• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Flow and distribution, an engineering problem...

PsycoGeek

PsycoGeek

[H]ard|Gawd
Joined
Jun 25, 2000
Messages
1,226
I had thought up a solution (I thought) to getting water to each block without having to pass it tohrough a single inline loop. I thought "MANIFOLD!" but the theory isn't proving to be as easy to put into practice.

The whole idea was to use a distribution manifold to separate flow of cool water to the CPU, North Bridge, and eventually the GPU. Then use a reclamation manifold to sent the heated water to the radiator for cooling (don't worry, I had planed on a custome built radiator capable of handeling the load). The system would look like this: (Start at the pump. There are two lines coming from the CPU because I have a 3 barb waterblock)


|-------------------------------------Radiator< --------|
| |
| ^
| |---------------->|
| |------->CPU| |
Pump----->Manifold--| |---------------->|
|------->NB-------------------->|


The problem I ran into in prototyping is keeping the flow rate even after the manifild. Can it be done without expensive and complicated regulating equipment? Can the flow be evenly distributed through the manifold design? Can it be done without restricting flow on one output line? The current non-working prototype is:

h l
^ ^
flow -----> ____ |__|_

In this design the first output has more flow, almost twice as much as the second one. I thought for sure while designing it that would be reversed. (h and l indicate High and Low flow.)

-edit- Crap, formatting of my fine asci art isn't working, and I don't know how to fix it... :mad:
 
Use the [c0de] tags.

The water won't be split evenly unless the restriction on each path is the same, and the only way you're going to get that is by changing the manifold restrictions over and over until it's right.

Because the temperatures vary very little in the loop, there's little to no benefit in not just inlining all the blocks. Plus you're going to have a nightmare tubing setup if you try to split and recombine the flow.
 
Code tags didn't work so well either. Will have to play with that some more.

I understand that with 2 blocks inline the temp doesn't vary that much, but after adding a GPU block (on a 6800 class card) it does seem to matter. I don't want to go into creating a completely separate loop for the GPU. But you are probably correct in that it will be too difficult to do as I add more devices to be cooled.

As far a a tubing nightmare, the whole idea of a manifold of this type it to organize the tubing and flow from a central point so that wye's and tee's don't have to be used, which are what cause routing nightmares. I know... I hate my current setup and want to change it.
 
Keep them all in series and just up your flow rate to yeild a small delta T across the system.
 
Flow rate is pretty high right now. So high in fact that I get bubbles in the system from the turbulance in the resivor.

I was thinking of going with a larger radiator as well if I have to stay the course with a single loop system. I may also plug off one of the out barbs in the waterblock to eliminate a wye to rejoin the flow before it hits the NB block. The bubbles from the resivor seem to collect in the topmost part of the block and when I pinch off a tube it pushes them through. The waterblock is an Asetek Waterchill Antarctica. I don't like the 3 barb design as well as I thought I would.

-edit- Was going to post a pic, but I can't get a clear picture of the resivor.
 
Even with a rise in temps between your northbridge and GPU, the temps should still be quite good so I wouldn't worry about it. If your temps are too high you probably need for radiator/more airflow.
 
If your pump is pushing that easily, inline it, and opt for the larger radiator you desire. Aqua computer makes the granddaddy of all rads, being able to fit 15 120mm fans on one side. I'm tempted to buy one for my next PC project....yeah that project...the one that keeps changing direction and never gets done :p
 
I was thinking of this one (radiator):

BIXIII_black_90cw-250w.jpg
 
mustang_steve said:
Aqua computer makes the granddaddy of all rads, being able to fit 15 120mm fans on one side. I'm tempted to buy one for my next PC project....yeah that project...the one that keeps changing direction and never gets done :p
Click to expand...

Why not just get a radiator from my neighbors Buic! Geez that thing is big! Definately not case mountable.....
 
Wouldnt the addition of 1 or 2 manifolds add a fair bit of restriction to the system?
 
I don't think so Mike. All a manifold does is distribute what is going into it. The only problem would be with the collection manifold, which increasing the tubing size vrom it to the radiator would at least partially solve.

The problem here is designing it to evenly distribute the water. That is what I can't figure out without the use of inline valves and digital flow meters. Ahhhh.... if only I was still working in the IT dept. at Ingersol-Rand... :rolleyes: I would have a bunch of engineers at my disposal...
 
PsycoGeek said:
Why not just get a radiator from my neighbors Buic! Geez that thing is big! Definately not case mountable.....
Click to expand...


It's case mountable, barely. You mount it on one of your side panels.

Actually if I get one, it's going on the side panel of my PC-72 that's been sitting around in the closet.
 
I played around with manifolds, flow meters and other stuff to do something very similar. The restriction (and resulting reduced flow) with an Eheim 1250 (not a weak pump) was amazing. Every extra split adds to the overall flow resistance. Parallel arrangement gave less flow throug the CPU block then the serial path. So I staid serial. Perhaps if you're using a 30 ft head pump you can get better cooling out of a parallel set-up, but I doubt you can do it with your standard aquarium pump.
 
whitewale said:
I played around with manifolds, flow meters and other stuff to do something very similar. The restriction (and resulting reduced flow) with an Eheim 1250 (not a weak pump) was amazing. Every extra split adds to the overall flow resistance. Parallel arrangement gave less flow throug the CPU block then the serial path. So I staid serial. Perhaps if you're using a 30 ft head pump you can get better cooling out of a parallel set-up, but I doubt you can do it with your standard aquarium pump.
Click to expand...


