A Primer on Water Cooling

[Zarathustra[H]]

While I have been overclocking since the 286 days (first overclock was my 8mhz 286 @12mhz, best percent wise OC I ever got was a 486 sx 25mhz @ 50mhz, you just don't get doubles anymore ) I have never dabbled in water cooling.

In my old age, I'm starting to have a newfound desire for silence. I have dabbled with fanless passive cooling, but the performance leaves something to be desired, so I am considering adding water cooling and rigging up a completely fanless SLI system, with a water reservoir in the closet, where it is out of earshot.

How are these things typically connected? With a CPU and two GPU's would you do some sort of triple split before them to make sure each gets fresh, cool water, or is it OK to run them in series? I have also seem people cool chipsets... Is this necessary?

I have an old fridge available I was going to stick in the closet as well, is there a real benefit in cutting holes and sticking my reservoir inside it, or am I just asking for trouble with condensation having cool water going to the computer?

Does anyone have a link on a good guide for n00bs to follow in rigging up their first water cooled system?

Much appreciated,
Matt

 

[ghost6303]

there are plenty of stickies at the top of this forum with a lot of the information you are looking for.

With a CPU and two GPU's would you do some sort of triple split before them to make sure each gets fresh, cool water, or is it OK to run them in series?

this is a very common misconception- the placement of the radiator in the loop or trying to feed the 'coldest' possible water to the CPU is really moot. the water temp in any given location in an average cooling loop wont varry by more then a few single degrees. if it really matters to you, there is no harm in making sure the radiator feeds directly into the CPU. but if its easier to route the water lines another way- again no harm no foul.

have an old fridge available I was going to stick in the closet as well[...].

refrigerators are only designed to maintain a low temperature on something (food) that doesnt produce its own heat. if you start dumping 300+ watts of heat into the refrigerator, it will not be able to maintain those temperatures and will burn out quickly.

people have done something similar with air conditioning units turned into water chillers, but as you said, then you have to take into acount the dew point of air and make sure to avoid condensation. and you also said you are trying to eliminate noise, sitting right next to an air conditioner is counter productive here.

I have also seem people cool chipsets... Is this necessary?

necessary- no. generally the better your cooling capabilities, the more voltage you can pump thru electronics, and the faster you can get them to run. it depends greatly on the specific motherboard if this is necessary or not, and what you intend to do with it (how high of an overclock are you looking for). the computer im building right now will have a water cooled motherboard just because i want to remove as much heat as possible.

 

[Zarathustra[H]]

there are plenty of stickies at the top of this forum with a lot of the information you are looking for.

Yeah, thank you, I looked at some of them, but they seemed more focused on what equipment to buy, and less focused on the more overall system questions I was asking.

this is a very common misconception- the placement of the radiator in the loop or trying to feed the 'coldest' possible water to the CPU is really moot. the water temp in any given location in an average cooling loop wont varry by more then a few single degrees. if it really matters to you, there is no harm in making sure the radiator feeds directly into the CPU. but if its easier to route the water lines another way- again no harm no foul.

Thank you for clearing that up. So I presume you hvae everything routed in series? I ahve heard of some people suggesting multiple lines to deal with equipment giving off a lot of heat. Are there advantages to this over just using one line with a beefier radiator?

refrigerators are only designed to maintain a low temperature on something (food) that doesnt produce its own heat. if you start dumping 300+ watts of heat into the refrigerator, it will not be able to maintain those temperatures and will burn out quickly.

people have done something similar with air conditioning units turned into water chillers, but as you said, then you have to take into acount the dew point of air and make sure to avoid condensation. and you also said you are trying to eliminate noise, sitting right next to an air conditioner is counter productive here.

Fair enough about the refrigerators. I pictured having a large enough reservoir in it that it would average out the work the fridge had to do over time, I can see what you are talking about though. I did some research, and there seems to be some opportunity with some relatively inexpensive fish-tank cooling systems!

necessary- no. generally the better your cooling capabilities, the more voltage you can pump thru electronics, and the faster you can get them to run. it depends greatly on the specific motherboard if this is necessary or not, and what you intend to do with it (how high of an overclock are you looking for). the computer im building right now will have a water cooled motherboard just because i want to remove as much heat as possible.

Good to know.

Thank you very much for your input.

 

[King Icewind]

Reservoirs don't make any sound so there isn't any need to put it in the closet.

What cards are you going to SLI? Two 9400's isn't worth it. Not even sure if there is SLI 9400's.

