chiller idea.....will it work?

my basic thought is to take like a few like 40 watt pelts, so you can cool them with heatsinks. Have a like res that is just a big PVC pipe, like 2 inches in diameter, and cut holes sin the side.

Then, put some heatsinks on the cold side of the pelt too, and have the cold side heatsinks sticking into the water. So, the pelt has heatsinks on both sides, but the cold side is in the water, and hopefully cooling the water. Then, just cool the other side with like stock Athlon coolers or something cheap like that.

Think it would work?

Also how many 40 watt pelts would it take? like 5?

Also, where could I buy pelts in the 60-40watt range?

Wasn't sure where to post this........

Sounds ok to me

i posted this under extreme cooling
{
i've had this idea to chill water, it is increably inifecent, but have never tried it.

get three water blocks, two pumps, a 200+ watt pelt, a HUUUUGGGEE rad and tubing, insulation (that stuff)

Put the pelt between two w/blocks

run the hot sides loop through the rad and the cold side to you proc.

it might take more than one pelt with modded waterblocks but you could chill water without the fridge.
}

Although u may be goin for the whole "lets be different" thing. Ur beetter off throwing those pelts where they should be, between cpu and waterblock. Much more effcient way of doing things dont cha think?

Mikeonatrike said:

Although u may be goin for the whole "lets be different" thing. Ur beetter off throwing those pelts where they should be, between cpu and waterblock. Much more effcient way of doing things dont cha think?

Click to expand...

Well, yes and no. This way you chill all the blocks in your system, while you may get lower temps just using pelts the original way, it would also take more power to do that, along with being more expensive.

You are basically describing a peltier chiller. Most solid-state water chillers (or baths) do this very thing. My chiller is based on 10 55w peltiers surrounding what would look to most like a HDD water block. The cold sides face the block and the hot sides share a massive heatsink. I am still working on finding the ideal flow and coolant, but so far my record was -8c on a stock CPU. I'm also using a standard heatercore inline before the chiller, so the coolant is very near ambient before it hits the chiller.

I visited the factory of Marlow Industries for some ideas... they are the TEC manufacturer for several peltier-based water coolers. They even have several peltier chillers on the Hubble Space Telescope. I was able to see a few of their water blocks (cold plates) in cross-section. Most of them looked like a cross between a maze3 snake path and a bumpy or pin-shaped rods for surface area. Sort of like the 'pin-fin' DIY water-block from yesteryear.

My core is a straight-through path with a buttload of rods inside. I took a pic before I reinsulated the whole thing, but the pic didn't turn out as well as I hoped.

You can kind of see the idea. It's about 10" long and that round hole is the entry point for the coolant. Each side mounts up against a 55w peltier and it gets COLD inside. The first time I set it up, I ran room-temperature water through it with a 110GPH baby pump. It froze up and stopped flow after 20 seconds or so. That was pretty sweet =p.

Anyway... yeah. It's a fun project to do but most people would probably say that phase-change is more sensible... especially when you consider the power consumption. I pull about 700w when I run the whole assembly. That's pretty stupid, but I'm working on a few ideas to lessen this. I doubt a normal phase-change system would cost nearly so much on electric bills =/.

wow, 700 watts. I'll have to think about this then, but if I had a heatercore in the system, I could turn off the pelts if I wanted to, right?

That's exactly what I do. Right now, the coolant just flies through the chiller at high pressure (750gph/21') with the peltiers off. Turning them on at such a high velocity honestly doesn't help as much as you might think. Every TEC-chiller article I've read SERIOUSLY hints at 'low-flow' being absolutely key, and I'm really seeing that. What I'd like to figure out is a way to introduce the sub-zero coolant into an already stable closed-loop >rad/waterblocks/pump> system without worrying about pressure problems or dealing with open-air issues. Condensation is also a concern, but only when I run the chiller for extended periods.

The brainstorming I've had so far includes some potentially STUPID ideas.

Like...

Have the chiller feed directly into a large res where the chilled water will be picked up by the standard cooling loop.

Setup some inner-tube manifolds that will allow a second independant low-flow chilling path while somehow allowing high-flow throughout the rest of the loop. Tricky...

Simply cut the input of the chiller-core to a VERY small diameter barb and run a second loop inside the standard cooling-loop via 2 Y-connectors. The flow would continue with the path of least resistance, the high-flow Y, but there would still (hopefully) be sufficient flow to make the chiller loop worthwhile.

So far all I've actually experimented with is making the whole system run at low-flow. This gets me very low coolant temperatures but doesn't overclock so well =/. The water apparently sits on top of the CPU block for too long. Also, as noted above, I've tried running the whole chiller in-line with 1/2" barbs at high-flow. In that setup, it basically acts as another large waterblock... but it sheds heat instead of adding to it. The problem is that the CPU is more efficient at adding heat than the 10 peltiers are at taking it away =/.

