Water Flow

[CoW]8(0)]

My reasoning is using a simple heat capacity equation:

DeltaQ = CmDeltaT

DeltaQ = heat added to water
C = specific heat of water
m = mass of the water
DeltaT = change in temperature of the water

Assume that an observation is done over a waterblock over a given time range. The 'DeltaQ', 'C', will be our constants. The 'm' will be our independent variable (x) and 'DeltaT' will be our dependent variable . Now changing the equation to a function of 'DeltaT' instead of 'DeltaQ' will yield the new equation:

DeltaT=DeltaQ/(Cm)

This equation shows that increasing 'm' (mass) will decrease 'DeltaT'. 'm' as a variable is the mass of water flowing through the waterblock. Thus, as flow increases, increasing 'm', the temperature difference will decrease.

This being applied to a waterblock shows that the increased flow will cause a lower temperature increase in water as it leaves the outlet, but also applied to a radiator, it will show a lower temperature decrease leaving its outlet as well.

This is why I stated adequate flow through the system is necessary, but having much higher flow will only yield a more constant temperature throughout the system. I agree that the lower temperature differential improves cooling performance, but it should be negligible if not overridden by the larger amount of heat deposited into the water by the stronger pump.

Also, can someone explain to me the idea of less resistence of a parallel system than a series system? I know this is true for electrical circuits. I don't have much knowledge in fluid motion.

 

[Erasmus354]

My reasoning is using a simple heat capacity equation:

DeltaQ = CmDeltaT

DeltaQ = heat added to water
C = specific heat of water
m = mass of the water
DeltaT = change in temperature of the water

Assume that an observation is done over a waterblock over a given time range. The 'DeltaQ', 'C', will be our constants. The 'm' will be our independent variable (x) and 'DeltaT' will be our dependent variable . Now changing the equation to a function of 'DeltaT' instead of 'DeltaQ' will yield the new equation:

DeltaT=DeltaQ/(Cm)

This equation shows that increasing 'm' (mass) will decrease 'DeltaT'. 'm' as a variable is the mass of water flowing through the waterblock. Thus, as flow increases, increasing 'm', the temperature difference will decrease.

This being applied to a waterblock shows that the increased flow will cause a lower temperature increase in water as it leaves the outlet, but also applied to a radiator, it will show a lower temperature decrease leaving its outlet as well.

This is why I stated adequate flow through the system is necessary, but having much higher flow will only yield a more constant temperature throughout the system. I agree that the lower temperature differential improves cooling performance, but it should be negligible if not overridden by the larger amount of heat deposited into the water by the stronger pump.

Also, can someone explain to me the idea of less resistence of a parallel system than a series system? I know this is true for electrical circuits. I don't have much knowledge in fluid motion.

You are thinking about it all wrong. The point of higher flow is NOT, I repeat NOT, to decrease the temperature difference between waterblocks. The point of higher flow is to increase the temperature transfer between the waterblock and the water. In ANY watercooling system the water temperature will come to an equilibrium temperature. This equilibrium temperature is wholly dependent upon the radiator alone, and how it is setup. The comments about increased flow increasing performance are not related at all to the calculations of the increase in water temperature between waterblock inlet and outlet.

The basic idea between less resistance in parallel is simple. If you have a 1/2" tube and you then split it into two 1/2" tubes, you now have less resistance as the water effectively has twice the area to flow through. In practice this doesn't exactly work out because the Y-connector you are likely using to split the tubes has its own resistance which balances out the effect of the tubes. However when you add waterblocks into the mix then there will likely be benefit of having a parallel loop for less restriction. If you have water in a bucket and you make one hole, there is lots of resistance to the water leaving the bucket, if you make multiple holes there is much less resistance

 

[CoW]8(0)]

The calculation shows that temperature difference in the water of a system does matter. There is a direct relationship of the mass of water going through the waterblock with the quantity of heat which is required to raise water 1C (or any relative amount). Therefore you're distributing the heat more evenly among all the water in the loop with higher flow. This will yield lower temperatures for the waterblock relative to the water and higher temperatures for the radiator relative to the surrounding air.

However given the equation: DeltaT=DeltaQ/(Cm) which graphed looks something similar to
http://www.rc.umd.edu/praxis/mitchell/images/asymptote1.gif with 'm' as the x axis and 'DeltaT' as the y axis you can see that 'DeltaT' begins to diminish as 'm' increases. This explains why increasing flow after a 'point of adequacy' will yield negligible amounts of improvement in cooling performance.

But once the tubes merge back into one, wouldn't the this cause all the resistance to return since you've decreased the area to flow through?

 

[CoW]8(0)]

Sorry mistake, meant to edit. Any way to get rid of this?