Putting the cards in one loop and the CPU in another will help keep the heat down. Just depends if you have the space for that or not.

 

[ghost6303]

Putting the cards in one loop and the CPU in another will help keep the heat down. Just depends if you have the space for that or not.

only if you are planning on two completely separate loops. but that entails buying two of everything. and in a mid tower case, there is almost always no room for two pumps, two radiators, two reservoirs, etc.

So I presume you have everything routed in series? I ahve heard of some people suggesting multiple lines to deal with equipment giving off a lot of heat. Are there advantages to this over just using one line with a beefier radiator?

i have had many many loop setups over the years. ive taken my system apart so many times in the last few months, that all i have on water currently is 1xmcp355 pump going into 1 north bridge block split and then two 1/4" hard drive coolers to cool a cluster of a half dozen hard drives, then back into an open tank. no radiator. evaporation takes care of that little bit of heat.

the only reason i would split one loop into two or more smaller loops would be if the pump i was using was not up to the task, or (as i have done in the past) wanted to run one small dedicated pump thru a very large radiator, then a separate high flow pump to push thru the cpu/gpu blocks. the radiator is normally the biggest source of water turbulence and pressure drop in any loop.

because of that i used my mcp355 pump, which has a very high pressure output and 3/8" fittings, to pump thru the massive radiator and then right back into the reservoir. then i used a D5 (mcp655) pump which doesnt have as much head pressure, but flows more water and uses 1/2" native fittings, to drive the CPU and GPU blocks, then back to the reservoir.

also before i have used a 5 gallon bucket as a reservoir with no radiator. if you are surfing the net/emailing, the computer didnt generate enough heat to need a rad. but if i were doing some hardcore gaming, i would throw a half gallon block of ice in it every once in a while (or frozen water bottle). there is enough of a thermal buffer in 5 gallons of water to absorb a great deal of heat for a good amount of time without need to radiate it anywhere. it would take a full 12-24 hours to warm up to stable operating temp. if you only have your computer on for a few hours at a time, that would be perfect.

you can literally set it up 1000 different ways. it pretty much boils down to what is easiest for you. if you had 5 single 120mm radiators, you could put one radiator between each waterblock if you really wanted to.

 

[Zarathustra[H]]

Reservoirs don't make any sound so there isn't any need to put it in the closet.

What cards are you going to SLI? Two 9400's isn't worth it. Not even sure if there is SLI 9400's.

Putting the cards in one loop and the CPU in another will help keep the heat down. Just depends if you have the space for that or not.

Not even sure which boards I'll be using yet. The 9400 will go. Its just in there for my temporary passive cooling. Heck, I haven'tneven decided if I'll be SLI:ing yet. I just want something reasonably powerful, and able to be perfectly silent...

I was looking at boards like the EVGA 480GTX's that come with a factory water block installed on them...

As far as the reservoir goes, the thought is that hsi would be a completely external unit. I ahve one of those shuttle small form fator cases so NOTHING is going to fit in the case.

The plan is to put pump, reservoir, radiator, etc. etc. all in the closet, or outside the room somewhere, and have two little tubes running to the case for the inlet and outlet water.

 

[ghost6303]

btw- if you do the math you can save ~$80 or more if you buy a regular 480 and a waterblock separately and just put it together yourself in half an hour over buying it pre-assembled.

 

[King Icewind]

Not even sure which boards I'll be using yet. The 9400 will go. Its just in there for my temporary passive cooling. Heck, I haven'tneven decided if I'll be SLI:ing yet. I just want something reasonably powerful, and able to be perfectly silent...

I was looking at boards like the EVGA 480GTX's that come with a factory water block installed on them...

As far as the reservoir goes, the thought is that hsi would be a completely external unit. I ahve one of those shuttle small form fator cases so NOTHING is going to fit in the case.

The plan is to put pump, reservoir, radiator, etc. etc. all in the closet, or outside the room somewhere, and have two little tubes running to the case for the inlet and outlet water.

I wouldn't recommend putting the radiators in the closet. There won't be any fresh air, or air flow at all. The closet will continually get hotter, and so will your temps. If you have any extra computer cases lying around you could put the rad's in the case, and then add some quiet fans to the rads. The case will help block the sound and you will get better temps. Just an idea.