Ramble on! Sorry, I go off on tangents when I talk about this because I keep hoping for some TEC-Chiller guru to step in and tell me I'm an idiot... explain the 'proper' way of doing this type of cooling =p.

The 700w figure is including a fairly efficient regulated PSU, 2 comair 6" fans for the chiller-heatsinks and the dual blower setup on my heatercores. With the chiller powered off, I'm back to 'normal' power consumption for your typical overclocker =p.

vonkaar said:

That's exactly what I do. Right now, the coolant just flies through the chiller at high pressure (750gph/21') with the peltiers off. Turning them on at such a high velocity honestly doesn't help as much as you might think. Every TEC-chiller article I've read SERIOUSLY hints at 'low-flow' being absolutely key, and I'm really seeing that. What I'd like to figure out is a way to introduce the sub-zero coolant into an already stable closed-loop >rad/waterblocks/pump> system without worrying about pressure problems or dealing with open-air issues. Condensation is also a concern, but only when I run the chiller for extended periods.

The brainstorming I've had so far includes some potentially STUPID ideas.

Like...

Have the chiller feed directly into a large res where the chilled water will be picked up by the standard cooling loop.

Setup some inner-tube manifolds that will allow a second independant low-flow chilling path while somehow allowing high-flow throughout the rest of the loop. Tricky...

Simply cut the input of the chiller-core to a VERY small diameter barb and run a second loop inside the standard cooling-loop via 2 Y-connectors. The flow would continue with the path of least resistance, the high-flow Y, but there would still (hopefully) be sufficient flow to make the chiller loop worthwhile.

So far all I've actually experimented with is making the whole system run at low-flow. This gets me very low coolant temperatures but doesn't overclock so well =/. The water apparently sits on top of the CPU block for too long. Also, as noted above, I've tried running the whole chiller in-line with 1/2" barbs at high-flow. In that setup, it basically acts as another large waterblock... but it sheds heat instead of adding to it. The problem is that the CPU is more efficient at adding heat than the 10 peltiers are at taking it away =/.

Ramble on! Sorry, I go off on tangents when I talk about this because I keep hoping for some TEC-Chiller guru to step in and tell me I'm an idiot... explain the 'proper' way of doing this type of cooling =p.

The 700w figure is including a fairly efficient regulated PSU, 2 comair 6" fans for the chiller-heatsinks and the dual blower setup on my heatercores. With the chiller powered off, I'm back to 'normal' power consumption for your typical overclocker =p.

Click to expand...

While I realise this would still have the same flow, what about using a Y-splitter, and running it as like a parrallel heatercore setup, just with pelts.

I was thinking of running a heatercore, and then after that and right before the cpu, have the pelts. But, it doesn't seem liek it chills the water very well for how much power it takes.

vonkaar do you really need the rad in there

i would think it would warm up the water before it went into the chiller?

just wondering

zebkoolindc said:

vonkaar do you really need the rad in there

i would think it would warm up the water before it went into the chiller?

just wondering

Click to expand...

At stock temp, it was counterproductive. The CPU wasn't heating the chilled coolant above ambient so running it through a radiator only brought room 'heat' into the mix. The radiator is there for when the CPU gets hot enough to bring the subzero coolant above room temp, which so far hasn't happened. I haven't done any serious overclocking yet for several reasons. For one, I haven't received my Cascade-SS yet so I'm hesitant in experimenting with flow and pressure variations until I have a real indicator of how the final system will be. I'm also not confident in any of my CPUs to be 'good overclockers' so I'd be embarrassed to post results on crappy chips =p.

If the chilled water is 'enough' to keep a highly overclocked CPU 'happy'... the heatercores will be bypassed (probably with some nifty bypass valves I just found) and the loop will only contain the chiller for the cooling. I don't think that's going to be the case because from what I've seen so far, the chiller is much more effecient at.. hmm... 'chilling' ambient water than... er... 'shedding' heat away off hot water. Like, if I can take 25° coolant and bring it down to -10°, great... 35° diff. But, if I started out with 55°, I don't get a 35° drop. I know that's probably painfully obvious to you engineer types, but the difference is even more staggering than I originally anticipated. The 'idea' was that the heatercores would bring the coolant back down to rooom temp and then I would get my nice 35° drop =p. Of course, bringing hot water down to ambient is the goal of any water-cooler... so... we'll see how well that pans out.

The last problem I'm seeing is the amount of heat that all this crap generates. I have the normal heat coming off the system... 3 HDD, 9800 and misc components. That goes out into the room via 92 and 120mm system fans. Then I have the heat from the 'standard' water-cooling loop. That gets sent out by way of my heatercores/blower setup. NONE of that compares to the amount of heat that a 10-peltier chiller puts out. At the present, that is vented outside the apartment but it still heats up my puter room pretty well. That, of course, reduces the efficiency of my heatercores and ... yeah... you get the point. Cycle repeats itself =/.