 

[zer0signal667]

The calculation shows that temperature difference in the water of a system does matter. There is a direct relationship of the mass of water going through the waterblock with the quantity of heat which is required to raise water 1C (or any relative amount). Therefore you're distributing the heat more evenly among all the water in the loop with higher flow. This will yield lower temperatures for the waterblock relative to the water and higher temperatures for the radiator relative to the surrounding air.

You're mixing principles of equilibrium and non-equilibrium heat transfer. There is a constant mass of water in the system regardless of flowrate. If you were piping water through the block and out onto the ground, THEN you would need to consider mass flow. In the case of a recirculating loop, as you said several times, any given water molecule spends the SAME amount of time in the block regardless of how fast, slow, how many times per minute it goes through it. Distributing heat throughout the water is the wrong thing to be thinking about, heat transfer characteristics of the water/block interface are the only properties that will affect the temperature differential once the system is at equilibrium.

This is bordering on being off-topic, but I think I'll let it slide as long as the OP has no objections AND the discussion remains civil (I'm not finger-pointing, but these discussions usually end up in flames)

 

[Mysterae]

This is bordering on being off-topic, but I think I'll let it slide as long as the OP has no objections AND the discussion remains civil (I'm not finger-pointing, but these discussions usually end up in flames)

I don't mind all this discussion, it's a good thing and really informs myself and other readers. Deciding on which is right is the difficult bit! But no one likes a flame-war so lets keep it civil!

Wow, a lot to get into! I think the main argument is does increasing the flow of the water increase the efficiency of the waterblock cooling, (transfering of heat from the cpu to the water alone, negatting leakage from other parts of the block). Concensus says it does, to a degree, then tails off following the Law of Diminishing Returns. The water in the loop will reach an equilibrium temp, and the radiator plays a big part in lowering that temp.

The more waterblocks (pipe length, joints, bends too) in a loop (be it series or parallel) will increase the backpressure (restrictiveness) of the system, requiring a good pump to overcome that backpressure and allow flow. The following quote is taken from this article from systemcooling.com:

For example: if we measure the pressure on the inlet and outlet sides of a waterblock (with water flowing thru it at a constant rate), we might see 2.5 PSI on the inlet side and 1.5 PSI on the outlet. This means the system water pressure is dropping 1.0 PSI as it flows thru the waterblock – creating 1.0 PSI of backpressure.

So what I really need to do is measure the backpressure my manifold is going to create because it's going to be much more than if I just connected the pump straight to the first block in a series config. Then measure the backpressure of each waterblock to see if they are equal (doubt it) and calculate the total bp of the manifold and waterblocks in both series and then parellel, assuming the electrical equation is applicable, or I've just made myself look like a right spanner:

series - Rtotal = R1+R2+R3+R4 etc
parallel - 1/Rtotal = (1/R1)+(1/R2)+(1/R3)+(1/R4) etc

Right, after all that, I'm thinking of changing the plumping lol! Don't freak and think all this has been done for nothing! I may get rid of one manifold (the hot one in my orignal diagram) and it's connected pump. For three reasons:

- the backpressure the second manifold is likely to cause
- a colleague in work mentioned cavitation which I had read about some pumps doing. This is likely to happen running two pumps because the chances of them running at the same flow rate is low. This could be solved by shunting them together (a pipe between them with a valve to limit the level of shunt) but this adds more pipework and you guessed it, more backpressure - although it would be before and after the pumps (depending on which one is where).
- I'd save time and money only having one pump and manifold

Instead of the hot manifold and pump, the seven outlets from the six waterblocks would all go to the reservoir before going to the radiator. I'll draw another piccy to show what I mean. But I feel like doing the original design just for the hell of it!

To show how well the system is (hopefully) going to perform, I'm in the lucky position to be able to borrow a thermal imaging camera from work That and the figures should be the evidence to prove it either way!

 

[CoW]8(0)]

Well the topic is still dealing with flow rate and the science behind it.

The water temperature in the entire system will never be the same everywhere, unless you're given infinite flow. There is indeed a constant mass of water in the system, but the system isn't closed regarding heat transfer, which is why you could technically think of the water as being pumped 'out into the ground'. The water entering the block in a watercooling system isn't necessarily 'new' water, but it is definetly water with different heat content than when it left. This is why the equation still applies and why the 'm' value is governed by the amount of flow through the block.

Thermodynamic equilibrium does not occur between the waterblock and the flowing water. The water will also not maintain a constant temperature, only a part of the watercooling system will (i.e. the CPU). Therefore the temperature of the water inlet and the water outlet will never be equal, the heat content of the water entering the block will be different than the water leaving. But having higher flow lowers this temperature difference.

