Water Flow

Hello there!

I've been lurking around the forum for some time gleeming as much info as I can about water cooling and overclocking. I built my last machine (XP2200+, Soyo Dragon blah blah..) about 3 years ago and it's time for an update!

Since I want this machine to be as near damn quiet as possible (unlike the hoover I have at the mo) and squeeze as much power as possible, water cooling is the only way to go!

The updated pc is going to be primarily for games, but a bit of other stuff too.

I haven't got or even spec'ed the hardware yet (waiting to see what ATI bring out in the next month or so, but may go the NV route, dunno) but I want to iron a few things out regarding the water cooling side of things.

Now, to get to the crux of this post, - water flows.I'm not convinced that linking all the water blocks in series is the way to go, but I could be wrong. I'm planning on water cooling almost everything : CPU, GPU x2, NB, HDs x2. Thats six water blocks. Heck, I might even water cool the PSU if I feel it will help! When you go for it, you gotta be [H]ard right?

Now, surely six waterblocks in series would be bad, because (IMH) you would need some serious flow not to transfer the heat from the first few blocks the later ones in the chain. So I was thinking something along the lines of the image below:



I might be barking up the preverbial, missing something or whatever, and I'd appreciate your views and comments on this!

Cheers!

Great illustration. There have been numerous discussions on this topic and it from my experiance, the heat gain of the water is pretty negligable in a series loop. The .1 degree increase in water temperature after it leaves the CPU does not seem to adversely effect other components down the line. I currently cool 2 separate CPU's with the same series loop. When attempting parrallel installation, the biggest challange for me was maintaining adequate flow thru both blocks.

Thanks Disarray, valid points. I'm surprised that it's only a .1 degree increase in water temp after the first block. If the water was heated by the first block, to say 35 degrees, the next block will have to start at 35 degrees, and be increased, and so on down the chain. The speed of the flow will have an impact to. Flow too fast, the water won't take the heat, flow too slow and the water will get hot (thus transfering it to the next if in series).

The main prob with the way plan on doing it is that each block will only have one sixth of the flow of the pump, am I right?

Mysterae said:

The main prob with the way plan on doing it is that each block will only have one sixth of the flow of the pump, am I right?

Click to expand...

Not necessarily. The water will take the path of least resistance, so that means that if you have a restrictive block in your loop, that particular component won't be getting as much water flow as the other components will. That means that it also won't be cooled as well.

If you are dead set on doing it that way (all in parallel), its going to take a lot of planning to get each component adequate waterflow so that it doesn't overheat.

Mysterae said:

Thanks Disarray, valid points. I'm surprised that it's only a .1 degree increase in water temp after the first block. If the water was heated by the first block, to say 35 degrees, the next block will have to start at 35 degrees, and be increased, and so on down the chain. The speed of the flow will have an impact to. Flow too fast, the water won't take the heat, flow too slow and the water will get hot (thus transfering it to the next if in series).

Click to expand...

Also... (I forgot to address this)

That is not how a water loop removes heat. Its not a matter of the "next block" starting at a temperature or anything. All of the water in the loop is approximately the same temperature. The water just has to get to a certain temperature (depending on your radiator) to dissipate the amount of heat energy being discharged into it. So, you aren't adding temperature at all, you are adding the total amount of heat energy. The efficiency of the heat dissipation mechanism will determine what temperature the water will be at to dump that amount of heat energy.

Also, I think you might be a bit confused with waterflow. Higher flow will ALWAYS remove more heat from a heat source than lower flow - all other things being equal. The water will absorb the heat energy no matter what, and at a higher flow, it can remove it from the source quicker, and dissipate it through the radiator quicker. However, many pumps that yield higher flow dump their own heat energy into the water which can decrease the efficiency of the loop. There are a lot of factors that can go into determining whether a high flow system or a lower flow system is the correct application for you.

There's plenty of threads you can search for that go into a good bit more detail of heat transfer in a water loop, but I figured I'd start with the general stuff. Good luck!

