Understanding & Mastering Water Cooling

[Chai]

From my experience, the less tubing you use, the less kink you are going to get, unless you are trying to work in a small case, which is very unlikely.

[Skip Da Shu]

So let's just pretend for a minute that some idiot had several mobo/cpu/hdd "computers" mounted in stackable plastic baskets. These "computers" and their air HSFs would add considerable heat to the room they were in. This in turn would cause a serious load on the A/C systems for that room. Which results in some extremely high electric bills during the long summer. What if this person wanted to build a piped water system, a single system to cool 10~12 CPUs. In this system the radiator would be mounted outside of the building and be plumbed back inside with PVC water piping. Inside the building would be a substantial AC powered water pump and a PVC based distribution system with T fittings and reducers to provide water flow to each CPU water block. Do you know of any lunatics who've already built such a system and posted info where one might read up on it?

I wouldn't even ask but I just read an article where a guy actually pumps water out of his swimming pool to a room in the upstairs of his house and cools 4 computers with it!

Too bad I don't have a pool.

I'm.. er, ah, I mean.. Suppose someone was thinking of using an old car radiator and a standard discount store 2/3m sized box fan to blow air threw it. Maybe a small swimming pool pump to move the water around. Such a contraption would so fit the motif of my website!

Ever hear of such a thing? Could it work? -- Skip

PS: See what y'all got started with your article! See information IS dangerous

[Chai]

Originally Posted by Skip Da Shu

So let's just pretend for a minute that some idiot had several mobo/cpu/hdd "computers" mounted in stackable plastic baskets. These "computers" and their air HSFs would add considerable heat to the room they were in. This in turn would cause a serious load on the A/C systems for that room. Which results in some extremely high electric bills during the long summer. What if this person wanted to build a piped water system, a single system to cool 10~12 CPUs. In this system the radiator would be mounted outside of the building and be plumbed back inside with PVC water piping. Inside the building would be a substantial AC powered water pump and a PVC based distribution system with T fittings and reducers to provide water flow to each CPU water block. Do you know of any lunatics who've already built such a system and posted info where one might read up on it?

I wouldn't even ask but I just read an article where a guy actually pumps water out of his swimming pool to a room in the upstairs of his house and cools 4 computers with it!

Too bad I don't have a pool.

I'm.. er, ah, I mean.. Suppose someone was thinking of using an old car radiator and a standard discount store 2/3m sized box fan to blow air threw it. Maybe a small swimming pool pump to move the water around. Such a contraption would so fit the motif of my website!

Ever hear of such a thing? Could it work? -- Skip

PS: See what y'all got started with your article! See information IS dangerous

Nothing is impossible, but it requires a lot of work, everything has to be much bigger, radiators, pumps, and a lot of maintainence work too.

I've only seen this.

http://www.dwpg.com/content.php?contid=3&artid=57

[Skip Da Shu]

Originally Posted by Chai

Nothing is impossible, but it requires a lot of work, everything has to be much bigger, radiators, pumps, and a lot of maintainence work too.

I've only seen this.

http://www.dwpg.com/content.php?contid=3&artid=57

Thanx, Reading it now... before you declare me nuts read this...

http://www.thebuehls.com/pool_cooling/

[neverthar]

Thank you all for reading my article and do appreciate much the feedback generated. I'm using this account which I established sometime ago so bear with me about it. There are a few misconceptions about this article which i'd like to address.

First about pre-packaged kits which many presumed I slagged in the beginning of the article. I wasn't slagging it but stating that for high performance usage, time and time again, many users have ended up disappointed with the results obtained. If for ease of maintenance and silence while at stock speeds and largely unstressed CPU's, then it would be ok to use. However, this would defeat the purpose of going watercooling as one would imagine much better temperature reductions most critically at full load which remains an issue. In fact many of such kit users have ended disappointed and gone back to aircooling. After all, if using let's say a Scythe Ninja with a silent fan nets a similiar result, which would you choose?