 

[Mysterae]

Using the resistance calculations I mentioned earlier, and assuming the following backpressure values:

Manifold - Rm = 1
GPU1 - R1 = 1.5
GPU2 - R2 = 1.5
NB - R3 = 1
CPU - R4 = 2
HD1 - R5 = 0.5
HD2 - R6 = 0.5

The figures above are completely made up BTW.

The Rtotal (or total backpressure) would be:

Series - 8 PSI
Parallel - 7.16 PSI

Thats 10.5% reduction in backpressure. Up the backpressure of the manifold up to somethiing silly like 10PSI (or add another manifold and pipework ) and the reduction lowers to 5% in parallel compared to series. Anyone convinced yet?

 

[Erasmus354]

Well the topic is still dealing with flow rate and the science behind it.

The water temperature in the entire system will never be the same everywhere, unless you're given infinite flow. There is indeed a constant mass of water in the system, but the system isn't closed regarding heat transfer, which is why you could technically think of the water as being pumped 'out into the ground'. The water entering the block in a watercooling system isn't necessarily 'new' water, but it is definetly water with different heat content than when it left. This is why the equation still applies and why the 'm' value is governed by the amount of flow through the block.

Thermodynamic equilibrium does not occur between the waterblock and the flowing water. The water will also not maintain a constant temperature, only a part of the watercooling system will (i.e. the CPU). Therefore the temperature of the water inlet and the water outlet will never be equal, the heat content of the water entering the block will be different than the water leaving. But having higher flow lowers this temperature difference.

I am not quite sure what point you are trying to get at...to me it seems like you are taking what has already been established and trying to find a way to make it more complicated. Try to think simple.

 

[Erasmus354]

Using the resistance calculations I mentioned earlier, and assuming the following backpressure values:

Manifold - Rm = 1
GPU1 - R1 = 1.5
GPU2 - R2 = 1.5
NB - R3 = 1
CPU - R4 = 2
HD1 - R5 = 0.5
HD2 - R6 = 0.5

The figures above are completely made up BTW.

The Rtotal (or total backpressure) would be:

Series - 8 PSI
Parallel - 7.16 PSI

Thats 10.5% reduction in backpressure. Up the backpressure of the manifold up to somethiing silly like 10PSI (or add another manifold and pipework ) and the reduction lowers to 5% in parallel compared to series. Anyone convinced yet?

Using your calculations if you were to look at just the HDD waterblocks the backpressure in series would be one, and the backpressure in parallel would be 4. I think you need a different measurement to avoid fractional numbers and the problems that arise when using them.

 

[Mysterae]

Opps, rather than edit my last post I'm willing to show the error in my calculations! I had calculated the series config with the manifold, which it would of course not have. Sorry!

So then:

Series would be 7PSI
Parallel would be 7.16PSI

An increase by 2.4%. Damn it!

However, in parallel each block would have the same temp of water flowing into it (rather than cascading down the chain) and the amount of flow into each blocks inlet would be equal (but not coming out).

 

[CoW]8(0)]

I am not quite sure what point you are trying to get at...to me it seems like you are taking what has already been established and trying to find a way to make it more complicated. Try to think simple.

Simply put,
higher flow => better distribution of heat of the water => more consistent temperature of the water in the system => better cooling (most likely a negligible gain compared with a system of adequate flow)

 

[Mysterae]

Told you my maths sucked. And so do my spreadsheet skills it seems!

Using:

or online makes 0.5 twice in parallel 0.25, not 4.

It throws my previous calculations in the trash too, giving an overall figure in parallel of 1.15PSI including the manifold. Wish that were true compare to a series backpressure of 7PSI for a series config.

That's an immense difference in backpressure and thus waterflow. Can't be right or everyone would be doing it parallel. Perhaps real world will be the only way to see.

(Again, sorry for all my crap maths).

 

[Bio-Hazard]

All this math is making me glad I dropped out of school in the 3rd grade.

 

[CoW]8(0)]

Told you my maths sucked. And so do my spreadsheet skills it seems!

Using:

or online makes 0.5 twice in parallel 0.25, not 4.

It throws my previous calculations in the trash too, giving an overall figure in parallel of 1.15PSI including the manifold. Wish that were true compare to a series backpressure of 7PSI for a series config.

That's an immense difference in backpressure and thus waterflow. Can't be right or everyone would be doing it parallel. Perhaps real world will be the only way to see.

(Again, sorry for all my crap maths).

Well as long as the relative values of the resistance of one waterblock (ie CPU) with another (ie GPU) are accurate, the ratio of total resistance of a series versus parallel will be accurate. Thus, simply set one block as equal to 1R (R being an arbitrary constant) and estimate resistance of the other blocks relative to 1R (or the 'control block').

