No offence taken here Chief... just giving you some advice that you'd do well to listen to on these forums. Better to learn and think on your own than to ask people to do it for you.

Are you talking about Graystar?

ooops
 

yea him, Greystar, sorry. Im going to so many different forums, its so crazy how much info one can gather on this topic. And keeping track of names.

Question
 

Got a question or lack of answer that is now bugging me since reading this thread, i'm sure you guys can clear it up for me.

Before i do what i do now the previous 8yrs i was employed as a control & instrumentation engineer for a UK power company. The power station i worked at was CCGT : combined cycle gas turbine, i.e Gas turbines produced electricity then exhaust gases from the GT's feed heat recovery boilers generating steam in turn feeding steam turbines & walla more electricity. The steam turbine condesnate then pumps back to a de-aerator tank, this tank then feeds the 6 boiler circuits/ drums & on & on....

Can sketch the config if it makes it easier to picture

The boilers have 6 circuits, the lowest in the stack (& hottest) are HP ( high pressure circuits), the highest the LP circuits. The very top (& coolest part of stack) circuit is the pre-heater circuit. The condesnate returning from the ST condensor passes through this pre-heater section for numerous reasons & leads me onto the question:

If the actuators (valves) controlling the pre-heater inlets control to wide open (to high flow) little heat is absorbed by the pre-heater section, throttle the flow right back & significant increases in heat are observed.
We where always told & up till reading this thread i took as true that the reason was the condensate was not in the pre-heater section long enough to absorb heat.

Can anyone clear this up for me? Can expand on descriptions & include sketches if it helps

TIA :)

quite correct
both sides of the heat exchanger are not operating at the same efficacy
and the lowest will dictate the performance overall

going to sounds dumb here but i'm still failing to comprehend the difference between more flow = good for a pc w/c system, & more flow = bad in condensate pre-heaters.

At the power plant all temperatures, pressures, flows etc where measured in triplicate & then voted by the control system so reliable data. With reduced flow we'd observe a) greater dt over pre-heater inlet /outlet & b) stack temperatures at the pre-heater level would reduce, i,e more stack heat absorbed into pre-heater water.

I'm no thermo wiz (v.basic grasp), but i see 2 instances of heat exchangers both with water in liquid state both behaving differently?

I think I get the idea.

In your case, you have a liquid coming into a heat exchanger, that you want to heat up. In this case, yes the liquid needs to remain in the heat exchanger for a longer period of time, so that it will absorb more thermal energy. In effect, you are looking for a maximum temperature differnce between the liquid before and the liquid after the heat exchanger.

However, in the case of water cooling, we are not concerned with the temperature differnce of the water. An ideal coolant would have an infinte specific heat capacity, and the temperature would not change at any point in the loop. We are concerned with the temperature of coolant required to dissipate the heat.

So in your case, you need to get the liquid to as high a temperature as possible, but this has to be in a single pass through the heat exchanger. If you speed up the flow, you still want the same thermal energy to be dissipated into the same volume of water. Obviously, as you have stated, this is not possible.

But for water cooling, if we speed up the flow, then we don't need to dissipate as much per unit volume of coolant.


So, in your case, if you want to double the flow rate, and have the liquid coming out at the same temperature, then you need to double the thermal energy dissipated by the heat exchanger, however, the heat dissipation in watercooling is constant, so the flow rate doesn't matter in the same way.

I hope this makes sense. It's quite hard to explain.

8-ball

A water block will transfer more heat if water is left in place, but you'll get a temp rise which is'nt wanted...

You were trying to heat the coolant/water no?. we're not. Leaving the water in place longer will cause a higher temp which will cause a higher DT(?) which will heat the water more but we are trying not to let the temp rise, but to cool it. Carrying the heat off at the lowest temp...

PS. This is'nt a statement, it's my guess. I don't understand fully heat transfer 'lingo/theory' much either. Unfortunatly I don't have the interest/drive to investigate it fully so I only dabble when I feel I need to :D ...