The bad thing about watercooling in parallel is that the most restrictive blocks are generally the CPU, the CPU is also the block that needs the most flow. However if you run it in parallel it will be the one that recieves the least flow. The heat gain from a normal processor is about 0.5-0.7C, less for gpu, less for other types of things. A typical loop will vary in temperature by about 2C from start to finish. So there really is no reason not to do series.

All you guys are making me think, and that was the point of the post! I didn't think that a series whole loop would be almost the same temp from begining to end, by about 2 degress. I thought the air cooled water exiting the rad would be significantly lower than that exiting a waterblock, and I reasoned that by giving the blocks cooler water it would absorb more heat energy, keeping the chips cooler.

On the subjec of the blocks restrictiveness. Yes, the water will follow the path of less resistance, slower in more restricted blocks, faster in less restrictive blocks. So what if I were to place flow valves at the output of each block so each could be 'tuned' to in effect balance the system? For example open the CPU valves up fully (or not fit them since you would want these fully open at all times) and adjust the other blocks valves until they do not affect the CPU's flow. This would mean some kind of measurement of flow or just monitoring the temps of all blocks. It's kind of like a domestic central heating system, instead this would be a central cooling system lo!

Mysterae said:

All you guys are making me think, and that was the point of the post! I didn't think that a series whole loop would be almost the same temp from begining to end, by about 2 degress. I thought the air cooled water exiting the rad would be significantly lower than that exiting a waterblock, and I reasoned that by giving the blocks cooler water it would absorb more heat energy, keeping the chips cooler.

On the subjec of the blocks restrictiveness. Yes, the water will follow the path of less resistance, slower in more restricted blocks, faster in less restrictive blocks. So what if I were to place flow valves at the output of each block so each could be 'tuned' to in effect balance the system? For example open the CPU valves up fully (or not fit them since you would want these fully open at all times) and adjust the other blocks valves until they do not affect the CPU's flow. This would mean some kind of measurement of flow or just monitoring the temps of all blocks. It's kind of like a domestic central heating system, instead this would be a central cooling system lo!

Click to expand...

That is exactly what you have to do if you run watercooling in parallel, you need to balance the flow which can be quite a daunting task. The disadvantage of running blocks in series is mainly that a loop run in series has more total restriction than a loop run in parallel. However most watercooling pumps these days have plenty of head pressure, which helps to overcome the pressure drop of a loop in series and still provide great flow.

It takes a lot of energy to heat up water, 4180 joules of heat to make one litre of water increase in temperature by 1C. A processor outputs around 100 joules per second, or 6000 per minute. A moderate flow loop will have at least 3 litres per minute of flow, meaning that you have 3 litres (12000 joules to raise temp by 1C) passing through the waterblock which is generating only 6000 joules of heat. Thus the temperature only raises by .5C. It is the incredibly high specific heat of water that makes it a good medium for heat transference.

Some things to think about are to run your main components in series, and then some of the minor ones in parallel. For example run the cpu/gpu/nb in series, and then the next component in the loop will be the two harddrive coolers in parallel. Since the HD coolers are likely identical you dont need to worry about balancing the flow, and you get the benefit of reduced restriction in the loop as a whole.

1st let me say, awesome diagram, that really brought to life what your trying to accomplish.

2nd, i think, that, in series will probably still be more effective, but not by a whole lot.. and lets face it, your way looks damn cool. So why not try it out, and see what happens? Test your theory. Whats the worst thing that can happen? you end up changing them to series.

I say, test your theory, post pics and results, and see what happens! You wont know until you try!

Erasmus, your maths really explain what is actually happening in terms of numbers and energy, thanks for taking the time. I can see now that a series loop will after a period of time come to some constant temp, with the radiator keeping the overall temp down. I understand it alot more now even after reading and searching through previous posts.

Making the system in a semi-series, semi-parallel system is a good idea, I've seen that a few times on this forum.

But, as bradyapda suggested, (cheers mate) I'm gonna give my way a go! This cooling system is really for my new pc which is a month or so off, but I reckon I'll fit it to my exiting pc. The only think missing is the extra graphics card but no biggie.

All I need to do is to make a measured drawing of the manifold and see how much it will cost. If it's too dear I'll make one up. What do you reckon to making it out of? Acrylic or ali?