Originally posted by sandman001
I was thinking of running a heatercore, and then after that and right before the cpu, have the pelts. But, it doesn't seem liek it chills the water very well for how much power it takes.

Click to expand...

Well, if I ran the chiller in a short-loop with no heatsource, the 'water' would get a heck of a lot cooler than I'm reporting. The bonus to all of this is, if the CPU-cooling idea fails, I've always got a GREAT chiller for drinks =p.

long thread, but lots of chiller test data here

be cool

I was *really* hoping you'd jump in on this BillA... you are the only person I know with actual real-world experience with TEC-chillers in a CPU environment.
Thanks for the link it's answering a lot of my questions.

thanks for the nice words, but . . .
I would say right now pHaestus has more experience than I

I've been debating starting a thread over on procooling on my chiller... mainly because the chiller is so damn pretty while the test bench is so fugly =/. I'll probably listen along on that thread to help me with my flow logistics problems and then I'll ask around. The simplicity of a water-cooled TEC chiller may end up making my whole project obsolete. Oh well, I've already spent enough moeny in the aesthetic appeal to justify continuing with an air-cooler chiller...

I do have one question though...

The MCW-Chill 452 is consuming 34 amps. Is that something that would be considered reasonable? I'm looking at 45-50amps at 'full-speed' and that was sort of scary. I know my wiring can handle it, but I've often wondered if that would freak the average consumer out. Maybe I should find out what Prommies and Vapochills draw and use that as a benchmark =p. I can always drop it down to 6 or 8 TECs instead of 10. I guess my question is... When ya'll designed the MCW-Chill, did the high current requirement throw up any alarms to the marketing people? I can't think of another way of saying that... I mean... is 34a reasonable? Everytime I talk about the 45a I'm pulling, I imagine people are wincing and thinking I'm crazy... but 34a isn't that much lower and ya'll are mass-producing it. So maybe I'm not crazy afterall?

for starters:
the MCW-CHILL was not designed for CPU cooling, rather an industrial application requiring chilled water
- I ran a simple test cooling an AMD heat die, now pHaestus is cooling an actual CPU

35A is a very significant current draw, and for this reason the MCW-CHILL will only appeal to TEC users who already have such power supplies (my opinion only)

equally significant is the heat that must be dissipated to the air,
quite enough to warm a room quickly

be cool

Awesome. So officially, TEC-chilling is still in the realm of the hobbyists. That's good to know =p. I basically put the MCW-Chill as an indication that TEC-chilling for CPU purposes has gone 'mainstream'. Apparently, it's no different (in purpose) than a Thermotek chiller or any standard industrial laser chiller. In that case, it's still up to we few hobbyists to adapt this proven technology to work efficiently in a PC environment. In that case, 34a is far from overkill.

Thanks again ^_^

Personally, I consider phase change >> TECs, as these second have poor efficiency...

That's only because Phase-Change systems are 'proven'.

The way I see it...

There is a highly competitive market for industrial water chillers. Some of them are phase-change, others are TEC based. If phase-change coolers are 'so superior' in the PC-cooling market, why don't the phase-change chillers dominate the industrial water chiller market as well?

One reason phase-change coolers are so popular for the PC crowd is because so many people have experimented and 'proven' the technology to be not only reliable, but practical as well. I have seen a very small fraction of the same effort applied in TEC chilling...

Either the general consensus among the PC cooling crowd is correct and TEC-chilling is simply not efficient, or there is still a lot of work to be done. Undoubtedly millions of dollars have been spent on laser-chillers alone... why didn't those highly-intelligent engineers come to the same conclusion? Why wasn't TEC-chilling abandoned for that market? I see phase-change industrial chillers just as often as thermo-electric chillers.

I may very well be proven wrong... but I believe that there IS a market for well researched TEC-chilling in the CPU-cooling circle.

Eh, gonna have to pas son this for now, it's just too much power usage for too little gain.

I don't say which are proven which not, but TECs have poooor efficiency.

...and it means...

They need lots of power to do things which phase change can do at lower power.
Situation may change when "these new TECs" will hit market.

efficiency (from some site, may be very inaccurate):
TEC - ~5%
phase change - ~40-45% (running at nice conditions)
"new TEC" - ~50-55%

Running 1kWatt cooling on comp doesn't sound nice for me

Good sides of TECs are that they're simple, silent and relatively cheap. Bad thing is efficiency.

So if we can truct "given data", new TEC of 20W would be as good as current common 226W. So this way TECs would be the only solution, with every aspect (probably not price tho) better than phase change. Till then I won't consider them a better solution for high-end cooling.