PPS. Halo BillA!!, hows work?. Getting your own way I hope!...

thanks
 

i think its sinking in, or something is :D:

varying flow in a pc-w/c single closed loop does not change time in the exchangers, quote cathars example

that said increased flow in same system does offer perfromance benefits in terms of decreased boundary layer, increased turbulence etc.

The power plant scenario though i 1st viewed as single closed loop, is far from it due to numerous cascaded loops & changes of states etc, so its the 'single shot' through the exchanger i think that explains it to me. That right? .... in laymans terms :D

Let me try and explain it this way.

There are two ways to use a heat exchanger.

1. Dissipate a constant power.
2. Cool a liquid down to a constant temperature.

For 1, if we increase the flow rate, then the energy dissipated per second does not need to change, so the energy dissipated per unit volume of liquid decreases.

For 2, then we need to dissipate a constant amount of energy per unit volume per second, so if we increase the flow, rate, then more energy must be dissipated by the heat exchanger.

Just different uses of a heat exchanger.

8-ball

More heat? Where does this magical more heat come from? Does the cpu stop and say, hmm the waters going slow today so I better put out more heat? No the cpu will give out the same heat no matter if the water is going slow or fast.
The water is like a conveyer belt. Picture a steady stream of sand landing on a conveyer belt. The flow of sand remains the same (the heat). Now increase the speed of the belt to an insane speed and what happens? The belt will look like it is almost empty compared to when it was going at a slow speed but the same amount of sand is being moved.

Would 'more heat units' be OK for you?...

No I don't think it would would it?. It'd have to be more heat units per unit of water would'nt it?...

PS: Was the word 'heat' really so ambivilent when taken in context with the rest of my reply?, or did you just read that one line and pounce? :p ...

Nice analogy. I'm going to have to remember that one.

Another Newb w/ questions...
 

(Good ones I hope)

This is an excellent thread, I've been learning quite a bit, thanks to all of you that have been 'experting' on it.

Let me run a few things by to see if I'm thinking correctly. I think I understand the theory, but I wouldn't try to repeat it ;-} I'm more interested in how it applies.

I'm in the process of designing a system which will be a heavy duty server/workstation configuration. I'm after high reliability and stability, so I won't be overclocking however I want to get as close to silent operation as I can. From what I've been reading, this implies needing high flow rates, and a big rad w/ multiple large slow running fans for high volume airflow. Correct?

To be as silent running as I can, I should use as few fans as possible, and make the ones I use as big as I can, correct?

I am planning to get as big a case as I can, and am thinking in terms of one or two 120's on the front sucking in (w/ filters,and trying to keep the case slightly pressurized for dust suppression), and a couple of 90's on the back blowing out through the rad, plus the PSU fan. I would use a digital doc or equivalent to keep speeds low unless things got hot. Does this sound like a reasonable plan?

I am planning to use a dual processor mobo, and high speed SCSI drives, but nothing fancy in the way of video. Judging from everything I've been reading, it looks like my primary heat sources will be the two CPU's, the hard drives (and possibly a DVD burner) and maybe the Northbridge (It's the only thing in the mobo picture with a heatsink) In order to keep the number of fans to a minimum, I expect I would need to put a heat sink on each of those sources. If I do, would cooling the rest of the system with the fan setup described above be adequate?

I haven't seen much description of multiple heat source plumbing patterns. I am making the following assumptions: Tell me if I'm missing anything?

1. The CPU's need the most cooling, and one should not feed into the other, so they need to be in parrallel.

2. To maximize the flow to the CPU's they should be plumbed w/ 1/2" tube, and my blocks (I'm going to make my own) should be designed to allow a high flow by staying fairly close to 1/2" worth of passage cross section while encouraging turbulence over the die area.

3. The drives and Northbridge don't make as much heat, so I can use smaller plumbing on them for lower flow.

4. Resevoirs seem to be neutral or a good thing in regards to cooling, and make maintainance and operation easier...

5. Optimal plumbing pattern is 'pump > rad > hot stuff > resevoir > pump'

Based on this, I am planning the following:

1. A pump (not sure what size, suggestions welcomed, but probably biggish) with an input barb as big as will fit it, and an output the same size as the rad input.