I also agree with this logic (engineer here, at your disposal).

edit: and thanks for hosting your own image, I was betting I'd have to remove that radiator pic :D
 
Now, whitewhale, that makes sense, and it's something I didn't think of.

Hey zer0signal, don't worry, I never steal someone elses bandwith. That's just cheap. Photobucket.com is great!

-edit- Where can I get a pump with a 30ft head? :D
 
PsycoGeek said:
Where can I get a pump with a 30ft head? :D
Click to expand...

Grainger.com - but think of the fact that 30 ft head means 15 psi on all your joints, your hoseclamps, your hoses, that o-ring around your barbs etc... I thought about it long and hard and figued I sleep better with an aquarium pump ;).
 
I'm currently building a manifold setup, using 3" acrylic cubes. One side has six tapped holes for hose barbs, the other side has one 1-1/2" tapped hole for the hose barb that's going to end at a ::cough:: pickup truck radiator. The whole idea is actually just a solution, I couldn't find a more elegant way to connect 1/2" ID tube to 1-1/2" ID tube.

I'm running a 3-barb cpu block, chipset block, gpu/vram block, and two hard disk blocks. That's only 5 lines on the input manifold, so I'm just going to plug one. The extra line on the output manifold is so I don't have to wye the cpu block output. Hopefully, this will decrease backpressure to the cpu block on the intake manifold, so it'll get a little more flow compared to the other blocks.

The great part about running a manifold system is *not* having equal flow to all blocks. You want more flow to the blocks that have to deal with more heat, so you want things like the cpu and video blocks to either get priority or be of lower resistance than the rest of the system. I haven't figured out how to get the water to like my video block more than my chipset block, so I may end up tubing the chipset block with 3/8" ID, as it's comparably cooler than the video block output when running. That would give me an excuse to run 3/8" to the hard drive blocks, and not have to match the 1/2" ID to 3/8" barbs.

The small advantage to running a dedicated-loop system is that every block gets cold water fresh from the pump (or radiator, depending on what order you set it all up). This has relatively little effect on final cooling performance, as water will be happy to absorb heat far beyond what a computer can produce without becoming saturated to the point of slowing it's heat-absorbtion rate.

If the small difference that running the coolest water possible to each block makes is enough to make or break your overclock, you're running too close to the edge anyway. If you're watercooling properly, you should have mountains of headroom in the area of heat dissipation when it comes to overclocking. The limiting factor becomes voltage and the hardwares' innate limits, and that discussion is for another forum area.

I run manifolds because when you make them out of acrylic cube like I did, they look really cool. Sorry I can't get pics, I'd love to show them here.

Most Important:

My first manifold idea was a simple, cheap solution (the whole manifold idea was the product of fixing another problem). I'd buy a short length of 1-1/2" pipe, plug one end, put the radiator hose on the other, and just jam hose barbs in holes I'd drill into the side of the tube. It was like a teenage hangout for stagnant water. Lots of internal volume, lots of space where water could just sit and create turbulence, no direct route from entrance to exit.

So don't make one of those.

You're still dealing with the same quantity of total waterblock resistance, the improvement of a dedicated-loop system is the multiple lines that allow the water more area to flow through the waterblock-portion of the circuit. Don't add resistance to the circuit by making a dumb manifold that will confuse the direction of flow and give you all kinds of air-pocket headaches. My manifolds are designed like a showerhead, they have little internal volume, and never change flow direction. It's the same advice as using slow curves in tubing instead of 90-degree bends, when you use bends, you're pushing water, as fast as possible, into a brick wall, and it has to build up that velocity all over again. That's a waste of pump work.

I think this is long enough. Thanks for sticking with me on our journey through hydrodynamics.
 
to add resistance to your chipset block's line to give other blocks priority, you could just use a bit of 1/4" ID tubing somewhere, or dual 1/8 ID, and see how that goes.
 
There's one serious flaw in the manifold argument and the "reduced resistance" over multiple paths gives you more flow: You cannot get more flow than your pump provides! Taking my Eheim 1250 experiments as example again, the unimpeded flow is 5 gpm. If I run serial CPU-GPU-radiator I get 2 gpm flow. If I run 2 parallel loops, cpu - GPU radiator, I get 1.5 gpm CPU and 2.5 GPU/radiator, if I go tripple I get 1/2/2. Yes, the total flow though my system is highest with a tripple split - but the flow through the individal components is lowest. And since the heatflow is proportional to the mass flow and the temperature difference, reducing the mass flow by 33% - 50% is much more detrimental to your cooling then having water that is 2 degrees warmer, reducing the temperture gradient by maybe 5%.
So unless you use a very high flow pump that can provide the maximal flow rate through all loops you put in you most likely get worse results with parallel.
 
Excellent point. You do have to remember to get a pump that matches your system, in the case of a dedicated loop, or manifold-type system, you can use a larger pump, provided you don't bottleneck it by running 1/2" tubing from the pump to the manifold. We're talking 3/4" and 1" or greater primary lines. You can upgrade to an Eheim 1260 or 1262, which both have 3/4" outputs, without forcing the water thats just coming from the impeller right into a bottleneck.

EDIT: Don't run the radiator on it's own loop. That's just... strange.
 
You must log in or register to reply here.
Back
Top