oooohhhkay, been thinking about this a while, here's a design.

[iroc409]

it's been discussed recently, so i figured i'd finally discuss what i've been thinking about for a while. this is similar to cnyc's deal, but for a little different reasoning.

initially i was going to go fairly large-scale with this several months ago, but decided against it. too much power requirements, yada yada. but, the idea has changed a bit.

the idea now is to build a mostly uncomplicated setup, with for the most part an extremely compact unit. the main cooling system would be about 80mm tall and 160mm wide, by whatever length deemed necessary. the intent is actually so this bolts into the stock hard drive cage attachment on the bottom of a lian-li case.

the idea is to get something that would come up with *fairly* good results in a fairly safe and compact device, and, no radiator. thus, a single-pelt cooling setup intended only to cool the cpu. which means 2 short hoses, 1 to the cpu and 1 return line. the pump would be housed in the reservoir... i was looking towards the pentair aquatics 800 or 1200. they're pretty small, but due to the simplicity of the loop and the short distance, i felt this would be acceptable. in any instance, it would be a pretty small loop. now i'll post a little picture to give better visuals:



now, you can see basically what i'm talking about the setup. probably looking towards a 120w pelt was my thinking, but perhaps that's all bunk. a 220 would be best, but i just couldn't think that an air hsf could shed all that heat.

the cooler inside the reservoir would be constructed of either small (like 1/8") copper tubing, or very thinly zig-zagged or corrugated copper sheetmetal.

the tubing would then be the rubber braid-covered fuel tubing. it's fairly insulating, and it looks cool


anyways, that's one of the ideas i have on the drawing board. maybe it should get left there?

i dunno, it was nice, compact design i thought just might work, but how everyone says it won't, maybe i'll work more on my other stuff, that included a radiator. i'd appreciate your thoughts though.

[SparkedFire]

you wont know if it will work untill you try it.

[Since87]

Originally posted by SparkedFire
you wont know if it will work untill you try it.

Not true. All the information needed to determine if this will work is out there.

I'd say this is an idea that should be left on the drawing board until a lot more research is done.

[iroc409]

yeah, but has there been anyone that has tried anything like this, or can anyone with a better engineering background describe more the dynamics of why it would or wouldn't work?

i've looked around a bit for this and i've not seen anything like it, so it's hard to base judgement on it to myself.

i will most likely try it sometime anyway, just out of curiosity, it's just my schedule right now is very tight for working on fun things like DoD and tweaking electoronics

i haven't played games now in a couple weeks

[SparkedFire]

let this exarament be the research!

[iroc409]

Originally posted by Since87
Not true. All the information needed to determine if this will work is out there.

I'd say this is an idea that should be left on the drawing board until a lot more research is done.

the best research is probably actually trying this. is there anything similar out there? i've yet to find it.

i have the resources that would most likely be able to build this pretty cheaply. doesn't need much more than a cutting device, a drill press, taps and a method to fasten it all together. all of which i have access to in our shop. although i've never soldered copper before, just electronic junk.

[SparkedFire]

like i said

[Since87]

Originally posted by iroc409
or can anyone with a better engineering background describe more the dynamics of why it would or wouldn't work?

This is hardly a tutorial, but this thread at Overclockers touches on a lot of the relevant issues.

[iroc409]

very interesting article/thread.

that kinda makes me think that this isn't going too well. so, i'm going to scope out the hardware stores around here and see what i can find for cheap, suitable parts. thinking about temporarily using a plastic reservoir. not sure that the plastic can withstand the temp variations, however from that article there may not be much of any severe variations anyways.

alright, not entirely sure if i'm going to do this yet or not, but depends on what's available, i just might. here's what i'm thinking of setting up.

basically just use a plastic reservoir, and cut a hole in the side for the internal hs, and drill it for a a mounting system. then basically build the internal heatsink, still using the tiny pipe for the "fins" (i'll get to this in a second). then, basically do the rest like i displayed, just not as fancy. then, just set a fan by it for now, and use a cpu heatsink to cool the pelt (for now). probably use a 120 pelt or so.

now the reason for the tube-style heat exchanger is kinda 2-fold. first, although tedious, i think it would be easier to construct from scratch than a fin-style exchanger.

