100W from your CPU. If you built this thing and turned it on, the magic smoke would probably get out of your CPU before it could shut itself down.

[treatmentx]

oh... i think i didn't explain myself very well...

My CPU has an Apogee GT to it already.

I want to add TO the existing loop, a copper block attached to the peltier found in the USB beverage cooler, on top of what I already have (MCR220x4 fans)... it seems like it'll really cool down the water, but not cold enough to hit dew point for condensation... given it's low 2.5-5W TEC.

[Propaganda]

This has been discussed extensively in the extreme cooling section.

When it comes to wattage of TECs used.... 120watts+ even then... not so good.

[Otter]

My CPU has an Apogee GT to it already.

I want to add TO the existing loop, a copper block attached to the peltier found in the USB beverage cooler, on top of what I already have (MCR220x4 fans)... it seems like it'll really cool down the water, but not cold enough to hit dew point for condensation... given it's low 2.5-5W TEC.
Ah, okay, so you're going to use the TEC to cool the water instead of the CPU. That would be much less of a fire hazard.

It shouldn't hurt anything, but how much difference do you think that little bitty TEC is going to make? What is its rated cooling power? Even if it could take 10W out of the coolant, I don't think you'd notice the difference.

[treatmentx]

According to this guy
http://semprongamerrules.110mb.com/semprongamerrules_gamer/index.php?option=com_content&task=view&id=14&Itemid=26

converting it to a molex vs USB (quick & easy 5 minute solder) will get you 15W. I have it at my desk, with an old memory card copper plate... i'd pick up the copper plate and it is COLD (relative to where I live, Los Angeles).

I think i'll research some more, and just do it, post my findings....

and i think this might do a better job than that CPU block.
http://lytron.com/standard/cp_presslock_CP15.htm

I dunno... what do you think Otter... you're wisdom is well recognized here.

[Otter]

Well, my wisdom on Peltier cooling is that I've never tried it and don't know much about it. For me, it's more trouble than it's worth. The folks in the extreme cooling section would be more help. But I think you're going to need a lot more TEC than that beverage cooler to make a noticeable difference in your load temps.

Your basic design has a couple of advantages. The CPU won't toast instantly if the TEC cuts out. And you could turn the TEC off to save energy when the extra cooling wasn't needed. But you'll need a lot more grunt and a fairly complex setup to make it practical.

One of the problems you'll be dealing with if you get a TEC powerful enough to make a significant difference is that TEC's use quite a bit of power. IIRC, it's usually about as much as the heat the TEC can transfer, hence you've got twice as much heat to get rid of. That will be partly offset by the higher temp on the hot side, but you're still going to need a serious cooling system for the TEC. You could do it, but I doubt it would really be worth it. For the same expense and effort it would take to make your ambient TEC-chilled hybrid, you could probably build a phase change waterchiller.

Edit: Oh, and the most improvement you can hope for with the hybrid is the difference between your coolant and the air entering your rad in your current system. Once the coolant went below ambient, the rad would start heating it. That's good, I suppose, because it makes condensation pretty unlikely. But what's your coolant temp now? A second rad would give you half that difference with very little added complexity and without the egregious inefficiency of a TEC.

[billb]

oh... i think i didn't explain myself very well...How about this for another bizare idea?

[Clockwork_Apple]

How about this for another bizare idea?
You're just adding another medium between the water and air.

water > heatpipes.

[oakstave]

The peltier effect coolers have not caught on to a vast extent because of the massive inefficiencies inheirant in the devices. You would get MUCH more of a boost in cooling if say, you were to put an ice bag on your reservoir.

I don't think you are going to get much effect with that wattage. But if you're going to do it just to do it... well don't forget to post your results!

Here's what you really need. A USB toaster! http://store.theonion.com/gotcha-box-usb-toaster-p-71.html

[XeonStrikeForce]

Better yet save your money and buy a real chiller that will drag it down to near-40C at fractions of the power sucked up by an equivalant pelt system.

[treatmentx]

? i thought the pelts were power mongers, which is why it hasn't caught on... (like oaksteve said)

and the USB only sucks 5v, Molex 12v, according to that guy to really juice it up.

The idea is to get it to ambient, sub-ambient on the molex perhaps.. and I'm watching the dew point in Los Angeles (10C currently), but given that it's not -40C, don't have to worry about getting it that low...