I'm talking about higher levels of performance demanded by enthusiasts who are into overclocking and high frequency processors which to develop a substantial amount of heat (For example, highly clocked Allendales/Conroes and massive heat dumping FX's). No doubt though those who choose to persevere using these kits can obtain some real gains but that would require a good machinist handy in your speed dial to be honest. You are all welcome to ask questions about improving such kits. Be glad to assist so as long it's constructive and clear.

As for Olle's comments, here's my counter arguments.

Firstly tubing, kinking can be combatted as seen in many online examples by merely using some form of tubing stiffeners like Swiftech sells or if on the DIY, zip ties which do the job very well actually. I've had someone who's using a WC'ed SLI setup (Utilising Maze 4) who had kinking problems due to his bending tygons cured by using zip ties to correct the diametral shape of the kinked tubing. Turbulence for short sharply turned tubings is considered negligible. Otherwise Dual PEG WC'ed setups would be suffering from such phenomenons.

As for shortened tubings, if one uses high flow and large pump head pressure pumps, then such gains are not realised. However, those on a budget using AC based pumps such as the Hydor L25/Nirox P2800, the gains are more obvious as these pumps do not have as much pump head as for example the Laing D5 or the DDC and therefore reducing the length of travel between pump and waterblock does reduce the pump head loss encountered. My observations were done using the following setup (note this was done 4 years ago as I've moved on since then):

AXP-M 2500+ (250*10 @ 1.725v)
Dangerden Maze 4
HWLabs Black Ice Pro
Nirox P2800
Dangerden tygons of 1/2" ID

Before

http://img.photobucket.com/albums/v4...C/DSC03017.jpg

And something like this after (the repostioning of the rad was done sometime later)



No remount of the waterblock was done (after all, who wants to disassemble and remount a Maze 4?) to maintain certain fixed variable. A mean of 5 average readings were done using MBM5 and Prime95 Small FFT test before and after the change in tubing length from 10 inches to 5 inches (My case is capable of converting from ATX to Semi-BTX, in other words I'm using the CM Stacker). And this was confirmed a workable tweak for various DIY kit users which used blocks ranging from a cascade style jet impingement block to a Swiftech MCW6000 and paired to budget end pumps.

As for the reservoir, do read again what I wrote as it was in reference to a particular reservoir made by person who went by the nick of Bulkhy as there might be still some who use it. Some have also replicated that design to save cost using the packaging Glade uses for the car air fresheners. However those who use CD bay style or MicroRes style reservoirs would not need to worry about it. And as for restriction, it's also considered negligible as first of all, the norm of arrangement would have one placing the reservoir rather far in perspective from the last component of the loop (usually the radiator) and hence thereforth backpressure wouldn't be quite a lot in comparison. Secondly, you wouldn't be starving the supply of water of the loop as a reservoir is what it's meant to be, to store excess water. We're just trying to combat air bubbles which if in sufficient quantitties would cause air lock which would be more difficult to solve.

Sorry for any typo as I couldn't use the Reply function and am typing this out in the Quick Reply window. Somemore my boss would kill me if he saw me spending too much time typing this

[Olle P]

The fan mounted just above the cooling block in the "after" picture isn't there in the "before" pic.
If that is because it didn't exist "before" then I suppose that could explain a lot.
Then another aspect is that IIRC the Mace 4 is a block that require a pretty high flow to function properly (get enough turbulence inside). Perhaps your adjustment was what it took to push the flow from "just below" to "just above" that limit, and had you had another type of block the difference would have been negligible.

... reducing the length of travel between pump and waterblock does reduce the pump head loss encountered.

Reducing the total tube length in the entire system does reduce the flow restriction, and that's what matters.
If you manage to shorten the tube before the block by adding the same (or more) length after the block there won't be a positive difference in the water flow.

Then there's also the issue about where to place the radiator to optimise all factors involved: Tube length (restriction), air flow, air temperature, audible noise, size, weight and mobility.