and secondly, this is a longshot, but i thought it might improve the exchange. because the water flow would come down over the pipes, the water inside the pipes would not be directly affected by the water flow. thus hopefully 2 things would happen. first, hopefully the water inside the tubes would stay longer and have more contact with the heat exchanger, thus removing more heat from the water. with this you'd get kindof a convection current through the setup, and probably also get suction from the passing water to draw the water out of the tubes. however, if it's too fast, it will just pull the water out of the tubes too quickly, without the heat exchanger doing it's job. so, this could be good, could be bad. not sure yet.

anyways, that's what i was thinking, probably make a trip out to menard's today, see what i can find.

[bigben2k]

Again... It's been done here.

The problem is that you can only effectively air cool a pelt that's no more than 80 Watts, and that's insufficient for today's CPU (the rating isn't comparable to the CPU power). You could get away with 2 * 80W pelts, both aircooled...

The clamping pressure on the Pelt is critical: OnDaEdg's article.

Fins/heatsinks are going to be important, on both sides. Keeping it all sealed might be a challenge, in the long run.

Energy consumption will become a factor.


Otherwise, it's a pretty cool thing to build

[Crosstrack16]

Otherwise, it's a pretty cool thing to build

No pun intended

[iroc409]

Originally posted by bigben2k
Again... It's been done here.

The problem is that you can only effectively air cool a pelt that's no more than 80 Watts, and that's insufficient for today's CPU (the rating isn't comparable to the CPU power). You could get away with 2 * 80W pelts, both aircooled...

The clamping pressure on the Pelt is critical: OnDaEdg's article.

Fins/heatsinks are going to be important, on both sides. Keeping it all sealed might be a challenge, in the long run.

Energy consumption will become a factor.


Otherwise, it's a pretty cool thing to build

yeah, i;m starting to think more and more that it's just not that feasible. although i wouldn't be attempting to cool near as much water as that french article (although i can't tell entirely, because i don't speak frenck ).

it would be pretty cool though, because it would be a very compact setup. power consumption is an issue, and i'd either have to build a psu, or buy one of those fancy $120 units. i'd probably just build it, somewhere i have plans for the same thing, assuming i can find it (my file system is a mess).

i'm just going to check around today, and if i can find what i'm looking for cheap enough, i just might build it.

i had another design that was considerably more complicated, wich included a chiller system, that could be done single-loop, or double loop with 2 pumps. the 2 pump setup would actually probably work a lot better (and this unit uses a rad).

i guess we'll see what i find first, and if anyone has any 80w pelts they want to unlead cheap, i'm all ears

[iroc409]

oh yeah, can you solder thin copper (for like the tubes on the hs) with a regular solder gun and stuff, or do you have to use the torch and all always? i could build a basic version of this a lot easier that way (although i suppose i could just use a prebuilt hs, but what's the fun in that?).

[bigben2k]

A torch would be more fun Wish Gmat was here


Personally, I'd go with an XJinn block (cross drilled), and run the coolant back and forth in it, by redirecting the flow from one hole to the next, with small bent copper tubing. Of course I never look forward to making a cross drilled block.

[iroc409]

Originally posted by bigben2k
A torch would be more fun Wish Gmat was here


Personally, I'd go with an XJinn block (cross drilled), and run the coolant back and forth in it, by redirecting the flow from one hole to the next, with small bent copper tubing. Of course I never look forward to making a cross drilled block.

yes, fire fun, burning down apartment complex not ;P. actually, it's just because i have to go buy one


the inner hs is going to be open-ended tubing, it's not a chiller but a slightly difference heatsink built to actually hold the water for some amount of time, probably with a ton of like 1/8" copper tubes.

[iroc409]

although.. just thinking about it, i *could* do kinda what you are talking about... makes for a quite interesting assembly, but it's possible... i could build the finned setup on top of a cross-drilled base, which the pump would pull it's intake water through... that might actually be more efficient. this is assuming you understand what i'm talking about. basically, looking at it from the side, you'd have thin coldplate->radiator-like flow through inner block (but would be thin as possible)->exchanger tubes.

have merit?

[bigben2k]

Yeah, you really have to take advantage of the pump here.