 

[silent-circuit]

Remember that if you put the rad and res far from the case you may need to consider head pressure of the pump you're planning to use. Sure, theoretically a closed loop requires very little head pressure to move, but you have to consider the initial filling and bleeding of the system, which could turn in to a total nightmare. Include a pair of quick disconnects so you can easily move the system without worrying about the connecting tubing. As others have said, avoid using the closet, instead mount everything in a dedicated case -- a mid tower should be big enough, and cheap. Use quiet fans and place it some distance from where you'll be working and it shouldn't be a problem.

 

[Zarathustra[H]]

Has anyone ever tried using a low end industrial water chiller? One of those portable things they use in labs to chill rotary evaporators and other equipment?

They would seem perfect for this application. You set a thermostat on the unit for the temp you want to come out of it and then you just plug in your hoses. Reservoir and everything is built right in.

I wonder if there are any affordable surplus units. I'd get like a 1/4 ton water chiller, with a 2-3GPM pumpand stick it in a well ventilated area away from the comp. I'd probably set it to operate at ~50-60F to avoid condensation. Sounds like a perfect application to me, (but probably not cheap....)

 

[Zarathustra[H]]

Something like this:



They have phase change cooling/radiator, reservoir and pump built right in. The thermostat can be set to hold the temps anywhere from 39-86F. The tolerance isn't published on this one, but typically they claim to within anywhere from +/-0.1C to +/-0.5C.

Just hook up your in and out hose to your water cooling case and power it up You could probably even route your hoses from your garage or whatever, and stick it in there to make your computer room completely silent. You might not even need to keep a case fan if you do GPU's, chipset and CPU, and have only SSD's in there for drives... Let convection take care of cooling the ram...

The 1kw unit is probably good for 1-2 computers, the 2kw unit for 2-4...

It would be a heck of a lot more than your average water cooling setup, but MAN would it be awesome. I wonder if water chiller units like these can be found cheaply at surplus auctions and liquidations...

I'd probably set mine to ~50F to avoid condensation if operating in a comfortable air conditioned room... Higher if in a non-air conditioned humid room...

if I get a winning scratch ticket for ~$2k or so, I'm totally doing this. (it would help if I actually bought scratch tickets, which I don't)

 

[Zarathustra[H]]

the only reason i would split one loop into two or more smaller loops would be if the pump i was using was not up to the task, or (as i have done in the past) wanted to run one small dedicated pump thru a very large radiator, then a separate high flow pump to push thru the cpu/gpu blocks. the radiator is normally the biggest source of water turbulence and pressure drop in any loop.

because of that i used my mcp355 pump, which has a very high pressure output and 3/8" fittings, to pump thru the massive radiator and then right back into the reservoir. then i used a D5 (mcp655) pump which doesnt have as much head pressure, but flows more water and uses 1/2" native fittings, to drive the CPU and GPU blocks, then back to the reservoir.

Interesting. What kind of flow rating do you find you need for a typical circuit? 1gpm?

you can literally set it up 1000 different ways. it pretty much boils down to what is easiest for you. if you had 5 single 120mm radiators, you could put one radiator between each waterblock if you really wanted to.

In the interest of silence, it would be interesting to see what could be done with a series of large passive radiators, thought this would probably increase the pump load, and add more noise on that end instead, correct?

 

[Zarathustra[H]]

Remember that if you put the rad and res far from the case you may need to consider head pressure of the pump you're planning to use. Sure, theoretically a closed loop requires very little head pressure to move, but you have to consider the initial filling and bleeding of the system, which could turn in to a total nightmare. Include a pair of quick disconnects so you can easily move the system without worrying about the connecting tubing. As others have said, avoid using the closet, instead mount everything in a dedicated case -- a mid tower should be big enough, and cheap. Use quiet fans and place it some distance from where you'll be working and it shouldn't be a problem.

Is pressure only a pump-side problem, or do I ahve to worry about building up enough pressure that I start blowing hoses off water blocks and fittings and shorting my box with water?

 

[silent-circuit]

Is pressure only a pump-side problem, or do I ahve to worry about building up enough pressure that I start blowing hoses off water blocks and fittings and shorting my box with water?

That's highly unlikely, though clamps are more important in a long loop like you're planning than a more typical in-case loop. I usually run 7/16" ID tubing over 1/2" barbs and don't bother with clamps. Considering the kinds of things you're looking at and the price range you seem to be considering the extra $50-70 over barbs for compression fittings would probably be a realistic option for you. That lets you forgo clamps entirely (and they're more secure than clamps), plus they look nicer. Don't know how much that matters to you.

