Single line water cooling or split line

[Vector86]

Would it be more effecient to split the water line comming from the pump multiple times to give each cooled component the exact same temperature of water instead of going from GPU to CPU to HD and so on. It sounds like it would be better in theory because the next item in the line won't receive water previously heated from the component before it. Would this be bad on water flow effeciency with all the splitters?

[fhorst]

It will be bad on water flow, not due to the splitters, but due to the extra tubing lines.
Lets say your flow is 100%, if you split it, it will have 50%, if you split it again, it will have 25% flow.

Less flow is less pressure, and we need pressure to create turbulance in the waterblock, to get a better cooling performence.

Look at it from an other point of few. Are you going to OC your hard drive? does it generate 100W of heat? most likely not. Your hard drive generates 5 to 10W of heat.

When you look at the point that needs the cooling the most, it's your CPU, so you want to cool that the best.
As you have cold water in your system, why not use it also for the rest, AFTER the cpu?

After the CPU, the GPU needs the most cooling, after this your North bridge, and south bridge, Mosfets, PSU, Memory etc.
Your hard drives don't need watercooling, but it will keep the temperatures in your system down, so that will help al stable OC. Also the lifetime of your harddrive will double when they are watercooled.

One other way to look at it-> why watercool all the rest? it will add more heat in your water, so your radiator needs to disipate more heat. If it is capable of handling it, no problem.
If it is not, get an extra or better rad, or don't watercool the rest

If you just want to go for a total watercooled system, and you are not into overclocking, your CPU temps will rize about 5 to 10 degrees, but still will be better then a standard intell air cooling

[killernoodle]

Actually, no. It is much like having resistors in a series vs resistors in a parallel circuit. Parallel resistors actually reduce the resistance vs a series circuit.

Think of it this way, if you split the lines you will have less water flowing through each block and with less velocity and pressure. Inline will cause more pressure on the first blocks, equal flow throughout the blocks, and an increase in temp from block to block.

It is best to split the flow of water after the CPU block, kind of like what bladerunner has done. For instance, I am going to put in a GPU and NB block in my system. Each will be connected to an outlet port of my whitewater so each gets about 1/2 of what water is flowing through the whitewater and with half the velocity. The cooling power will be reduced, but it will be easier to plumb and the restriction in the system will be far less, giving the CPU more cooling ability.

[Althornin]

Originally Posted by fhorst

Less flow is less pressure, and we need pressure to create turbulance in the waterblock, to get a better cooling performence.

less flow is...less flow. Pressure DROP (refered to as "head") is related to flow...but you seem to be mixed up.

When you look at the point that needs the cooling the most, it's your CPU, so you want to cool that the best.
As you have cold water in your system, why not use it also for the rest, AFTER the cpu?

Your water temps will rise less than a degree after passing through your CPU waterblock. Seeing as a millileter of water takes a calorie to raise one degree C, and assuming 2 lpm of flow, you can deal with 140 watts of heat before you raise a single degree per pass. Order in the loop is insignificant, really.

Also the lifetime of your harddrive will double when they are watercooled.

source?

If you just want to go for a total watercooled system, and you are not into overclocking, your CPU temps will rize about 5 to 10 degrees,

lol, how do you figure?

[#Rotor]

If you have a couple of restrictive, high turbulence block, that you want hooked up, putting them in series will inevitably result is a very very low overall flowrate. The pump will be constantly doing some unpaid overtime, and we know that ain't good.

putting them in parallel, on the other hand, will give the pump some breathing room, and in certain instances you might even find that the per-block flow-rate might increase. For sure the overall flow through the pump and Radiator will dramatically increase...

if however you are stuck with easy-flow blocks, then series is unfortunately propably the best way for what you have.

[fhorst]

Originally Posted by Althornin

less flow is...less flow. Pressure DROP (referred to as "head") is related to flow...but you seem to be mixed up.

Less flow is lower pressure drop, as it is less restricting.

Originally Posted by Althornin

Your water temps will rise less than a degree after passing through your CPU waterblock. Seeing as a millileter of water takes a calorie to raise one degree C, and assuming 2 lpm of flow, you can deal with 140 watts of heat before you raise a single degree per pass. Order in the loop is insignificant, really.

How do you get that idea? :shrug:
This is depending on a lot of things, design of the block, water velocity, number of watts produced by the CPU, temperature of the CPU and water etc.
With your theory, you state it as a fact that all heat produced by the CPU is transferred into the water. And you are right.
Only problem is that you’re CPU will be about 50 degrees.... depending on the waterblock design. With a 20-25 degrees in temperature difference between the water and the CPU block, a 2lpm flow will absorb the calories, and your temperature won't go higher, if your water temp does not rise.