2. A rad, (I'll probably use a heater core for cost reasons) the largest one I can fit onto the outside back of the case with a shroud to space it an inch or two away from two fans mounted inside the case blowing through it (preferably 92's, but will use 80's if I must) *Note that the shroud will need at least an internal separator between the fans, and possibly flaps over one or both to prevent 'air shorts' especially if one is shut down when system is cool.*

3. On the output side of the rad, a manifold. I am thinking of making the manifold out of 1.5" PVC pipe drilled and tapped for barbs as needed for input and outputs for each circuit. My thought is this should minimize pressure drop from the plumbing and ensure that each CPU gets about the same flow.

Circuits as follows: one for each CPU (1/2") one or more 1/4" for drives, northbridge, etc. (how many?)

4. Heat sinks (I'll probably start a seperate thread for those later)

5. A mirror image of the manifold in 3 feeding into:

6. A resevoir, made from yet more PVC pipe, w/ extra barbs for a sight glass tube, and a cleanout plug at the top for filling.

Does this sound like a reasonable plan? Can anyone see anything that I'm missing or screwing up on excessively?

Thanks,

Gooserider

Re: Another Newb w/ questions...
 

This is quite a lot to comment on, so I thought I'd just take this one piece.

Not necessarily. This would be important if the water is heated up a few degrees by the first CPU, since the second CPU would be operating at a higher temperature. But generally the flow rate is such that heat input, even with multiple heat sources, is usually far less than a degree.

You might be better off putting them in parallel to reduce the pressure drop - two resistances in parallel create less pressure drop than two resistances in series - but the result of this is half the flow rate across each block, and typically worse cooling.

You'd have to try to be sure, but it's my opinion you'd get better results putting the CPU blocks in series.

1/2" cross section seems decent. Again, it's only my opinion, but I was always a fan of the blocks with big internal volumes.

If you're putting all these fans in the case I doubt you'd need to watercool the northbridge or HDD. Even a SCSI should be decent if it's in a position with plenty of airflow around it. If it's not . . . then yeah, WC the thing.

Let's not forget their aesthetic appeal. :)

Yeah, but again, the change in temperature of the water along the loop is very small, so what's most important is arranging things to prevent flow restriction. If putting the radiator right before the CPU block requires a 90 degree elbow, then you might be better off doing something else.

Good questions.

Alchemy

Thanks for the response Alchemy. I hope it shows that I've been trying to do my own thinking on this stuff, but it's kind of new to me.

I'm surprised on the CPU's, but that's why I'm asking... I'd only seen one or two other discussions that involved multi processor boards, and they did parrallel.

Well, what I'm planning on is running the fans at low speeds and/or with most of them in off/standby mode, so I wouldn't have as much airflow as might initially appear.

I had thought of trying to cool the northbridge just by jamming some copper tubing in between the fins on the existing passive heat sink. Would that work, or would it cause more problems than it would solve?

The hard drives I'll be running are going to be either the 10 or 15K rpm Seagate Cheetahs. Seagate's website specs them at 18 watts each which doesn't seem that much to me, but the stuff I've seen on drive cooling says it's needed on any of the over 7200 rpm drives. I'm planning to put them in the internal 3.5" bays of whatever case I get, and those usually don't get much ventilation to speak of.

True, but I'm more a function over form kind of person :) It probably isn't politically correct to say it around here considering some of the pix I've seen but IMHO "Beige is Beautiful" ;)

I agree, and I think that may be one of the advantages of those PVC pipe manifolds I mentioned. They will essentially be cylinders that I can tap into at whatever points along their sides that will give me the best shot at the target block.

Gooserider.

A little help here please: Alchemy, 8-Ball, Skulemate?

Aren't manifolds bad for pressure drop? (in this situation as described)

Since every abrupt change in diameter causes pressure drop and a manifold has multiple changes in diameter, where would one expect to benefit from using a manifold over using carefully selected components and just running them in series?

I know there are way too many variables to give a difinitive answer but some thoughts on the subject would be nice. Thanks. And hey it's still 'on topic'.

