Question about TEC's (Peltiers)

[Malogato]

Ok, I have a question, say I was going to buy a 226 Watt Peltier, Like featured in this for sale thread or This ebayer..

My understanding is that the way these devices work is there is a "very cold, and very hot" side. I would assume, also, that in the situation, the cold side is down, and the hot side is up near the water barbs.

Now.. I have also read all about the fact that you will get condensation if you have the tec inside your rig, and you need to therefore pad it with neoprene, dielectric grease, etc..

The brilliant (or perhaps asinine, and probably not unique) idea I have had is.

Use the TEC as an external water chiller. Put the cold side (or several of them...) against your radiator, in a 'conventional', externally located, water cooling rig.

Since the tec itself is not *in* the computer, the condensation issue is reduced. (Especially if you use anti-freeze). Also, yes, I know, it seems partly wasteful of efficiency,
but, I don't think a tec that's "on the outside of the radiator" - an inch plus from the water, will freeze the water. - It will, in my opinion, most likely .. "chill" the radiator, making the radiator somewhat more efficient in it's heat-removing capabilities.

If the cold side is positioned properly, (which, in my mind, is on the opposite side of the radiator from the fan) you should be 'getting a cold breeze' - It won't immediatly "chill" the radiator, but after time, the metal temp will drop.

My watercooling setup has already BEEN planned as external, with it's own external powersource anyhow.. So, adding a tec to it won't be much of a chore.

Whattya all think?

Questions, comments, and corrections appreciated. Flames and the like are not.
I'm an extreme cooling newbie, so, be patient with me.

 

[monkeysauce]

Do you plan on running two loops? The pelt will be dumping tons of heat into the water that cools it, and if it's just one loop this really negates all cooling that the "cool" side is contributing.

The condensation will appear on any surface that is sub-ambient temperature (at the "dew point" i believe, could be wrong). Antifreeze inside the tubing makes no difference.

You can buy water chillers from aquarium store for a few hundred dollars. People have also used the refridgeration units from water fountains (think school) to get cool water through their system.

I would make sure when planning your external WC system that you have a large radiator (automobile heatercore that will fit 2x120mm fans) and a very strong pump. Focus on pressure and head over GPM. Does this help at all?

 

[Malogato]

Do you plan on running two loops? The pelt will be dumping tons of heat into the water that cools it, and if it's just one loop this really negates all cooling that the "cool" side is contributing.

The condensation will appear on any surface that is sub-ambient temperature (at the "dew point" i believe, could be wrong). Antifreeze inside the tubing makes no difference.

You can buy water chillers from aquarium store for a few hundred dollars. People have also used the refridgeration units from water fountains (think school) to get cool water through their system.

I would make sure when planning your external WC system that you have a large radiator (automobile heatercore that will fit 2x120mm fans) and a very strong pump. Focus on pressure and head over GPM. Does this help at all?

Can't I let the hot side, just, well, have a big heatsink and fan on it? Or, hell, let it cool into the room? Is it a "requirement" that the hot side be cooled?, or just a "good" idea. -

I can buy a 350watt, -40c peltier for under $30, So, "a few hundred dollars" is more than I want to do.

I've got an eheim 1250, and a decent sized radiator. The entire rig is going external, with just the tubes (and a criticool to turn on the pump) going into the rig [and, well, of course, the blocks] But the reservoir is external [and not small by any means] I was hoping I could, instead of using a $300 CC25 chiller, use a $24 -40C peltier. If it is allowed to run 24/7, it will .. "Eventually" lower the ambient temp of the water

To make myself clear, none of the pumping, reservoir, radiator, or peltier will be within 2 feet of my tower.

I planned this because I do not want to have a magnetic, heat making, additional power requiring pump inside my machine. If I have it "away" I can power it with some other power source, and it's ambient heat will disperse to my ROOM.

I was not intending to have the peltier "in" the water.. I was intending to have it "on" the radiator which has water going through it. - The cold side, only, mind you.

 

[Deezus]

You'll have to cool that big of a pelt much better than a HSF. That's why people mount them IN their rigs on top of the CPU and have the water loop actually cool the hot side of the pelt.