With a good design waterblock and system, the temperature difference between the CPU and the water can be as close as 5 degrees or less. (Full load)

All new blocks (RBX, Whitewater, Cascade) are designed to work with turbulence, as this dissipates the heat the best. For turbulence to work, you need pressure (=head). Water will go in the least restricted way. If you split before the CPU, the water "won't" go via your CPU block.

As there is only little water passing your CPU block, your temps will rise. Just kink your 1/2 tube, and you will get about a 2lpm flow. Put some load on your CPU, and see what happens! (this is how you will get a 5 to 10 degrees higher temp)
If you are right, a L20 (600l) should perform the same as an L30(1200l), with only a 0.5 degree of temperature change.

Just to keep things in the right perspective:
A really good designed watercooling system will keep your CPU temps about 5 to 20 degrees lower.

Compared to stock Intel CPU coolers, even a bad designed water cooling system will work better.

Ask yourself how far are you willing to go to get those few degrees out of your CPU core temp.

For the Hard disk statement, it's hard to find that one back.... :shrug:
I saw it some time ago on a hard drive vendor page, MTBF times compared to operating temperature. A hard drive running at 80 degrees had half the mtbf as the one running at 40 degrees.

[kusojiji]

If you have a big pump, parallel can work. I used a 560 gph pump with 3/4 inch hose. Split it with a Y that necked down to two 1/2 inch legs to the two water blocks. Recombined to 3/4 inch to the radiator. The 3/4 inch gave me volume, necking down to 1/2 inch gave me velocity. System worked well enough to put me at room temp at idle and 2.5 degrees above that at load.

[Skulemate]

fhorst, your last post is quite hard to follow. However, I will try to offer some help if I can. As Althornin alluded to, the pressure drop encountered with a system is directly related to the flow rate achieved, and is generally a function of the flow rate squared. And he is correct when he states that there will be a very small rise in the coolant temperature after cooling a CPU, given a typical flow rate and heat load.

I think that you may be confused by the relationships between flow, pressure and the fluid's velocity. All of the blocks you have mentioned work mainly because of a jet impingement effect that is the result of carefully placed jets that serve to increase the water velocity, aiming it toward the area directly above the die of the CPU. Note that it is the increased fluid velocity that is key here, not the pressure, or even the (system) flow rate. In each of these blocks the increased velocity was achieved by decreasing the diameter of the nozzle (or nozzles in the case of the Cascade)... this serves to increase the water's velocity for a given flow rate, but there is a cost involved, and that is the high head loss associated with such a restriction.

This is indeed where pressure comes in, as the pressure gradient caused by the pump is the driving force causing flow within the system. Note that I said pressure gradient and not simply pressure... the actual pressure within the system (or the block) probably doesn't matter a whole lot in terms of cooling (i.e. you could pressurize the entire system if you liked, but there would be little point as the pump would still produce the same flow driving pressure gradient).

Your comments regarding the linear scaling that "would" occur by doubling the rated flow of one's pump shows that you have only a passing understanding of how a pump interacts with a given system. I would recommend that you take some time to read pHaestus' ProCooling Pump Comparison, as there is a very clear illustration that shows this interaction far better than I could explain it here. Keep in mind that there are far more variables that need to be considered besides the maximum (zero head) flow. I hope this helps.

[Gooserider]

My reccomendation for all multiblock systems, especially if there is not much data on a given combination of blocks is to do "Bucket testing" of any reasonable combination of components to see what gives the best flow rate results.

Bucket testing consists of simply building up a system with the components you'll be using, with the exception of the pump intake and system output. The intake should connect to an effectively unlimited water supply, the out put of the water system (which would normally go to the pump intake) should be a hose that can be aimed into a bucket. Ideally the block positions and hose lengths should match what will be used in the system, but it is OK if they don't as long as they stay about the same for all tests. To do a test, simply set up a test configuration and get it running long enough to purge any air out of the setup. Then put the output hose in a bucket, and time how long it takes to fill. If you know the size of the bucket, it is simple to calculate the flow rate.

Try different combinations to see what gives the best flow rates, both for the system and the individual blocks. IMNSHO, a few tests will beat all sorts of theoretical discussions.

That said, here are some of the guides that I use...

1. When picking a pump, high head is more important than high flow, almost to the point where the flow doesn't matter that much.

2. Priority should go to CPU cooling, then GPU cooling, followed by everything else.

3. In a series system, the flow will be NO MORE than what is allowed by the most restrictive block.

4. In a parallel system the flow will be the sum of what can flow through each block, (up to the input flow)

5. A restrictive block in series with a non-restrictive block will kill the flow in the non-restrictive block.

6. Restrictive blocks will allow more flow in parallel, non restrictive blocks it doesn't matter that much between serial and parallel.

7. In general, you want all the flow through the CPU blocks, but can afford to split up the flow through the other blocks.