I'd imagine acrylic rather than introduce another metal to the mix

There is more to designing an efficient manifold than just having multi- branches. If you search around perhaps pipeline design sites you may find the math which is an arm long. But as I see it you need your distribution manifold to be a sort of reservoir or accumulator tank, which is of a greater volume than the demand of the six outlets example you have a 1/2" pipe fed at one end with six outlets along it 1/2" each depending on pressure the fiest one will get the flow or the flow will be divided by six. There is way more to it than this i Know all said and researched you will find above advice of series loop is best or do complete seperate loop and rads

Duly noted jman1. I had considered making the centre bore 1/2 (this will be the inlet on the cold manifold, and outlet on the hot manifold) and the six branches 1/4, but I don't think I will.

Like the vicar said to the prostitute, I'm going to have to suck it and see

I look forward to seeing the end result, good luck, and as long as you are careful with what you are doing it should all perform just fine. It seems like you are putting a lot of thought behind this and that should bode well for the end result

Sounds like a fun project, I'm looking forward to seeing some custom design/machine work. Got a theme for your case? I suggest something involving octopi or centipedes

sorry for gettoing up the nice skills you have.. (what program is that any way???)

any way,scrap the first manifold and split the seconed into 2 zones and use the center port for the res. you can put flow control on the ports comeing off the block's the pump will do the rest, makeing the mainfold taller will give the air in the water time to rise out of the loop and into the res, (same as a filler tube) the best way i can think of to make said manifold is out of cast acrillic (err.. resin) and simply devide it down the middel with a piece of plexy to create 2 zones, the only place where the 2 xones will interact is in the very top at the filler tube, but it should be restrictive enought that it wont hurt any thing if the plexy touches the barb at the very top (1/8th in plexy)

i was considering what would happen if you didnt seperate the minifold, and i have a feeling that the pumps would fight each other in this loop, so best to keep it sepperate

lost my train of thought.. ill post again if/when i remeber

thore

Mysterae said:

Hello there!

I've been lurking around the forum for some time gleeming as much info as I can about water cooling and overclocking. I built my last machine (XP2200+, Soyo Dragon blah blah..) about 3 years ago and it's time for an update!

Since I want this machine to be as near damn quiet as possible (unlike the hoover I have at the mo) and squeeze as much power as possible, water cooling is the only way to go!

The updated pc is going to be primarily for games, but a bit of other stuff too.

I haven't got or even spec'ed the hardware yet (waiting to see what ATI bring out in the next month or so, but may go the NV route, dunno) but I want to iron a few things out regarding the water cooling side of things.

Now, to get to the crux of this post, - water flows.I'm not convinced that linking all the water blocks in series is the way to go, but I could be wrong. I'm planning on water cooling almost everything : CPU, GPU x2, NB, HDs x2. Thats six water blocks. Heck, I might even water cool the PSU if I feel it will help! When you go for it, you gotta be [H]ard right?

Now, surely six waterblocks in series would be bad, because (IMH) you would need some serious flow not to transfer the heat from the first few blocks the later ones in the chain. So I was thinking something along the lines of the image below:

I might be barking up the preverbial, missing something or whatever, and I'd appreciate your views and comments on this!

Cheers!

Click to expand...

Interestingly enough that’s exactly how they do multi zone heating systems in homes. I too have been thinking along those same lines except for multiple machines.

Love the concept and diagram.

If you really want to run everything in parallel, make one or two reservoirs with multiple inputs like this:



You could make a reservoir in any design you want, this is a great guide about them from Overclockers.com.

Since you're already thinking about multiple pumps, just run a second loop with a single 120mm or even smaller radiator, as chipsets and hard drives don't make too much heat anyway (you could probably just get some good passive coolers that would cool your chipset fine with some airflow, but if you want to water cool them, go ahead). This second loop could easily run well with a quiet Eheim 1048 opr even 1046, but you would probably want a higher performance pump for your cpu and gpus loop, such as a Laing D5 (aka Swiftech MCP655).