Also the resistance of the manifolds must be added on directly since they are actually in a series in the overall loop.

Perhaps if you post the blocks you plan to buy, others can give you accurate estimates its relative resistance to another block.

 

[zer0signal667]

Simply put,
higher flow => better distribution of heat of the water => more consistent temperature of the water in the system => better cooling (most likely a negligible gain compared with a system of adequate flow)

Ok, well we agree that higher flow=better, however you're still not taking the right path to get that answer. Are you now saying that the more consistent water temperature is beneficial? Earlier it seemed like you were saying greater temp differentials were better. Regardless, heat flow across the system is always the same, equal to the output of the processor (and pump, to get anal about it). Increased flowrate increases heat transfer at each interface, decreasing thermal resistance, decreasing temp differentials, while heat flux remains constant. There are equations to back this up, run a search for Reynolds, Prandtl, Nusseldt and you'll find a plethora of information on convective heat transfer. I've yet to see an equation that mentions any kind of property or parameter regarding heat distribution, and frankly I can't see how you even make the link between that and better cooling.

 

[zer0signal667]

Told you my maths sucked. And so do my spreadsheet skills it seems!

Using:

or online makes 0.5 twice in parallel 0.25, not 4.

It throws my previous calculations in the trash too, giving an overall figure in parallel of 1.15PSI including the manifold. Wish that were true compare to a series backpressure of 7PSI for a series config.

That's an immense difference in backpressure and thus waterflow. Can't be right or everyone would be doing it parallel. Perhaps real world will be the only way to see.

(Again, sorry for all my crap maths).

Don't forget that while your system will have less resistance in parallel, it will also have increased effective cross-sectional area... overall flowrate may increase, but fluid velocity will be slowed within each block. Going back to the electrical circuit analogy, there is less "push" (voltage, pressure drop) across each "component" (resistor, waterblock). And then don't forget the theme of the whole discussion, fast flow=more efficient heat transfer.

 

[Mysterae]

Bio-Hazard, me too. I spent too much time in art class and not enough in maths...

But it's never to late to learn . CoW]8(0), that's how I calculated it (finally). You want to check my maths Sir?

I've replaced each blocks bp with 1 as you suggested and added the second manifold as per my original idea. 2.17PSI backpressure compare with 6PSI for series. Big difference. Okay so add in a bit for the extra pipework and some valves and I still doubt it will be close to 6.

Right, out of maths class and into art class. Here's how I would do it with only 1 manifold and 1 pump:

It's debatable wether the pump should go between the reservoir and radiator or as I have shown it here. Between the res and rad would mean any heat from the pump to the water would be dissapated by the rad, but the rad would add it's own backpressure to the equation (and I aint feckin doing that again).

As for blocks I'm still looking and reading. I liked the look of the cooler master Aqua suite of blocks, but they are 1/4 fittings and reviews are thin on the ground - just check the thread I started lol!

 

[Erasmus354]

you want to have the pump directly after the resevoir, that is the entire point of the resevoir. The only function of a resevoir is to make priming the pump and filling/bleeding the loop easier. Preferrably you want to have the pump directly after the resevoir in the loop, and also below the resevoir physically.

 

[CoW]8(0)]

Ok, well we agree that higher flow=better, however you're still not taking the right path to get that answer. Are you now saying that the more consistent water temperature is beneficial? Earlier it seemed like you were saying greater temp differentials were better. Regardless, heat flow across the system is always the same, equal to the output of the processor (and pump, to get anal about it). Increased flowrate increases heat transfer at each interface, decreasing thermal resistance, decreasing temp differentials, while heat flux remains constant. There are equations to back this up, run a search for Reynolds, Prandtl, Nusseldt and you'll find a plethora of information on convective heat transfer. I've yet to see an equation that mentions any kind of property or parameter regarding heat distribution, and frankly I can't see how you even make the link between that and better cooling.

More consistent water temperature isn't necessarily beneficial, but more consistent water temperature shows that the heat is being distributed evenly in the water. With increased flow , you're giving more water molecules a share of heat energy. Basically, the equation is showing that one would increase flow in order to have the lowest increase in temperature in heat areas (waterblock heat energy => water). You're trying to dump as much heat into the water without raising its temperature. To minimize this, you would indeed distribute the heat as evenly as possible in the water via increased flow. This would cause more consistent water temperature.

 

[CoW]8(0)]

Bio-Hazard, me too. I spent too much time in art class and not enough in maths...

But it's never to late to learn . CoW]8(0), that's how I calculated it (finally). You want to check my maths Sir?