Anyway, working towards this, the Lytron plates are way too big and I emailed them requesting a custom plate, but they don't make custom plates this small, so I have go with a cheapo block... chipset block or a koolance cpu block on ebay

http://cgi.ebay.com/KOOLANCE-CPU-300-H06-CPU-Cooler-Cooling-Water-Block_W0QQitemZ160133409503QQihZ006QQcategoryZ5105 6QQrdZ1QQcmdZViewItem

Will def post results, which I assume won't be anything staggering, but it's a cheap solution... $10-15block, $15usb coolit, $4tube/fittings. It's a similar idea to Coolit's Freezone (chilled water) ... which we've all seen the reviews. Given the strength of the pelts on those systems, a TEC cpu block is needed... whereas this USB is weaker, hoping I can escape on a molex mod without the condensation. (aiming for 20C, above dew point 10-15C year round high/lows in LA)

anyway, i'll do it, will post results when i get it up.

[Denbob99]

I would like to see results for sure. How many watts does water in a watercooling loop have?? :P

[Drew@PSU]

Well, the thermal output of the QX6700 at stock volts/clocks is 130 watts, add in overclocking/volting, I have a feeling that for just the CPU, you're at 200+ watts. TECs are really only good for moving heat from one place to another, ie. a heat pump. However a 6 watt TEC will just annoy you, I'm thinking. You'd need a decent sized copper heatsink on the hot side too.

[treatmentx]

Well, the thermal output of the QX6700 at stock volts/clocks is 130 watts, add in overclocking/volting, I have a feeling that for just the CPU, you're at 200+ watts. TECs are really only good for moving heat from one place to another, ie. a heat pump. However a 6 watt TEC will just annoy you, I'm thinking. You'd need a decent sized copper heatsink on the hot side too.

I have an extra Zalman CNP9700,since I moved into watercooling, I was going to mod/put onto the hotside...

looking for more information/input before embarking on this sillyness lol... I feel like the zalman's been good to me... ? shrug.

[treatmentx]

After finally putting together this contraption, this is the conclusion

First off, I had to wait for ANOTHER block to come in, the Koolance off of ebay was the incorrect 1/4"ID vs 1/2"ID with the rest of the loop, that was a pain in the dingaling

My contraption was a USB Coolit Beverage modified with hotside with a Zalman 9700 cooling it, then a MCW30 on the cool side. I'll refer to this as "My USB Beverage Contraption" or "MUBC" from here on.

So i was mostly interested in getting my temps in my GPU down, my loop was

MCP655-->MCR220 (4x120mm fans in pushpull)-->Apogee GT-->MUBC-->MCW60 (with Uni-Sink)-->MCP655

Prior to MCW60, my XFX XXX 8800GTX GPU temps were 72C idle/80C load at the stock 630/1000 speeds.

With the MCW60, but MUBC turned off (USB not plugged in), my temps were 50C idle/ 58C load

With the MCW60, MUBC turned on, my temps were 48C idle / 57C load

in short, it made no difference. The 1-2C difference could simply be due to testing error.

All tests are based on idle speeds of 30minutes of non-graphical / non-cpu use. Graphical tests were measured with NVMonitor AND Rivatuner running in the background with 3dmark05 and 3dmark06 allowed to run only tests with graphical stress (not the CPU tests) and load temps were averaged from beginning high to ending high.

*close*

[Denbob99]

Well done for trying it :P I didn't think it'd do much tbh, waaaayyy too weak in terms of watts.

[samuknow]

Pick up a bigger pelt off ebay. They are cheap. You already have the setup. 80 watts should make a little dent. I really want to see if this works.

[Sam__]

samuknow is right....prlts are bloomin cheap on ebay...i got a 92W one for about $15.

I intent to use in on my nb some time when i get a good enuf heat sink...ovbiosly id only run it at 5v giving it a wattage of about 45.

Anyone tried this? pelt on the nb.

[treatmentx]

Pick up a bigger pelt off ebay. They are cheap. You already have the setup. 80 watts should make a little dent. I really want to see if this works.

Thx for the encouragement :beer:

Actually, the voltage going off the USB is probably about giving 5W to the pelt. Reading around the web, I found out if you rewire the USB power source to a molex power source (right from your PSU), you'll get up to 15W to the pelt, and will definitely make a dent.