If you do make the manifold parallel loop, give pics

I think what we are all trying to say is... go ahead and do it!! Its a great design. Will it be optimal cooling? Maybe, maybe not. Will it definately give you good cooling? Probably... but you wont know your results til you do it!

I think we all would love ot follow along and see you do this, and see your results, and what you do.

Go for it mate!

Thanks for all the interest guys!

zer0signal667 - the case that I'm thinking of is a bit of a secret at the moment. Well, in truth I've not settled on one yet but have a fair idea. It's going to be clear for sure, to show off the gubbins inside!

thore - ghetto-up my drawings anytime mate! The prog I used for that was CorelDraw 11, but Visio (only have this at work) would have been easier. I see what you are saying in the amended drawing, the only thing that puts me off is the second rad. I was hoping to get away with a 2 fan sized rad, nice idea though.

BillR - that's what I thought too, I wonder how you would apply it to multiple machines.

ikellensbro - thanks for that excellent link to Overclockers.com, I didn't find that article on my web search for manifolds. Of the manifolds I did see for sale were long narrow ali blocks that I thought wouldn't give an even distribution of water. I'm checking out pumps at the moment, either of the two manufacturers you mentioned, Eheim or Laing. The two pumps I use will of course be identical, but the power concerns me as I would like to avoid having to fit another psu (it's already going to have to power two gfx cards) and don't want to go down the 240V route.

bradyapda - you're the devil on my side lol!

I spoke to my plastics guy today and he said send him the drawing of what I'm trying to do and he'll cut up the plastic and supply the glue (or I'll use silicone). The first part is to make a test manifold to see if it works before making it for real, if you know what I mean. just need to decide on the pumps and the rad. The blocks I'm going for are the Cooler Master Aqua's, but I made another thread for that discussion!

Something like this might work well for your manifolds...........



http://www.usplastic.com/catalog/product.asp?catalog%5Fname=usplastic&product%5Fid=18541

LOL... nice find!

Just for the record... I decided not to cool my GPUs in parallel in favor of feeding the top card with the coolest water (being that it's generally the warmer card) running the line out from that one right to the bottom card and they both run at exactly the same temp!! 39c idle and 44c load (or there abouts).

Bio-Hazard, it was seeing a manifold like the one you show (infact, here it is, just found it)



that kicked it off in the first place! Although I was concerned about each branch getting equal amounts of flow. For example, where do you fit the inlet...in the middle? Then, I could be talking pants here, the furthest away outlet(s) may not get enough flow. You would have to be carefull which block you fit where (in terms of restrictiveness). Those manifold blocks do look sweet though.

revenant, been watching your project log keenly! Awesome attention to detail like so many on this forum. When mine gets off the ground I'll prog log it too, but the bar has been raised so many times.

Erasmus354 said:

That is exactly what you have to do if you run watercooling in parallel, you need to balance the flow which can be quite a daunting task. The disadvantage of running blocks in series is mainly that a loop run in series has more total restriction than a loop run in parallel. However most watercooling pumps these days have plenty of head pressure, which helps to overcome the pressure drop of a loop in series and still provide great flow.

It takes a lot of energy to heat up water, 4180 joules of heat to make one litre of water increase in temperature by 1C. A processor outputs around 100 joules per second, or 6000 per minute. A moderate flow loop will have at least 3 litres per minute of flow, meaning that you have 3 litres (12000 joules to raise temp by 1C) passing through the waterblock which is generating only 6000 joules of heat. Thus the temperature only raises by .5C. It is the incredibly high specific heat of water that makes it a good medium for heat transference.

Some things to think about are to run your main components in series, and then some of the minor ones in parallel. For example run the cpu/gpu/nb in series, and then the next component in the loop will be the two harddrive coolers in parallel. Since the HD coolers are likely identical you dont need to worry about balancing the flow, and you get the benefit of reduced restriction in the loop as a whole.

Click to expand...

I'm not sure if your calculation makes much sense. There may be some problems such as:
1. The specific heat you mentioned is for pure water at around 4C.
2. Assuming the heat of the processor is completely transferred to the water.
3. The temperature of water would only be raised by that small of a value in an open water loop. Unless your radiator is able to dissipate 100% of the heat energy the water gained, your water temperature will be rising by much more than .5C. Therefore your conclusion can only be applicable to one loop of the water.