A pelt can only run effecient if the hot side can get the heat away quickly, otherwise it will interfere with the cold side and either not cool at all or cook the whole set-up.

Someone else will be able to explain this better than me.

But by the logic you are using, your saying the same thing as just running the CPU without anything on top. Pelts get just as hot as a CPU.

 

[Malogato]

A pelt can only run effecient if the hot side can get the heat away quickly, otherwise it will interfere with the cold side and either not cool at all or cook the whole set-up.

But by the logic you are using, your saying the same thing as just running the CPU without anything on top. Pelts get just as hot as a CPU.

Yes, the analogy of "hot side cold side" is similair to the cpu itself, but, I was not aware that, like a cpu, the hot side needed to be "evacuated" . I honestly believed that, (I'll use a farsical example here) if someone wanted to just have a peltier hanging from a hook that while being turned on (I mean, like, "out in the open air") that would be perfectly safe for the peltier. It might not be safe for someone who accidentally touches it, but I did not think a peltier would "burn up" if you didn't "forcibly remove heat" from the hot side. I thought just air moving around a room would be sufficient.

I thought the reason people put water cooling on top of pelts "inside the case" was so that the hot-side wouldn't raise the in-case ambient temperature, not because the pelt itself would burn up.

I wasn't actually intending to use the 350watt pelt.. It was an example of price differences.

I'm thinking of another option, honestly.. I have a left over 5 gallon water cooler... That'd make a really nice reservoir..

 

[Deezus]

Found this, a bit dated but the info is good.

http://www.epinions.com/cmd-review-28A-188D6BFC-39D8E8B4-prod3

 

[DFI Daishi]

are you still thinking of doing this build?

tons of info and experience that i could kick in, if you are.

first two points:
1) pelts are cheap, the dedicated PSU that you need to power them isn't.
2) a 5 gallon watercooler has a dinky little heatpump in there that will not get the water chilled, if it doesn't actually let the water get hot, and burn out pretty quickly under continuous heat load from the computer.

 

[Postal Dude]

Yo dude. I made something very similar to what you are describing. Despite everyone telling me it couldnt be done. I put the cold side in my actual res and cooled it with an AC 64 freezer. The freezer cools the pelt quite well (110W)
I saw moderately decent temp drops. But nothin to write home about. Bout 3C load temp drop.

As for what u said with the anti-freeze, if the water goes below dew point. u will get condensation.

Go with what ppl are saying mate. Stick a 226W (or more) on a cold plate, then on your cpu. Water cool the hot side with a decent water coling system. Stuff some dielectric grease in your socket, neoprene underneath the socket with more grease, And neoprene round the outside of the socket. Pelting the rad will not give you hardly any results. Nothing anywhere near worth the money

 

[Malogato]

Thank you everyone for your comments. I am not putting a pelt, or any potential condensation maker inside my case, no matter what precautionary steps are taken. Yes, I realize that the result will be less efficient than if I had the pelt right on the cpu...

Currently, I am looking at making a very large aluminum cold plate with an integrated TEC (or set of them) that will "freeze" this "Giant flathead nail"

I already have the aluminum, and, hell, access to the metalworking tools.

Anticapated capacity of water-bottle reservoir after all this metal is placed in it.

4 gallons

so a full 20% of the bottle will be filled with chilled metal.

In my office, I have an unusual situation where I have a half-depth filing cabinet that I have flush with the other storage, so.. "behind" the filing cabinet is a good place to keep the water bottle reservoir, pump, and radiator out of site. - All that you'll really see is two clear tubes coming out, along the wall and going into my computer.

That "chickenwire" area won't be a hard, solid block.. It's meant to be something like "netting" that I can stuff into the bottle's (smaller) neck, and then have it spread out like an umbrella opening up.

The 12-15" Top of the "nail head" will be warm, as the TEC's will be sandwiched in there. (in addition to the one on the radiator, folks)

I don't know how many tec's will be needed to make this giant "cooler core" but I am sure it is cheaper than a $300 water chiller which uses freon.

the pump (after radiator) will drop the slightly warmer water into the top-ish area of the bottle..
and the "cold" outflow will be located near the bottom for gravities aid in assisting the outflow.