I retried some tests this morning, and it's coming up to 2-3C difference on the GPU, so I believe there is some impact, but very little. If that's true, then let's say we have an average of 2C difference on a 5W. We multiply that power by 3, we get 15W, i'd say you would see at least 6-8C difference. It does work, but the thing that people should keep in mind is to keep the hotside cold. The cooler you can keep the hotside running, the better it'll run. I live in Los Angeles, it's been hitting 90F-100F lately, the Zalman could only do so much :(

Anyway, it's not that hard to set up, cheap pelt on Ebay, cheap waterblock on ebay or classifieds, old HSF (I had a Zalman), Home Depot material to clamp it, then good to go. I wouldn't put it directly on a computer component (NB or what not), you'll run the risk of condensation ON the computer. the point of my experiment was to chill the water enough outside of the case, close to ambient, without running the risk of condensation... to do that, you would have to research about condensation and dew points etc in your local area. (mines was 10C)

[XeonStrikeForce]

Q=M*C*dt^

Say about a litre of water = 1.01Kg
Specific heat of water is about say 4.2
DeltaT Say (30-25) 5

Q= (1.01/3600)4.2(30-25)= 6W Now thats just water with no heat being added! So 6W+ say 200W from the PC pluse 25% efficiency loss we end up with = 260Watts rounded up to make a real differance.

[itstemo1]

Nevermind what I said

[Otter]

Q=M*C*dt^

Say about a litre of water = 1.01Kg
Specific heat of water is about say 4.2
DeltaT Say (30-25) 5

Q= (1.01/3600)4.2(30-25)= 6W Now thats just water with no heat being added! So 6W+ say 200W from the PC pluse 25% efficiency loss we end up with = 260Watts rounded up to make a real differance.
Actually, the amount of water doesn't matter. You can estimate the temp drop with this equation:

Td=(Tc-Ta)-(Tc-Ta)(Qs-Qu)/Qs)

Where
Td is the drop in temperature for the mod,
Tc is the coolant temp,
Ta is the temperature of the air entering the rad,
Qs is the total heat load on the original cooling system, and
Qu is the the heat transfered by the modified USB cooler.

Plugging in some wild guesses and a TEC that transfers 80w in your actual setup, we get
(10c)-(10c)((180w-80w)/180w)=4.4c

So yes, with a powerful and adequately cooled TEC, you could push your temps down a bit. But the much simpler option of adding more radiator can do that too. For adding another MCR220 the drop would be roughly

Td=(Tc-Ta)-(Tc-Ta)/2

or in this example, 5c. And replacing the mcr220 with an mcr320 would knock off about 3C. As you increase the heat load, you'll need a bigger TEC to compete with a larger rad. To equal the drop for doubling the rad, you'll need the TEC to get rid of half the heat load.

While the USB cooler doesn't make sense from an engineering perspective, it's good experiment and a way to get a feel for what would be involved in making a more serious TEC/ambient hybrid. I still like the idea of being able to turn the TEC off without blowing the CPU. Thanks for sharing the project with us, treatmentx. :beer:

[XeonStrikeForce]

thats a machanical refrigeration formula, it ignores air heat exchange and is only interested in the flow of energy of the refrigerating device and size needed to achieve said delta! thats why mass is important along with specific heat, water hold tons of latent but since we're not freezing it we can ignor that.

[Otter]

thats a machanical refrigeration formula, it ignores air heat exchange and is only interested in the flow of energy of the refrigerating device and size needed to achieve said delta!
And that's why it's not relevant to the USB cooler project. All we're interested in here is the equilibrium temperature. If treatmentx had a big res, it would take the rad and TEC longer to cool the water down after folding for an hour, but his load temps would be the same as if he ran a t-line.

[XeonStrikeForce]

I was showing the tech makes no diff at the watage it is at. The Rad even further mitigates the techs cooling. for it to actualy make a differanc it would have to be in the 260 watts range, unles he simply wants to cool the water while the computer is off then it will cool the water.

[Otter]

Cooling the water while the machine is off is what your refrigeration formula is for, eh?

I don't know much about thermoelectric cooling, but 260W sounds like a reasonable estimate for cooling by TEC alone. I think all treatmentx is looking for is a boost to his regular WC system to get the temps closer to ambient, though.

[treatmentx]

Whoa, went away for the weekend, didn't think this thread was still alive.

Yeah, it was just for educational/experimental purposes so I, as well as everyone who's reading, can gain a knowledge or two.

I already had the MCR220, my setup was already pretty cool. I was just experimenting with a USB powered 5W pelt to see if it made a difference, which it probably did to a degree that's unmeasurable (1-2C) to have any real significance in the real world performance of the GPU.

I really like the formulas posted, keep it coming! :beer:

[XeonStrikeForce]

Cooling the water while the machine is off is what your refrigeration formula is for, eh? No.

I don't know much about thermoelectric cooling, but 260W sounds like a reasonable estimate for cooling by TEC alone. I think all treatmentx is looking for is a boost to his regular WC system to get the temps closer to ambient, though. That is what the formula says we need to do just that

A 260 Watt pelt will give a 5C drop in water temp leaving it.