Remember, in a WC system, the specific heat applies also to lowering the temperature.

A radiator dissipates at about -55.1 joules/min. With 3 liters/min of water passing through a radiator, the change in temperature is about .0044 C (for pure water @ 4C).

Surprising? It should be because both equations are incorrect. Neither can be applied because there are MANY more variables in the environment. But mine should at least account for heat conductivity since the radiator's value is from manufacturer specs.

CoW]8(0) said:

I'm not sure if your calculation makes much sense. There may be some problems such as:
1. The specific heat you mentioned is for pure water at around 4C.
2. Assuming the heat of the processor is completely transferred to the water.
3. The temperature of water would only be raised by that small of a value in an open water loop. Unless your radiator is able to dissipate 100% of the heat energy the water gained, your water temperature will be rising by much more than .5C. Therefore your conclusion can only be applicable to one loop of the water.

Remember, in a WC system, the specific heat applies also to lowering the temperature.

A radiator dissipates at about -55.1 joules/min. With 3 liters/min of water passing through a radiator, the change in temperature is about .0044 C (for pure water @ 4C).

Surprising? It should be because both equations are incorrect. Neither can be applied because there are MANY more variables in the environment. But mine should at least account for heat conductivity since the radiator's value is from manufacturer specs.

Click to expand...

1. The specific heat I mentioned is for water at around room temperature. Between like 12 and 30C water has a specific heat within 50 joules (ie it doesn't really change).

2. Once the loop has been running for awhile almost all of the energy the processor puts out is absorbed by the water. A very little bit will be dissipated into the air from the sides of the waterblock, but that is negligible. Otherwise if all the heat wasn't absorbed then the waterblock would heat up until it melted

3. The water will always heat up by about that much when going from waterblock inlet to waterblock outlet. However as the loop runs for a bit the water will slowly raise its overall temperature. It will raise in temperature until it reaches its "equilibrium" temperature. This is when the delta T between the water and the air is enough for the radiator to dissipate all the heat being put out by the components in the loop. Once at this point the water temperature will not fluctuate very much at all. However throughout the entire time the water will still heat up by roughly 0.5C each time it travels through the waterblock.


My calculations have nothing to do with what the eventual equilibrium temperature will be. That cannot readily be determined as too many factors come into play. What radiator are you using? How big is the surface area? What airflow are you getting through the radiator? And what is the ambient temperature? My calculations are much more applicable than your calculations for the radiator. The reason being that the specific heat for water is much more constant throughout the temperature ranges we are dealing with than is the radiators ability to dissipate heat. Also, the heat from the processor has to go into the water, and it has to go into the water at roughly the same rate as it is being pumped out by the processor.

That being said, yes there are issues with laminar flow and whatnot causing certain parts of the water to heat up more than others, but overall the calculations are very applicable because the variables you are thinking about are negligible in their effect.

The specific heat of water may be more constant throughout the temperature changes. But a WC doesn't contain pure water.

And if you question the changing value of the radiators ability to dissipate heat at a given temperature range, you should also question the ability of the waterblock to dissipate heat.

Boy, all this math is way to much for me............... For me it doesn't matter what it looks like on paper, what does matter is the maximum overclock I can get out of a system. and the math on paper and the way a system performs when it's installed isn't always the same...................

Yeah, maths isn't my greatest subject either.

Brain Fart Time - IMO manufacturers of water blocks should test and specify certain things in the datasheets for their products. For instance, the flow output of a block for a given input of GPM/LPM, thus measuring the resistance or restrictiveness of the block. Or what about the amount of heat (energy) it transfers to the water, for a given temp, flow and time. Instead we are reduced to "oh, that looks pretty" or "that looks rock" or waiting for a review on the 'net to tell us how it performs.

Mysterae said:

Yeah, maths isn't my greatest subject either.