If this turns out to be a bad design (what, with a plastic reservoir and all) then I'll adapt a beer keg.

I'm looking at pelts because I can get them a LOT cheaper than a water chiller.. anything decent, my experience, starts at the $300 range.

 

[Malogato]

Just to explain why I'm considering this way..

I'm very paranoid about having anything inside the machine itself that "Isn't natural" - I mean, hell, I'm not even putting the eheim pump inside because I'm worried about the heat it produces.

I'm going to have the two tubes (in and out flow) going in, water blocks, and the criticool, and that's it..

 

[2uantuM]

You do realize that nomatter where you put the pelt you're probably gonna get condensation, right? Condensation occurs where warm meets cold. If the cold water is running through the tubes, the outside of the tubes can get condensation. You can make as fancy of a design as you want, you're not gonna get anywhere without insulation.

 

[Malogato]

You do realize that nomatter where you put the pelt you're probably gonna get condensation, right? Condensation occurs where warm meets cold. If the cold water is running through the tubes, the outside of the tubes can get condensation. You can make as fancy of a design as you want, you're not gonna get anywhere without insulation.

You do realize that I won't really care if I get condensation 2 feet away from the computer chassis, right?

The tubes themselves will be insulated as they go into the case. I'm not expecting to drop the temps to -40, or, hell, for that matter, even +10.. My basic concept is to be able to more easily maintain 25-30 that is the same as my room ambient.

If you are going to try and tell me that this is going to cause condensation AT THE PROCESSOR, I'll be confused. as my cpu will already be at that temp on air.

I'm not trying to lower the temperature my cases ambient temp to something silly.

I'm trying to make sure that when the processor would want to run at 80c, it stays at 30c.

 

[DFI Daishi]

I'm not trying to lower the temperature my cases ambient temp to something silly.

I'm trying to make sure that when the processor would want to run at 80c, it stays at 30c.

true enought as you have stated things, however, you are trying to do something MUCH sillier than chilling yout processor to -20.

you're looking at dropping around $170 US for a pair of dedicated 12 volt power supplies to run a few of those moster pelts you linked (@about $22 each, also added to your budget), at least $45 for a decent waterblock, $70-80 US on a good DC pump (or what say $40-50 for a reasonable AC pump) , add a tubing, clamps, whatever you plan to cool the pelts with (which you "nail" diagram does not indicate). that puts your budget up over $300, all inclusive.

so, why are you complaining about paying for a freon water chiller? you're talking about spending a whole lot of money, to build something that will not work as well as a pre-made unit costing roughly the same, that will work a whole lot better. that is just silly.

as far as keeping a processor in the 30's........even hear of water cooling? straight up water cooling, with high-end parts can get you those kinds of results. i mean, you talk about being worried about the heat dump of having an eheim pump in the case being a problem, which means that you have done very little reading on the subject of watercooling. i would suggest that you do so. you can have nice opperating temps, with a whole lot fewer complication than your proposed loop.

 

[Malogato]

you're looking at dropping around $170 US for a pair of dedicated 12 volt power supplies to run a few of those moster pelts you linked (@about $22 each, also added to your budget), at least $45 for a decent waterblock, $70-80 US on a good DC pump (or what say $40-50 for a reasonable AC pump) , add a tubing, clamps, whatever you plan to cool the pelts with (which you "nail" diagram does not indicate). that puts your budget up over $300, all inclusive.

I don't know where you get your #'s from

I have a radiator, and eheim 1250 pump that I paid $45 shipped for.
(and that's got a criticool, too)
Don't need to pay $170 for a pair of dedicated 12volt psu's

The money spent in a waterblock would be spent if I was chilling, too, so that's not "extra" - I would need a waterblock for a regular watercooled setup. As well as the tubing, clamps, etc..

The cost of the water "chiller" is $300 OVER AND ABOVE the cost of the watercooling hardware.

I am looking at spending $40 on psu's, not $170.
and $44 on TEC's - MAYBE that much.