People need to sit down and do the math before they build :)

Say a high flow system moves coolant at 1 ft/s through a 1/2ID system. That means every second 9.5 cubic inches of coolant move through. That comes out to 155 grams of coolant in metric.

Assume water with a specific heat of 4.186 j/g. That means a heat capacity of 650j/C. So you'd need to input 650 joules per second (watts) to heat the coolant 1 degree. With a hundred watt CPU that means the second CPU gets heated all of .15C hotter then the first!

Thats why its safe to assume temperature is more or less constant inside the loop. And those numbers are conservative. 1 ft/s isn't all that fast, and few CPUs actually put out 100w continuously even when OCed.

Ok, Sounds like series is it...
 

OK Redleader, that makes sense, so It sounds like series it is, as I certainly am not going to worry about less than a quarter degree temp difference between CPU's.

I'm not sure Murray13, they are probably bad, but I don't know just how much harm what I have in mind would do.

My design would really have just two diameter changes per manifold. The one that splits the flow will have one big increase in diameter as it enters the manifold, and then a restriction on exit as it goes back down to the tube ID for that circuit. The one that recombines the flows would do the reverse.

My intent with the manifolds was twofold. One was to split the flow into different circuits, and the other was to use it to reduce the diameter of the radiator lines (probably 3/4" or 7/8" judging by most of the heater core specs I've seen) to the diameter of the circuit lines.

I still think I need two cooling loops, as I don't think the NB or drives need 1/2" plumbing or flow levels; nor do I want to try to get 1/2" into the drive bays. Also if I was to try to make drive cooling plates with 1/2" passages I'd end up with plates that were bigger than the drives :(

I have trouble thinking a manifold splitter would be more restrictive than adding a lengthy loop of 1/4" plumbing in series with the 1/2" CPU plumbing.

Of course, the arguement that says I should put the CPU's in series would also apply to my NB and drives, suggesting that I really only need to have two circuits, one 1/2" loop for the CPU's and a 1/4" loop for the other stuff. Manifolds seem a bit of an overkill approach for so few splits, so I might end up going with "T" or "Y" fittings instead. How would the restriction from those compare to what a manifold would produce?

Even more to the bottom line, If I go with a good high flow pump (say an Eheim 1048 or 1250) how much would I need to care? If I can get say 1-2 GPM flow through the system, does it REALLY matter?

Alternatively, given the list of things to cool I mentioned in my first post (2 CPU's, 1 NB, 3 drives) what would you do to plumb it?

Gooserider

Many people have more than 1 exaust from there CPU block. They then either Y it back together, or use each exaust line to cool a different heat source effectively turning there Wb into a manifold. I assume people use multiple exausts to lessen resistance or to get a better flow pattern.

I am planning on making a WW inspired block. The exausts will be used to cool the NB and GPU in parallel. The Intake will be 1/2in. Should the exaust be:

2 x 3/8in 0.220893233 in2
2 x 1/2in 0.392699082 in2
4 x 3/8in 0.441786467 in2

Does anyone have any comments on which one would be the best? Thanks for any help.

I think I'll stir this up a bit with this question... :dome:

If one pump and one loop (basically) in your systems works okay... what if you put TWO loops and join them at a reservior with two different pumps one for the radiator loop and one for the cooling loop (CPU block, GPU block, HD block, etc.)?

Would a differential in flow rates of the two loops to compensate for the different surface areas in each set of thermal transfer devices increase performance?

I'm no engineer, but manifolds wouldn't cause a loss of pressure. In a true closed loop system, pressure would be the same in the entire system, pressure isn't bound by size. A manifold would decrease flow rate. Smaller tube, faster flow rate, bigger tube, slower flow rate. Manifolds can be benificial if used correctly.

Not true.

...and to complete the answer...

The pressure is actually different at every point in the system, and ranges from a low, which may be slightly below atmospheric, to a high, as the pump provides.

*Flow* however, is the same everywhere, except where it is split.

A manifold is indeed *bad*, but only because it adds a little bit of flow resistance. To imagine it, picture a 12" (yes, 1 foot) pipe, with water flowing through it, then imagine 1'000 (1 thousand) infinitely small pipes, that run the exact same flow: which do you think will resist flow the most?