Brain Fart Time - IMO manufacturers of water blocks should test and specify certain things in the datasheets for their products. For instance, the flow output of a block for a given input of GPM/LPM, thus measuring the resistance or restrictiveness of the block. Or what about the amount of heat (energy) it transfers to the water, for a given temp, flow and time. Instead we are reduced to "oh, that looks pretty" or "that looks rock" or waiting for a review on the 'net to tell us how it performs.

Click to expand...

Well, all waterblocks transfer the same amount of heat to the water....exactly the same amount of heat that the processor outputs. The issue here is efficiency. Some waterblocks may be able to transfer more heat at a lower temperature than other waterblocks, which is what results in the differences in performance. Inefficient designs will cause the waterblock to heat up until there is enough of a deltaT for the same amount of heat to be put into the water as the processor is creating.


It all comes back to conservation of energy and what not. You cant just simply lose the heat energy. And while there are other paths for it out of the system (dissipating into the air from the sides of the waterblock, or through the tubing) the amount of heat lost in those ways is very small. Therefore essentially all the heat created is transfered through the loop and out the radiator.

Erasmus354 said:

Inefficient designs will cause the waterblock to heat up until there is enough of a deltaT for the same amount of heat to be put into the water as the processor is creating.

Click to expand...

From what I gather, what you are saying is that the quicker the block absorbs the heat from the cpu, and the faster that heat is taken away by the flow of water, the better. The way you explained just confused me a little and I had to read it a few times, especially the bit I quoted. I think you mean you don't want the water to be at the same temp as the cpu, as the main point of all this watercooling malarky is to bring the temp of the cpu down, or control it at least.

Yea it is kind of confusing. Basically all I am trying to get at is that heat leaves the water at the same rate that it goes in. This applies to the loop as a whole, as well as to the individual waterblocks. The heat going in to the waterblock from the processor leaves the waterblock (into the water) at the same rate that it goes in. The same way that the radiator (eventually) dissipates the same amount of heat from the loop that is introduced to it by the various components.

Well - while I think there are cooling advantages to distributed parallel cooling like this, you'll have to make sure you radiator(s) are doing their job well... because it's still a sealed "system" and will only cool as well as it's ability to shed heat... but I think getting the coolant from the rad to each block at the same time is definitely optimal, and would help balance temps, like on SLi systems and such... oddly enough I was able to achieve the same temp results with my gpus cooled in series... anyways, with high flow systems the water stays in the blocks for such a short time, and with only 2 or 3 blocks to cool, not sure if you would see the positive impact that much.. but with 4 or 5+ devices getting cooled, then I think it would really help balance the temps. anyways.. just a few fragmented thoughts..

Dark Ember said:

Also... (I forgot to address this)

That is not how a water loop removes heat. Its not a matter of the "next block" starting at a temperature or anything. All of the water in the loop is approximately the same temperature. The water just has to get to a certain temperature (depending on your radiator) to dissipate the amount of heat energy being discharged into it. So, you aren't adding temperature at all, you are adding the total amount of heat energy. The efficiency of the heat dissipation mechanism will determine what temperature the water will be at to dump that amount of heat energy.

Also, I think you might be a bit confused with waterflow. Higher flow will ALWAYS remove more heat from a heat source than lower flow - all other things being equal. The water will absorb the heat energy no matter what, and at a higher flow, it can remove it from the source quicker, and dissipate it through the radiator quicker. However, many pumps that yield higher flow dump their own heat energy into the water which can decrease the efficiency of the loop. There are a lot of factors that can go into determining whether a high flow system or a lower flow system is the correct application for you.

There's plenty of threads you can search for that go into a good bit more detail of heat transfer in a water loop, but I figured I'd start with the general stuff. Good luck!

Click to expand...

Higher flow will increase the amount of water through a waterblock which will increase the amount of heat required to raise the water temperature. But because it is a closed system, having a higher flow through the radiator will increase the amount of heat needed to dissipate to lower the water temperature. This is a reason why you won't see very strong pumps on WC systems, only adequate flow through the waterblock is needed. Thus, higher flow will simply provide lower temperature differentials in water.

You can also think of it this way:
In a given period of time, a parcel of water will still spend the same amount of time in a waterblock and a radiator regardless amount of flow in a system.