The giant 15" nail head would.. uh, be a heatsink, ya think? a $9-$13 box-fan with 1000CFM blowing through it will do quite nicely [such as http://www.walgreens.com/store/product.jsp?CATID=100476&navAction=jump&navCount=0&id=prod1225333]

So.. barring the cost of the aluminum... (free.. )
I would be spending $44 on tecs, $40 on psus, and say, $13 on a fan to blow over the giant pizza pie sized heatsink. so.. what, $100 vs $300?

 

[DFI Daishi]

I don't know where you get your #'s from

I have a radiator, and eheim 1250 pump that I paid $45 shipped for.
(and that's got a criticool, too)
Don't need to pay $170 for a pair of dedicated 12volt psu's

The money spent in a waterblock would be spent if I was chilling, too, so that's not "extra" - I would need a waterblock for a regular watercooled setup. As well as the tubing, clamps, etc..

The cost of the water "chiller" is $300 OVER AND ABOVE the cost of the watercooling hardware.

I am looking at spending $40 on psu's, not $170.
and $44 on TEC's - MAYBE that much.

The giant 15" nail head would.. uh, be a heatsink, ya think? a $9-$13 box-fan with 1000CFM blowing through it will do quite nicely [such as http://www.walgreens.com/store/product.jsp?CATID=100476&navAction=jump&navCount=0&id=prod1225333]

So.. barring the cost of the aluminum... (free.. )
I would be spending $44 on tecs, $40 on psus, and say, $13 on a fan to blow over the giant pizza pie sized heatsink. so.. what, $100 vs $300?

where i get my numbers: http://www.sidewindercomputers.com/samam32inpsu.html which is the best price i could find quickly, for a US store. i mainly track canadian prices, however this is the best that i found after a quick look.

http://www.dangerdenstore.com/product.php?productid=90&cat=32&page=1
http://www.swiftnets.com/products/S320-12KIT.asp

were other sites i looked at. where have you found a dedicated 12V PSU with better than 300 watts of output, in order to run each 270 watt pelt, for that kind of price? i was also under the impression that you were planning on using multiple pelts, hence i double the price of a single unit, at the best price i found. maybe i'm wrong on that count.

i don't know what you are thinking that you need the rad for, exactly. good that you have one kicking around. that's some kind of wicked deal that you found on that pump and rad. that pump normally retails for around $70. to have gotten both a rad a that pump for that price, with a relay controller thrown in rather blows my mind.

as for the nail head acting as an effective heatsink......there is a reason that heatsinks have fins, you know?

as i am understanding your diagram at the moment, i'm seeing you having a few pelts on the nail head, chilling that down, and conducting the cold down into the wire mesh, which provides ample surface area to chill the water. fine. now, what do you do with the hot sides of the pelts? you need some way to get rid of at least 300 watts of combined pumped and waste heat for each pelt. to run the pelts in their most effective range, you need to keep that hot side well below 50 C, preferably below 40 C. that means that you need quite a lot of surfance area, probably in the form of finned copper.

 

[hughJ]

You do realize that nomatter where you put the pelt you're probably gonna get condensation, right? Condensation occurs where warm meets cold. If the cold water is running through the tubes, the outside of the tubes can get condensation. You can make as fancy of a design as you want, you're not gonna get anywhere without insulation.

What he said^^

If you're running chilled water below ambient(case temp), you risk condensation as your dewpoint can vary anywhere between 5 and 25 C (depending on humidity) below your case temp.

If you're running water AT (or above) ambient(case temp), than your performance is only going to be marginally better than regular water cooling with an air-cooled rad.

 

[Malogato]

as for the nail head acting as an effective heatsink......there is a reason that heatsinks have fins, you know?

as i am understanding your diagram at the moment, i'm seeing you having a few pelts on the nail head, chilling that down, and conducting the cold down into the wire mesh, which provides ample surface area to chill the water. fine. now, what do you do with the hot sides of the pelts? you need some way to get rid of at least 300 watts of combined pumped and waste heat for each pelt. to run the pelts in their most effective range, you need to keep that hot side well below 50 C, preferably below 40 C. that means that you need quite a lot of surfance area, probably in the form of finned copper.

Maybe I should elaborate.
"Nail Head" = "Circle with a bunch of spikes all over it"

i.e. "Bed of nails"
Most likely, by the time it's done, there should be around 7000 'nails' that are 3-4 inches tall.