I've replaced each blocks bp with 1 as you suggested and added the second manifold as per my original idea. 2.17PSI backpressure compare with 6PSI for series. Big difference. Okay so add in a bit for the extra pipework and some valves and I still doubt it will be close to 6.

Right, out of maths class and into art class. Here's how I would do it with only 1 manifold and 1 pump:

It's debatable wether the pump should go between the reservoir and radiator or as I have shown it here. Between the res and rad would mean any heat from the pump to the water would be dissapated by the rad, but the rad would add it's own backpressure to the equation (and I aint feckin doing that again).

As for blocks I'm still looking and reading. I liked the look of the cooler master Aqua suite of blocks, but they are 1/4 fittings and reviews are thin on the ground - just check the thread I started lol!

I'm assuming your math is already correct But I'm not sure if your ratios are. Is a CPU block 4 times more resistant than a HD block. Is a GPU block 3 times more? You can't have an accurate comparison if the ratio's between each block are not accurate.

Here's a vague estimation: If there is a large difference in resistances between each block, you'll see much less total resistance in a parallel system than a series. If there is a small difference in resistances, you will see less of a difference of total resistance between the two.

 

[zer0signal667]

It's debatable wether the pump should go between the reservoir and radiator or as I have shown it here. Between the res and rad would mean any heat from the pump to the water would be dissapated by the rad, but the rad would add it's own backpressure to the equation (and I aint feckin doing that again).

As for blocks I'm still looking and reading. I liked the look of the cooler master Aqua suite of blocks, but they are 1/4 fittings and reviews are thin on the ground - just check the thread I started lol!

The radiator is going to add backpressure no matter where the pump is. Erasmus was suggesting that the pump be next the the res for a few reasons, I think, namely cavitation prevention and ease of filling the loop. Physical placement can really make a huuuge difference in the amount of aggravation you get from trying to fill and bleed the loop.

and CoW]8(0), I get the impression that you're putting together different ideas each time you try to convince me of the same thing, and I honestly no longer know what you're trying to convince me of so I politely concede. Until you have some concrete facts or principles in terms of thermodynamics, fluid flow, or heat transfer (besides heat capacity, which is only the tip of the iceberg) then I'm not getting your point, sorry

 

[CoW]8(0)]

In no way am I trying to drag out a debate, I'm simply trying to show an idea. I could very well be incorrect, but I have yet to see someone give me a direct reason (or even a scenario) as to why my idea could not apply. Users have been only telling me of other reasons why an increase in flow will yield better performance. That doesn't necessarily invalidate my reasoning, or vis versa.

Mysterae, do you have any cooling parts chosen yet? That would be a helpful start, at the moment we can only guess (with very little accuracy) of how your cooling will perform in comparison to a series.

 

[Thor]

flip your res and your manifold and alot of the problems will be fixed, if you make the res tall thin and deep (say 12" t by 2"w by 8" d) you can put ball valves at the very top on the res one for each channel, only keep the res half full or so , then you can open/shut the valves, by how much water you SEE comeing out of them... its by no means scientific, however, it should be very effective, and it will eleminate alot of the guess work as to your math, and trying to figure out how much water is acctually comeing out, no flow meter needed the trouble with this is, while the res is part way empty , it will introduce air into the line (the taller the res is the better chance it will give the air time to escape, befor the pump sucks it in , a baffold in the res would help as well) , so a seconed res would be needed to compleatly fill the loop once the flow rates are figured out, or you could leave it part way full so you can moniter it and it will make for a nice water feature..

option #2 would be to put more barbs on your rad, one for each line, and keep the manifold where it is, with the res just befor the pump... you could possibily make the manifold huge (like my last drawing) and use it for the res as well

hmm.. do both..

i might have to run with this myself.. when i have the money.. lol naw.. this pc is close to 80lbs as it is.. much heavyer and ill need a fork lift.. lol

thore

 

[SpoogeMonkey]

flip your res and your manifold and alot of the problems will be fixed

To add to that, change the flow of the pump to go to the rad as well: pump=>rad=>manifold=>water bloacks=>res=>pump

 

[Mysterae]

Thore, you got a wicked idea there! Something like this?

but with a few more inlets! Nice idea, although I'd have to stop playing BF2 every few minutes to go for a wee! It wouldn't make the pc silent anymore but the sound of running water is meant to be soothing, yeah?! Interesting all the same....

CoW]8(0), the blocks I'm looking at, as I've said before but to a shocked silence, are the Cooler Master blocks. They're pretty new so not many reviews, but I've never been one to go with the flow (sic).