CoW]8(0) said:

Higher flow will increase the amount of water through a waterblock which will increase the amount of heat required to raise the water temperature. But because it is a closed system, having a higher flow through the radiator will increase the amount of heat needed to dissipate to lower the water temperature. This is a reason why you won't see very strong pumps on WC systems, only adequate flow through the waterblock is needed. Thus, higher flow will simply provide lower temperature differentials in water.

You can also think of it this way:
In a given period of time, a parcel of water will still spend the same amount of time in a waterblock and a radiator regardless amount of flow in a system.

Click to expand...

Your first paragraph is a little confusing and contradictory. Your second paragraph is right on. The part you're missing is that higher flowrates always yield more efficient heat transfer. I think you're trying to say that in a radiator, higher flow results in lower temp differentials which reduces heat transfer. The case is that higher flowrate increases heat transfer, which in turn allows for a lower temp differential- this is a GOOD thing because it lets your water reach cooler equilibrium temps. The same exact principle applies to heat transfer through a waterblock.

what zerosignal said

What I'm trying to say is that because more water passes through a waterblock, the heat capacity is much higher. This in turn requires more heat to raise the temperature of the water. But this in turn requires more heat to be dissipated by the radiator.

Current thinking on flow rates is:

Higher flow rates don't increase cooling, As stated above the water stays in the block(s) and rad the same amount of time no matter what the flow.

Though there are some blocks that require high pressure (ie Cather's G5 and now G7) due to their restrictive "impingement" technology.
http://www.systemcooling.com/swiftech_storm-04.html
Systemcooling.com lists flow resistance for many blocks they have tested.
Also check out Cather's posts on procooling.com

Water flow through a rad must be above a minimum to avoid laminar flow, above that minimun additional flow has dimminishing returns.
http://www.overclockers.com/articles1163/

And just a comment: While your "manifold" system is interesting, without tremendous expense for valves , flow meters and temp sensors you will be just stabbing in the dark with all the variables you are introducing and you will need some kind of "super pump" to get decent flows with all the restrictions.

But, like others have said, it looks interesting so why not try.
Air flow through a rad has a great effect on cooling capacity.

http://www.procooling.com/articles/html/pump_comparison__-_phaestus__1.php

This analogy isn't totally accurate of a water cooling setup as there is no radiator to dissipate heat, but the same principals apply

Take a 2 gal pot of water and put it on your stove. Turn the stove on to 150F. Now you can stir nice n slow, or flog it until your eyes roll back in your head, but eventually the water will become 150F, since there is no active cooling on it.

Now think of the pot as your waterblock (if you've ever noticed the best pots are all copper) and the stove is your cpu. If you can imagine the total volume of water in your loop as 1 piece, and not just "the bit thats currently in the block" this is the same thing. Without anywhere for the water to go and get cooled, it will become 150F, just like the water in the pot.

CoW]8(0) said:

What I'm trying to say is that because more water passes through a waterblock, the heat capacity is much higher. This in turn requires more heat to raise the temperature of the water. But this in turn requires more heat to be dissipated by the radiator.

Click to expand...

*sigh* What you are trying to do is compare apples to oranges.

Ignore for a second all of the talk about flow so far.

Now take this fact : Every single waterblock in existance so far (and probably ever) more efficiently moves heat away from the processor the faster the flow through the block is. This is easily seen by looking at the C/W vs Flow charts at ProCooling. As the flow increases, the C/W decreases, or in laymans terms--more flow = better cooling.

Ok now that we have that established lets think about what the radiator is doing in a watercooling loop. As was stated by the overclocking.com article referenced earlier, so long as you keep above a certain flow rate, flow rate has very little impact on the cooling performance of a radiator. What does have an impact on how much heat a radiator can dissipate however are three main factors.
1) How much surface area the radiator has, and to a lesser degree how thick it is.
2) How much air flow is moving through the radiator.
3) Finally, the temperature difference between the air and the water. The higher the temperature difference the more heat can be dissipated.