I have access to a cnc machine that can cut this out of pretty much any material I provide a block of. (within reason) .. I happen to have the aluminum.

Think something like This.. but pizza pie sized, with the same size "spikes"
my calculations about quantity of nails might be off some.
Yes, it's going to be heavy, that's why I am opting for aluminum instead of copper.
as for the airflow over it? I think I mentioned a box fan.

That is, of course, if I find that I even need to go that large.
I'm initially going to start with 1 piece.. maybe CD sized, and one peltier in it.


Oh.. and the rad, pump, and criticool were bought here, on the forums.. $45 shipped.


Why the rad? Well, cause, simply.. why not? I don't imagine it could "hurt" the functionality of the system to have the water additionally air cooled through a radiator before being cooled in the waterbottle reservoir..

Like I said, I'm not shooting for -50c .. I'm shooting for maintaining 20-25c at all times.
If my water, and case, and room are all 20-25c.. I don't see how condensation will form.

I don't know how many times I have to say that I'm not going to be shooting for "below case temp".. I'm aiming for "near" case temp.

If I Can keep the water at 20-25c inspite of the fact that the cpu would want to normally be at 70c, then I don't see how condensation would form.

The only place I can imagine condensation is at the "nailhead" which, as explained before, would not be an issue.


at overclock "X" on plain air, I might sit at 52c
on plain water, I might sit at 45c [the waters temp itself will rise to 45c]

With the right quantity of pelts.. I can keep the water at 25c.

 

[DFI Daishi]

regarding weight: didn't even bring that up.

regarding that swiftech sink: okay, that's a decent 'sink, although thermalright's designs have been doing better with copper base and copper plate-fins, for some time now.

airflow: i didn't bring up ariflow. to cool a 226 watt pelt and an 80 watt pelt, plus system load, two 120.2 rads are not unreasonable, to maintain decent noise levels. i was able to do it with a single 120.2, with loud, 2x115 CFM fans. who knows how your custom sink will fare as compared to a decent rad?

the rad: assuming pretty average water flow rates, you're only looking at a degree or so of temperature difference from the hottest part of the loop to the coldest. if you have enough pelt pumping capacity to handle the load of your components, then the rad will only hurt performance, bringing chilled coolant back up closer to room temperature.

given the temps that you are talking about, i assume that you are running a prescott. with a budget that i would consider to be typical of a pelt cooled solution, purchasing top-end watercooling parts, i think that you could reasonably expect to keep a prescott below 40 C load. you are getting a lot of stuff used, so i'm not so sure how much that that will affect you final budget.

all of that being said, i think that you will need more than a single pelt and more than a single PSU to keep your desired temps. the theoretical heat output on my old barton, while overclocked, was about that of a stock prescott, and with a 226 watt pelt and coldplate, i was only able to keep temps down to around 20 C under load. given that your proposed system is going to get less out of each pelt used, i think that it would be highly unlikely to get your stated target of 25 C with a single 270 watt pelt, a larger heat load, and higher hot-side temp than i was running.

i'm not making empty jabs at your proposed design, i'm trying to help you to come up with something workable, and hopefully with reasonable cooling value, for the money spent.

 

[Postal Dude]

I think your design and idea's look cool mate. Like i said before i went thru the same sort of process to chill my water.. view this thread http://www.techimo.com/forum/t144360.html
or this one http://www.hardforum.com/showthread.php?t=913176 (only slight difference in posts and opinions)
I have had it set up and running. And im tellin ya. I wish i hadda jus stuck a pelt on a cold plate on my CPU. The effort and time taken to do it all, in relation to the results. It jus wasnt worth it. It was fun to build. Taught me alot about pelts and stuff. But results were not that great.

The temperatures i reached on them everest shots were from leaving pelt running with the pc off for a couple of hours to get really cold water in my res. Then booting up. Condensation everywhere! within an hour my temps were jus a couple of degree's below usual. Load temps went from 43 to 40. Whoop dee doo.