 

[Thor]

from my experiments with my custom case, i can tell you, you will need to ditch the cilinder idea unless you plan on makeing it better than 3" in diameter and 8 " tall with a baffold, 12 with out or it will suck in air (this is for a single inlet , you will have more inlets at a lower flow, so i am assumeing that lower flow=less force (force being = to how much air the stream drags in with it) now, this would mean that air penatration should be less with more streams but there would be more air at once , so the hight would still have to be about that same (my ehime will create a vortex and suck in air when placed in 4 " of water (above the top of the pump inlet... so about 7 " of water total)

also, if you have the inlets along the side like that, tygon has a max bend radious of about 4 " add that to the outside of the tube +2 in for barbs and the tightest radius you can make is about 6 " from the center of the tube 12" from end to end if htere in a hexagon around the tube you will run into problems if your trying to put it in a standard PC case (hell.. i dont think i could get that in my YY cube even if it where empty) ,

if you use a cilinder and put them all at the top, force will be increased (slightly, due to gravity)and the tube will have to be bigger... i figure about 5" in diameter so the you can acctually get a wrench in there to tighten the barbs, and the clearence at the top would be greater, (8" pluse 2"for the barb and 4" moreto get the tubeing to bend to where you need it. getting the barb thread cut would be toughter (with a clearence of only 1.5 in inbetween each barb, the chances of cracking the top of the tube(if plexi ) are pretty good mabe not on the seconed hole, but sooner or leter... it takes a slight mirical to cut threads like that, (or a machine and still some luck), but, if you plan on makeing it out of cast acrillic, you could put the barbs in place befor hand and hope you dont have any leaks, getting a baffold in place in a cast piece may be more difficult,, or it could be easyer... i dont know for shure..

how ever, make it rectangle,while it dosent solve any thing for clearence at the top, it makes life alot easyer all around

barbs are easyer to place , spaceing and less chance of the top cracking, as well as installing the flow control, and finding the right baffold config is easyer... it can be mounted to the side of the case (as the window??? )

/edit: sorry for the hughness of the pic ill resize it in the morning... gotta run (reedit:fixed)

thore

 

[Mysterae]

Thore, you do come up with some wicked ideas, and words of wisdom. I didn't think about the acrylic cylinder cracking when installing the barbs or hose connections. The manifold(s?) and reservoir are probably going to made from 10mm or thicker acylic, hexagonal instead of circular.

And I've seen the vortex effect you speak of in an article found here, the image below is taken from that article.

Although it looks pretty cool, all the air won't help much.

So I need to avoid such an effect like you say. The design you posted would work a treat, and I'm going to keep that one filed!

As for a case, I haven't decided on that yet as it's not going to be your standard pc case. My last case was a ClearPC and I liked it, but the basic shape of a pc case hasn't changed for donkies years and it bores me now, as it does many of the forum members, hence the reason so many folk mod 'em. I'm making the design at the moment, but need to finalise the watercooling as that will shape the look of the rig.

The waterblocks are ordered (Cooler Master CPU, NB, GPU x2, HD x2, out on a limb there), the plumbing stuff ordered tomorrow, first draft of the manifold in a week or two hopefully. Got a lot of testing to be done to see how it all flows, what kind (and shape) of reservoir is best. Just looking at rads now; probably a 2x120mm fan sized rad.

I might move this off this thread and start a project build, would that be better?

 

[Thor]

Thore, you do come up with some wicked ideas, and words of wisdom. I didn't think about the acrylic cylinder cracking when installing the barbs or hose connections. The manifold(s?) and reservoir are probably going to made from 10mm or thicker acylic, hexagonal instead of circular.

thanks

And I've seen the vortex effect you speak of in an article found here, the image below is taken from that article.

Although it looks pretty cool, all the air won't help much.

now make the pump an ehime with 7 " of water over head in a sink with the same vortex... my littel ehime can drian a sink faster than the faucet can fill it... as the space gets smaller the effect gains faster, that little vorex probably took a good 10 to 15 seconeds to form, ehimes in my situation can do the same in the larger volume in about the same time

So I need to avoid such an effect like you say. The design you posted would work a treat, and I'm going to keep that one filed!

the beauty of that one is it minimise's splashing into the res water, and should keep bubles out of the system for the most part, it also will help with vortexing (yes im not shure thats a real word) think of it like a rectangular wheel. the water only gets up so much momintum berfor it gets spun out into the wider part of the res, and then it doest have the energy when it comes back to the other side (that may not make alot of sence, im tired.. lol , i hope you get what im getting at)

As for a case, I haven't decided on that yet as it's not going to be your standard pc case. My last case was a ClearPC and I liked it, but the basic shape of a pc case hasn't changed for donkies years and it bores me now, as it does many of the forum members, hence the reason so many folk mod 'em. I'm making the design at the moment, but need to finalise the watercooling as that will shape the look of the rig.