This last factor is the most important when trying to understand how cooling works in a watercooling loop. When you first turn on your computer, the water will be at the same temperature as the ambient room temperature. This means that the radiator dissipates zero heat. As time progresses the water heats up, and as the water heats up the radiator dissipates more and more heat. Keep in mind when I say heat I am really referring to Watts which are measured in Joules per second. The water will continue to heat up until the difference in temperature is enough so that the radiator is dissipating exactly the same amount of heat that is being put into the loop. All of this will happen in any watercooling loop regardless of what type of flow rates you are getting.

Now that we understand how the actual cooling occurs in the watercooling loop we can look at how the waterblocks function. The point of this thread was parallel cooling versus series cooling. One of the arguments for parallel cooling over series cooling is so that the last components in your loop are not getting hot water. I will reference my earlier post for this :

It takes a lot of energy to heat up water, 4180 joules of heat to make one litre of water increase in temperature by 1C. A processor outputs around 100 joules per second, or 6000 per minute. A moderate flow loop will have at least 3 litres per minute of flow, meaning that you have 3 litres (12000 joules to raise temp by 1C) passing through the waterblock which is generating only 6000 joules of heat. Thus the temperature only raises by .5C. It is the incredibly high specific heat of water that makes it a good medium for heat transference.

Click to expand...

Notice when considering the temperature increase in the water due to a single waterblock is dependent upon two things flow and the heat being produced. Lower flow will result in a higher temperature increase, as will an increased heat output. The design of the waterblock does not have an effect upon the amount of heat being put into the water. More efficient waterblocks can remove the heat better, meaning that the processor is kept at a lower temperature, but the same amount of heat is being removed.

The point of the calculations for the waterblock is to show that once the water has reached its equilibrium temperature--that is the temperature at which the radiator is dissipating the same exact amount of heat equal to the sum of all the components--that the temperature of the water varies by very little throughout the loop.



Now that my little science lecture is over, let me add a little bit to the thread. As I have just shown, temperature should not be the primary concern when debating series or parallel. The other big difference between series and parallel is resistance. In this manner you can think of a watercooling loop as being similar to that of an electrical circuit. Two equal resistors in parallel have a lower resistance than either of them, while the same two in series have a resistance equal to the sum of their resistances.

What does resistance have to do with watercooling? The higher the resistance the slower the flow rate. The slower the flow rate the less effective the waterblock is at removing the heat from the processor, and thus the higher the temperature the processor will be. Therefore the advantage of running components in parallel is to increase flow rates. However there is a big caveat to this. Just as in an electrical circuit the current will follow the path of least resistance, the same is true for watercooling. This brings up the problem where if the CPU path of your loop has the most restriction, it will receive the least flow. Of course that is exactly the opposite of what you would want, therefore you must be very careful when planning your loop and make sure that components get enough flow.

Why then do people always recommend running your loop in series? The hassle of balancing flow in parallel is a very tricky one. Modern pumps now have enough power that the restriction of running two or three, or even four waterblocks in series is not an issue. We have the luxury of being able to take the easy way out and run everything in series while still maintaining excellent flow rates.


I am very interested in seeing how the OP's project works out. True parallel loops are rarely done because balancing the flow is tricky, however if it is done properly it should yield some nice results. Will it be better than the same loop in series? It is hard to tell, but some custom water manifolds will certainly make the thing look like it means business.

I am very interested in seeing how the OP's project works out. True parallel loops are rarely done because balancing the flow is tricky, however if it is done properly it should yield some nice results. Will it be better than the same loop in series? It is hard to tell, but some custom water manifolds will certainly make the thing look like it means business.

Click to expand...

Yeah what he said, and al the science stuff too

I agree I would LOVE to see the orginal OP create his 1st design.

Why?

For one, its different and creative. It looks like fun to build and test, and isn't that why we mod in the 1st place? For enjoyment? Whether its discovering something new, or trying to get ever last OC we can.. we do it for the enjoyment.

And becuase i think it will look awesome and like the poster above said "look like it MEANS business" i like it.

I think his orginal design is plausible enough, and will give fine cooling based on the science of water cooling as discussed above.

As Nike said.. "just do it"