If you really wanna try it to find out results. Go ahead, its fun. But you'll be disappointed. So much so i wanna take extra res out of loop. The pelt dont even run anymore

 

[Malogato]

regarding weight: didn't even bring that up.

regarding that swiftech sink: okay, that's a decent 'sink, although thermalright's designs have been doing better with copper base and copper plate-fins, for some time now.

I wasn't actually intending to use that swiftech sink, it was just an example of what my creation would be.

airflow: i didn't bring up ariflow. to cool a 226 watt pelt and an 80 watt pelt, plus system load, two 120.2 rads are not unreasonable, to maintain decent noise levels. i was able to do it with a single 120.2, with loud, 2x115 CFM fans. who knows how your custom sink will fare as compared to a decent rad?

Guess you missed, twice, the "1000cfm box fan' concept.

the rad: assuming pretty average water flow rates, you're only looking at a degree or so of temperature difference from the hottest part of the loop to the coldest. if you have enough pelt pumping capacity to handle the load of your components, then the rad will only hurt performance, bringing chilled coolant back up closer to room temperature.

I don't understand how the rad can bring temps back up.
It's going to after the heat load, and before the reservoir. It's purpose is to lower the temps of what's being dumped into the waterbottle, not to lower the temps of the water being dropped over the blocks.

given the temps that you are talking about, i assume that you are running a prescott.

Nope. A64 - I'm currently running @ 27, but that's because I'm not doing any serious overclocking. I anticpate however, when I decide to try for 3.3ghz that the temps will want to, naturally, rise. My goal is not to drop my cpu to -50, but to maintain room temperature even with a 50% speed bump. [I'm shooting for 300x11 instead of 200x11 with a 3:2 memory divider, leaving the ram at 200 the entire time]

As for overclocks i'm comfortable with, I've already hit 270x9 on low cfm AIR, and no voltage bumping. I know 300x11 is a far ways along from 270x9, (3.3 ghz vs 2.43ghz] but that was accomplished at only a 6 degree temp boost.
280x10 was acheived at a 15degree temp boost, but I am not happy with "Mid 40s"

i'm not making empty jabs at your proposed design, i'm trying to help you to come up with something workable, and hopefully with reasonable cooling value, for the money spent.

I realize this, I didn't think you were "Flaming" Me, i'm just trying to point out the things I think you are missing in my idea. I don't know the math of how much pelting I need.. My concept however is that I think I can start with 1.. and see what it does to the temp at stock..
slowly up the cpu's core temp.. and see how effective the pelt is at keeping room temp.

If I can only get to say.. 3.0ghz, before I hit 45c, then I'll have to add pelt #2
I don't want to do LN2, and the rest. also, like I said, I think with my design, the condensation issue will not be "inside my case".. I think the condensation issue will definately exist.. [I am envisioning the bottom of the nail head being forever damp] but that's a couple feet away from the case proper.

I'm thinking the eheim1250 has the chutzpah to drive this. There is that to consider as well.. I might be wrong about the 1250... But, that's ok.. I have a pond pump [24/7 duty cycle] that has an 850GPH flow [1" barbs though..ugh.. finding those adapters will be fun]

 

[Malogato]

I think your design and idea's look cool mate. Like i said before i went thru the same sort of process to chill my water.. view this thread http://www.techimo.com/forum/t144360.html
or this one http://www.hardforum.com/showthread.php?t=913176 (only slight difference in posts and opinions)
I have had it set up and running. And im tellin ya. I wish i hadda jus stuck a pelt on a cold plate on my CPU. The effort and time taken to do it all, in relation to the results. It jus wasnt worth it. It was fun to build. Taught me alot about pelts and stuff. But results were not that great.

The temperatures i reached on them everest shots were from leaving pelt running with the pc off for a couple of hours to get really cold water in my res. Then booting up. Condensation everywhere! within an hour my temps were jus a couple of degree's below usual. Load temps went from 43 to 40. Whoop dee doo.