hmm.. shoot some ideas out here and lets see what you have in mind, i have a few that would look awesome for this project, but this is your project not mine.. but.. some basic spec's should get you started, to house that much water cooling, you'll eather need a full tower (duble height if you will.. you know.. the monsters that are about 4ft tall) or.. you could go cube, be prepared to have this thing be very heavy and basicly unmoveable... (you should see me take mine to a lan party... its fine once i get it set up... but the trek from and to the car are something else.. it takes 2 people to carry this thing... and a third if a door is involved.. )

The waterblocks are ordered (Cooler Master CPU, NB, GPU x2, HD x2, out on a limb there), the plumbing stuff ordered tomorrow, first draft of the manifold in a week or two hopefully. Got a lot of testing to be done to see how it all flows, what kind (and shape) of reservoir is best. Just looking at rads now; probably a 2x120mm fan sized rad.

a 2x120 will do you well

I might move this off this thread and start a project build, would that be better?

as long as you post a link so we can follow it

use this as a speculation thread, figure out what you want how your going to set it up, then post to the build thread once you make the choise, so that thread stays clean

 

[Mysterae]

Speculation this thread is!

Right, mind dump of the few ideas I have, and if you know they have been done already, please supply links so I can see how they executed it, see the problems before hand and get more ideas!

The name of this build is going to be called H3 (Hexagonal, H20 and [H]ard) or HEX (not after the cool series on SkyOne with the fit Christina Cole), but for reasons you can probably guess after reading below.

Idea 1 - to put the computer in a fish tank. Seriously, but not with any water, that's idea number 2! I've been looking for hexagonal fish tanks, made of acrylic (no way glass), big enough to fit everything in but small enough to fit on my table. I don't want a monstrosity that has to hide under the table. All this work to hide it away? No Sir! The manifold(s) (acrylic hexagons) suspended above the motherboard, as would the reservior, with the water feeding to and from the waterblocks. That would look quite awesome!

Idea 2 - to have the computer inside a hexagonal fish tank, which is inside another larger hexagonal fish tank that contains water. Well, the water would be the coolant, acting as the reservior! Damn that first fish tank had better be water tight from the outside-in! This design would give the illusion that the whole case contains water. This would be easy to fill and empty, but weigh a considerable amount, not to mention the water surrounding lots of electrical circuitry - I doubt you'll see this design in in PCWorld...

Idea 3 - a bit of both ideas 1 and 2. This is for portability. Using the two hexagonal fish tanks, the inner tank would unconnected to the outer, so being easily removable just by lifting it out. This way I could pump air in a tube at the bottom of the outer tank with holes in it to create bubbles....have I gone too far now lol?!

Fish tanks are expensive and difficult to get in the sizes I'm looking for, so my plastic's guy would probably make them to the exact size I'm looking, in say 5mm or thicker. All electrical feeds (power, video lead, usb) would come in from the top. In all ideas the rad would be placed at the top.

This is the first custom pc I have made for myself in 3 years, and this H3 or HEX is going to last me another 3 years, so it's got to be future proof (yeah right). It's going to be an AMD x2 4400 or greater, but I'm waiting to see what ATI does in the way of Crossfire and motherboards before I submit. If ATI isn't better than on-par (I don't believe the rumours), it'll be a Nvidia SLI rig. So at the moment all the testing with water cooling is to be done on my ageing XP2200, Soyo Dragon Platinum, Radeon 9700Pro.

Which ever design I decide upon, it's ambitious to say the least and is certainly going to take a couple of months to complete!

 

[Mysterae]

*bump*

Just so you know I'm not sleeping on the job here, I have decided what the pc is finally going to look like amongst other things:

Highlights:

- plenty of space for another pump should I need it.
- easy to fill at the top
- easy to empty with tap at the back (not in picture)
- crossflow fan to cool the mb bits
- hot air and rad air expelled at the top
- waterflow visable in almost all blocks
- kinda compact (350mm wide 500mm tall)
- all in perspex (some construction may be ali)
- may tint the perspex from the inside with one-way mirrored vinyl (car window stuff)
- should look pretty cool!

The manifolds should arrive early next week so I'll start a log around that time. Just thought I'd give you guys a little taster since you all helped so much!

 

[Erasmus354]

looks neat, I dunno how well scaled that all it is but it looks like you need more space between the Disc Drives and the PSU (if those are to scale). Also you might want to consider tilting the PSU on its side. Most PSU have a bottom 120mm fan or something to draw in air. Make that fan flush with the case, and have the PSU basically cool itself and expel the hotair out of the case.

 

[CoW]8(0)]

Or just flip the PSU upside down if it isn't already...