 

[ghost6303]

Interesting. What kind of flow rating do you find you need for a typical circuit? 1gpm?

1gpm (60GPH) is very low. for instance the MCP650 pump flows 350ish gallons per hour with no restriction. you would want atleast 2-3 gallons a minute, but it would vary depending on the block. some water blocks like low speed, nice laminar flowing water. others like high speed lots of turbulence flowing over the heat dissipation area. pin impingement blocks seem to like faster flow rates as they have a huge surface area and transfer heat very quickly.

In the interest of silence, it would be interesting to see what could be done with a series of large passive radiators, thought this would probably increase the pump load, and add more noise on that end instead, correct?

i cant access my photobucket from where i am, but in the past i had an evaporator from an industrial fridge (two parallel 18-pass 1/2" OD copper pipes, about 4 feet long) that i cut out and used as a passive radiator. it had very low restriction which was nice, it held about 4-6 gallons of water. i just stood it up in the corner of the room. there was only a single 140mm low rpm fan blowing across the mobo for some sort of air movement. nice and quiet. fans are generally louder then pumps, especially because you can always wrap a pump in soundproofing, something you cant do with a fan that needs to move air.

 

[Zarathustra[H]]

1gpm (60GPH) is very low. for instance the MCP650 pump flows 350ish gallons per hour with no restriction. you would want atleast 2-3 gallons a minute, but it would vary depending on the block. some water blocks like low speed, nice laminar flowing water. others like high speed lots of turbulence flowing over the heat dissipation area. pin impingement blocks seem to like faster flow rates as they have a huge surface area and transfer heat very quickly.

Interesting. I didn't think the waterblocks in a computer would have contributed significantly to head pressure. I was more concerned with the height of the desk, if the pump is outside on the floor.

All that taken into account though and disregarding the pump rating, what have you found to work as a final actual flow rate with all the head presssure and such taken into account? Do you use an inline flow meter to measure this?

I am aware that the flow rate is of less significance than many think, due to that any rise in water temps over the cpu's and gpus caused by slower flow will also lead to more cooling over the radiator to equal slow flows there, but we also need to take heat transfer into account, and heat will transfer more efficiently if there is a greater Delta T at the cooling location, so ideally you wouldn't let your water temperature rise much locally, and thus you'd want a higher flow rate. I would imagine this reaching the point of diminishing returns rather quickly based on the theory above with equal time in the radiator vs. the water blocks, I just don't ahve enough experience with these things to know where that point is...

I apologize if I am going over-board here, but I am not used to the home DIY setups, but more with manufacturing equipment and process chilled water

 

[silent-circuit]

Nothing wrong with that. It's interesting to get another prospective anyway. Sadly I've never bothered with flow rate measurement since the loops I'm dealing with are fairly simple (1, 2, 3 blocks, 1 or 2 radiators, 1 pump per loop) but I know others here have.

 

[Zarathustra[H]]

Nothing wrong with that. It's interesting to get another prospective anyway. Sadly I've never bothered with flow rate measurement since the loops I'm dealing with are fairly simple (1, 2, 3 blocks, 1 or 2 radiators, 1 pump per loop) but I know others here have.

It would be interesting to look into.

I just speced out a pump, but I think it may be overkill for this application. It has a relatively low flow rating (1.4GPM) but it promises to keep it at up to 60 psi, which I don't see a water cooled system ever hitting

 

[ghost6303]

Interesting. I didn't think the waterblocks in a computer would have contributed significantly to head pressure. I was more concerned with the height of the desk, if the pump is outside on the floor.

If you think about it, as much pressure as the pump is forcing out, it is equally creating vacuum on the other side. as long as you have a fully closed loop, pumping uphill shouldnt create, theoretically- any (realistically- very little) head pressure as the water has an equal amount of energy coming down as going up. head pressure matters more when you are trying to pump one way up hill, or pump thru a restriction like a water block or even a pipe with rough inside walls.

Ive never used flow meters in any of my loops. the water moving in the resevoir indicates that it is actually pumping, and to test flow rates of a particular setup i just get a bucket and watch how fast the water comes out of the tubing. kind of a subjective measurement but it works enough for my purposes. if i had to put a number on it, i would guess that 3-5 gallons/min is a good flow rate for a 1/2" ID water loop. any more and you would be getting diminishing returns, any less and you wouldnt be cooling as efficiently as possible.