(now it's not really fair for me to put it that way, because I'm not saying anything about having the same cross section either way, but ya'll get it, right?)

Now consider Cathar's latest design, with 50+ mini tubes. The tubes are short enough so that the pressure drop is minimal, and that's the key. In fact, "Cascade" has a lower pressure drop that "White Water"...

Am I misunderstanding the term "manifold"? Isn't it just a larger container to which you distribute from? It was my understanding that pressure was constant on all sides in a "container" but velocity or flow changed with volume of the container. Look at the attached drawing. Am I wrong? If so, please explain further.

Longwinded ideal explanation of pressure inside a system:

Water will exit the pressure at a given pressure say 100rpu(random pressure units). Then it runs through 10 inches of tubing each inch it drops about 0.5rpu so by the time it gets to the rad its at 95rpu. The radiator is a compilation of a very long pipe with plenty of 180 degree bends(horible for flow). So be the time the water exits it at 80rpu. Then it goes through 12 inches of tubing on its way to the cpu. Down to 74. Then it hits a WW which has a very restrictive flow and turbulent area. The drop inside the cpu is 35rpu. So the sytem is now at 39rpu. The WW splits the flow into 2 seperate channles. From now on each branch will receive half the water but the pressure will be 39 at the begining of each loop. One side runs 16 inches to the gpu so its down to 31rpu before the block and 14 after. The other brack runs 8 inches to the NB so its down to 35. After the NB WB the pressure it 16 rpu. Both run back into a resevoir and exit with 10rpu to spare.

That is of course the IDEAL situation. The flow through each branch is related to the resistance in that branch and it would take some effort to eqaulize. Also any bends in the system will resist flow. The hose barbs commonly used also do a fair amount of resisting themselves. There are plenty of small things that complicate the flow inside a system.

I'll admit my model has 1 mistake. The water exits with exesse pressure and I don't think this will happen. I think the pump will compensate for not haveing to push as hard and will push more water.

Ok, think I understand now. Guess need to put in electrical terms.
Voltage(pressure)=Current(flow) X Resistance(flow restrictions). That makes sense. Guess I was taking the whole system in account. Thanks for the explaination.

Redleader: You are off in your calculations. You calculated based on a 1/2 radius tube, not half inch diameter. Assuming everything else is the same, you'd have the second processor being .62*C hotter.

BO(V)BZ

Damn. Thanks for the fix :)

Wow, now the series CPU doesn't look as good!
 

Well, that does make the series plumbing for the CPU's a bit less attractive, but I'm still not that worried about it, as it seems to me that other issues might still give bigger differences.

Either way, for reasons I mentioned earlier, I will still be splitting the output of my pump into multiple circuits, and perhaps the question should be rephrased....

Given that ANY kind of plumbing fixture will have a negative impact on flow rate, which of the following is WORSE?

The Manifold (note output number and placement is only a guess for illustration purpose only, I might put in several outlets)
............_______________________________
...........|...............................|
...........|........Manifold chamber.......|
-----------|....(approx 1.5" d X 3" L......|
IN (1/2"d or larger).......................|
____________...............................|
...........|...............................|
...........|...............................|
...........------|......|-------|...|------|
.................|..out.|.......|out|
.................|.1/2".|.......|1/4"|

Or the T fitting:

--------------------------------------------
IN (1/2"d)..........................OUT (1/2"d)

------------------|....|----------------------
..................|out.|
..................|1/4"|

Bonus question: What if there are several T fittings in a row?

Now I will be the first to admit my understanding of flow dynamics is not great. However my take on it is that the manifold is better, especially if there are multiple outlets.

To (mis)use the electrical analogy here, I see the manifold chamber acting in some ways like a battery or capacitor, it takes the flow from the pump and 'charges' to the pump pressure, then acts in it's own right as the pressure source that supplies the different outlets, where as the "T" fitting is like a spliced junction in an electrical wire that just pulls power off the main line, and adds some impedance losses.

Does what I say make any sense, or is my understanding full of bovine byproduct???

Gooserider