/Olle

[neverthar]

Thanks Olle for the observation but it wasn't. I did put in the caveat whereby the radiator was repositioned sometyime later. The fan added on had nothing to do with cooling my proc but rather to cool the PWM of my nForce2 which was getting hot while trying to achieve overclocks using BH5's last time.

Indeed the Maze 4 has some flow restrictions, but then again such flow design blocks are what the common enthusiasts would be using unless he/she had higher aspirations.

And for the tubing issue if post block the length was increased then yes. But in my case, it wasn't. I'm very aware of changing too many variables would affect my observations.

As for the rest of the factors you've mentioned, I did relaise that and later had my whole setup revised when I was using a 754 Athlon 64. Basically the whole setup binned for a new one which is the following

Swiftech Storm G4 Rev. 1.0
Swiftech Silent series radiator (double row) = A replacement by a mate for damaging my BIP
Cyber Aqua AC pump (Rated 3m HMax, 2650 lph QMax)
Panaflo 120mm*38mm fan (H series)

and the radiator is mounted outside the case. Such is the flexibillity of my casing

[Olle P]

Originally Posted by neverthar

The fan added on had nothing to do with cooling my proc...

Not intentionally, but as a matter of fact.
The fan does produce an airflow on the cooling block, which does improve the CPU cooling.
How much of an improvement I can't say.

Originally Posted by neverthar

... such flow design blocks are what the common enthusiasts would be using unless he/she had higher aspirations.

Exactly! And they all have different "sweet spots" in the flow chart. A typical flow/efficiency relation can be seen in the attached image. Users should have a flow above 2, the exact value of which differs between models, and my guess is that you started out closer to 1.

Originally Posted by neverthar

And for the tubing issue if post block the length was increased then yes. But in my case, it wasn't. I'm very aware of changing too many variables would affect my observations.

My point is that you stress tube length between pump and block, whereas it's only the total length in the entire system that matters.

/Olle

[neverthar]

Originally Posted by Olle P

My point is that you stress tube length between pump and block, whereas it's only the total length in the entire system that matters.

/Olle

I would have to disagree with you on that as we're talking head pressure loss which is proportionate to the distance of water travelled. As I've stressed, utilising higher head pressure pumps would not exhibit such differences but with weaker pumps, it's quite well exhibited. Once must remember the Delta charts done measuring flow restriction versus thermal dissipation. Certain blocks have steeper curves compared to others. those who read systemcooling's well done reviews would be able to understand why. Cascade style and jet impingement based blocks are very dependant on head pressure at inlet and with weaker (i.e cheaper) pumps this is further exacerbated in terms of effect. If I'm not mistaken, you are assuming an ideal state based on Bernoulli's equation for incompressible flow. However, that is with certain assumptions for calculation benefit. But with the introduction of restriction, the most possible way to maintain decent flow is to increase proportionately the pressure to overcome restriction follwing Newton's Law on energy conservation. What i'm trying to explain here is the dynamic pressure which will be lost over distance due to work required to overcome wall friction.

[Chai]

Originally Posted by neverthar

I would have to disagree with you on that as we're talking head pressure loss which is proportionate to the distance of water travelled. As I've stressed, utilising higher head pressure pumps would not exhibit such differences but with weaker pumps, it's quite well exhibited. Once must remember the Delta charts done measuring flow restriction versus thermal dissipation. Certain blocks have steeper curves compared to others. those who read systemcooling's well done reviews would be able to understand why. Cascade style and jet impingement based blocks are very dependant on head pressure at inlet and with weaker (i.e cheaper) pumps this is further exacerbated in terms of effect. If I'm not mistaken, you are assuming an ideal state based on Bernoulli's equation for incompressible flow. However, that is with certain assumptions for calculation benefit. But with the introduction of restriction, the most possible way to maintain decent flow is to increase proportionately the pressure to overcome restriction follwing Newton's Law on energy conservation. What i'm trying to explain here is the dynamic pressure which will be lost over distance due to work required to overcome wall friction.

Not too sure about that, but Cascade blocks will work fine on weaker pumps, still better than, say Swiftech's offering using the same pump.