Why such a big reservoir? If it's going to be that size, you might want to consider making it out of something that conducts heater better, (ie anondized aluminum)...

And it doesn't look proportionally correct...

 

[Mysterae]

Fair points, the psu is close to the dvd drives just to keep the inside dimensions as low as possible. I'm looking for a power supply at the moment that is near as silent, so only 1 fan and would probably remake the cables to suit - in exit, size and numbers etc.

Swinging the model round you can see the closeness (don't you just love Sketchup!):

The res is so large because - it can be! Also it makes filling and emptying a lot easier, plus adds that WTF bling factor I guess. It's not doing any harm being that size.

 

[Erasmus354]

If you want some premade computer parts to fiddle around with in your model there is a thread over on bit-tech.net that links to a site full of sketchup computer models. Lemme go see if I can find the link directly for you.

EDIT : http://forums.bit-tech.net/showthread.php?t=46003

that is the bit-tech thread, the site with the components appears to be down atm.

 

[Mysterae]

Cheers for the link Erasmus, didn't know there was such a following of Sketch3d in the pc mod scene! I've already imported the sata drive to my model!

I don't think (or don't know how) to get things as accurate to .1 mm so I might be shifting to another 3d package to get final dimensions for my cutting list to be a bit more exact.

 

[SledgeMakeGood]

I've got to stop leaving this forum for so long at a time. I did this, in a much uglier way. I used hot tub manifolds, PVC bits with a 1-1/2" or 2" thread on one end and parallel with it four or five 3/4" smooth ports on the other (I use a CPU block with one-inlet/two-outlet and hoped giving it twice the outlet hoses would decrease backpressure and improve flow tot the hottest block, hence four hoses from the pump, and five returning to the radiator).

I didn't need to adapt the large ends because my radiator came out of a pickup truck and the pump I bought to be able to supply strong flow through four (actually five, one line was split to two 3/8" hard drive blocks) waterblocks, which had a 1" outlet to begin with. The small ends easily took 3/4" male smooth ends to 1/2" threaded adapters, which in turn took 1/2" hose barbs. Brass hose barbs. These manifolds could stun a burglar.

The manifolds were designed to provide the least amount of backpressure, being inline in flow all the way through and having egg-shaped bulbous protrusions directly facing the incoming flow to help redirect it to each outlet without reducing it's velocity too much. And they worked, being part of a rapidly evolving technology used by gigilos and spring-break kiddies alike (the hot tub).

The problem was, I would have been (and, in fact, now am) much better off with everything the same (including the pump), but in series. Same temperatures, same flow rate measured at the radiator inlet (turns out the extra lines don't do much to help flow when measuring on the scale of what the blocks take out of it). The only difference, besides ease of setup and maintainence, was that I almost halved my number of possible leak-points. I also managed to run far fewer lines from the inside to the outside of my case (can't quite fit that radiator in an ATX).

It looks less flashy. I do miss that, but atm I felt like having a clean look (getting a Lian-Li case after I saw just how shiny they are in person).

Now, ymmv. Don't take my experience as proof of anything. And most certainly don't think you won't see a difference if you try it in series at some point for comparison. I just didn't notice any difference worth reporting. I skimmed through the thread and think you're running 1/4" hoses from the manifold, that alone will probably change your results in comparison relative to mine. Heck, I'm running a pump that could probably burst a hose if given enough backpressure. Realizing what I just typed, I think I'm gonna go perform some preventative duct tape additions.

Here's to a darned cool idea, good luck and take lots of pics!

 

[Mysterae]

Thanks for you input SledgeMakeGood, knowing somebody has tried this sort of thing before gives me less feelings of trepidation!

Hot tub manifolds are a great idea, but those fitting sizes are monstorous! Would have given you great flow though. Got any pics or threads of your set up before you went back to series?

I am wondering about the pump, the D5 I have selected may have to be paired up as shown in earlier pics. Got a lot of testing, trial and error ahead! The one downside in my proposed set up, as you mentioned is the number of connections. So got to check and double check them!

 

[CoolNQuiet]

All this talk about balancing the flow in a parallel system has got me wondering simpler parallel systems. For example, in a system where you use a Y connector to split the flow and send one path to the CPU and the other path to the GPU and chipset, do people worry about balancing to flow in the 2 paths. If so, what do you use to balance the flow?

 

[Mysterae]

What I mean by balancing flows is to restrict one (by a valve of some sort) to increase the available flow to the other. Not sure if people do this in the set up you describe because they want max flow in both the cpu and gpu as they're both pretty hot runners. I'm doing it to see how much the temps of the gpus' affect cpu temp. If it's fine to have max flow through the gpus' and cpu, the valves on the gpus' will be fully open or removed.