If you really wanna try it to find out results. Go ahead, its fun. But you'll be disappointed. So much so i wanna take extra res out of loop. The pelt dont even run anymore

No offense, P.D., but you say you had a pair of 80watt pelts for this? I'm starting with a 300watt, and going from there. Also, I'm going to be having a much larger water source to dissipate heat through, so I think the thermal delta might be easier to manage.. [I envision that your ammo can had, what.. a 3/4 gallon capacity? I'm looking at 3 gallons of water. ]

 

[DFI Daishi]

I wasn't actually intending to use that swiftech sink, it was just an example of what my creation would be.

and what i was trying to point you is that you might be able to get better results with a thick copper base plate and thin copper plates soldered onto it to make fins, than an aluminium base plate with all sorts of spikes coming up and off of it. you might need enough radiating power and premitivity that copper is the only way to go, and copper isn't so easy to machine. there is also the zalman type process of bolting a whole series of copper plates together, two different lengths.

Guess you missed, twice, the "1000cfm box fan' concept.

didn't miss at all, actually. very nearly all of the 230 CFM that i was moving got forced throught the fins of my rad, there was nowhere else for it to go. the majority of the air that your boxfan moves will be going aroud the 'sink and fins, rather than through them. the fans is going to be bigger than the 'sink, and air follows the path of least resistance. you would get more out of that 1000 CFM if you had a fan that was the same diameter as the 'sink, and strapped to it with a duct to force the air through the fins.

I don't understand how the rad can bring temps back up.
It's going to after the heat load, and before the reservoir. It's purpose is to lower the temps of what's being dumped into the waterbottle, not to lower the temps of the water being dropped over the blocks.

that's the thing: even the "hot" water coming back from the block is going to be at most 0.5 - 1C warmer than the "cold" water coming out of the res. since you should be shooting for water temps around 5 C lower than your targeted CPU temps, having the rad in the loop will only hurt matters, if you have enough pelt power to attain your goal in the first place.

Nope. A64 - I'm currently running @ 27, but that's because I'm not doing any serious overclocking. I anticpate however, when I decide to try for 3.3ghz that the temps will want to, naturally, rise. My goal is not to drop my cpu to -50, but to maintain room temperature even with a 50% speed bump. [I'm shooting for 300x11 instead of 200x11 with a 3:2 memory divider, leaving the ram at 200 the entire time]

if you have the budget to do a weird, wacky, and expensive build like this, i should think that you can afford some decent overclocking ram. running 1:1 doesn't give the gains that it used to, so a lot of reviewers have said that ram should be the last thing that you upgrade. that being said, as compared to a loop that keeps your proc at 25 C, overclocking ram is more likely to see a concrete performance gain for the money.

As for overclocks i'm comfortable with, I've already hit 270x9 on low cfm AIR, and no voltage bumping. I know 300x11 is a far ways along from 270x9, (3.3 ghz vs 2.43ghz] but that was accomplished at only a 6 degree temp boost.
280x10 was acheived at a 15degree temp boost, but I am not happy with "Mid 40s"

take a good, hard, look at the swiftech apex kit. it keeps a64s cool enough to get as much out of them as you are going to be able to sustain.

I'm thinking the eheim1250 has the chutzpah to drive this. There is that to consider as well.. I might be wrong about the 1250... But, that's ok.. I have a pond pump [24/7 duty cycle] that has an 850GPH flow [1" barbs though..ugh.. finding those adapters will be fun]

so long as you have a reasonably unrestrictive block like the swiftech MCW6002, the 1250 should be good enough to get you into the optimal flow range. having a pump with more balls might actually hurt your performace, because it actually dumps more heat into the water, which can negate the marginal gains of increased flow as compared to what the 1250 puts out.

No offense, P.D., but you say you had a pair of 80watt pelts for this? I'm starting with a 300watt, and going from there. Also, I'm going to be having a much larger water source to dissipate heat through, so I think the thermal delta might be easier to manage.. [I envision that your ammo can had, what.. a 3/4 gallon capacity? I'm looking at 3 gallons of water.]

volume of water means bugger all in terms of opperating temps. once the system has been running for a while, and temps are at equilibrium, the only factors that matter are coolant flow, heat pumping power and thermal permitivity. the only advantage that you might see with your large res if if you run the pelts of a good, long while, with the computer off, to chill down that water colder than your sustainable opperating temp, to allow for a brief period of extra cooling immediately after startup.