In my system, I had two moderately restrictive CPU blocks, and 4 highly restrictive drive blocks (no GPU or NB blocks at present) my pump is an Iwaki MD20RT, which has 14' of head, and more free flow volume than I could use. The drive blocks were so restrictive I knew I didn't want them in series with my CPU blocks, but that if I put them in parallel they wouldn't reduce the CPU flow measurably. This violated guide #7, but in this case I can get away with it. With the CPU blocks in series, I had a total flow through both blocks of 3GPM. If I put them in parallel, I got a lower flow of 2.5GPM per block, but the total of 5GPM was better for the system overall. Splitting up the drive blocks also improved their flow, so I ended up with 4 branches, two CPU branches at 2.5GPM each, and two drive block branches at about 0.2 GPM each, giving me a net system flow of about 5.4GPM. This is an extremely good flow rate, and even approaches the upper limit of what is reccomended for my size tubing....

Gooserider


I found that if I put

[Skulemate]

Originally Posted by Gooserider

...
3. In a series system, the flow will be NO MORE than what is allowed by the most restrictive block.

4. In a parallel system the flow will be the sum of what can flow through each block, (up to the input flow)

These points are perhaps a bit misleading. The resistance from blocks, fittings etc. is cumulative, so even though you may have a very restrictive CPU block, that GPU block will contribute to the system resistance, and will alter flow rate. In a parallel setup the flow rate will balance itself across the multiple paths such that the pressure drop along every path is equal. And again, the resistance of each path is cumulative, meaning that if there is both a northbridge and GPU block on a line, both will contribute to the resistance.

[fhorst]

Originally Posted by Skulemate

In a parallel setup the flow rate will balance itself across the multiple paths such that the pressure drop along every path is equal.

As I said... If your CPU block is restrictive, and your GPU/NB is not.... what will that do to the flow (and pressure gradient) in the CPU block?

You are right about "pressure gradient" (I guess, I'm not native speaking English)
For mee its all the same. The force that you need tho push the water through the tubes, rad and waterblocks
If you split the hose, thsi force will be split also. This force is needed to create a good jet. A better jet will give better temps!

[fhorst]

Originally Posted by kusojiji

If you have a big pump, parallel can work. I used a 560 gph pump with 3/4 inch hose. Split it with a Y that necked down to two 1/2 inch legs to the two water blocks. Recombined to 3/4 inch to the radiator. The 3/4 inch gave me volume, necking down to 1/2 inch gave me velocity. System worked well enough to put me at room temp at idle and 2.5 degrees above that at load.

Great pump& setup, even better results!

[Skulemate]

Originally Posted by fhorst

As I said... If your CPU block is restrictive, and your GPU/NB is not.... what will that do to the flow (and pressure gradient) in the CPU block?

You are right about "pressure gradient" (I guess, I'm not native speaking English)
For mee its all the same. The force that you need tho push the water through the tubes, rad and waterblocks
If you split the hose, this force will be split also. This force is needed to create a good jet. A better jet will give better temps!

Each component in the system will have a characteristic relationship between the flow rate and head loss. If your CPU block is the most restrictive thing in the loop then it may dominate the pressure drop of the system, but the NB block will also contribute. Note that for a given flow rate the pressure drop of the CPU block (any restriction actually) has only one value... it does not depend on the other components of the system.

If you have multiple paths you are not really splitting the pressure gradient. As I said before, this will be the same for all paths, assuming they split from and rejoin with each other in the same places. Obviously what is split is the flow rate of the fluid.

[Althornin]

Originally Posted by fhorst

This is depending on a lot of things, design of the block, water velocity, number of watts produced by the CPU, temperature of the CPU and water etc.
With your theory, you state it as a fact that all heat produced by the CPU is transferred into the water. And you are right.
Only problem is that you’re CPU will be about 50 degrees.... depending on the waterblock design. With a 20-25 degrees in temperature difference between the water and the CPU block, a 2lpm flow will absorb the calories, and your temperature won't go higher, if your water temp does not rise.

With a good design waterblock and system, the temperature difference between the CPU and the water can be as close as 5 degrees or less. (Full load)

All new blocks (RBX, Whitewater, Cascade) are designed to work with turbulence, as this dissipates the heat the best. For turbulence to work, you need pressure (=head). Water will go in the least restricted way. If you split before the CPU, the water "won't" go via your CPU block.

As there is only little water passing your CPU block, your temps will rise. Just kink your 1/2 tube, and you will get about a 2lpm flow. Put some load on your CPU, and see what happens! (this is how you will get a 5 to 10 degrees higher temp)
If you are right, a L20 (600l) should perform the same as an L30(1200l), with only a 0.5 degree of temperature change.

Ok, you are seriously mixed up here.
I gave you the math already. Look, the water absorbs the heat put out by the CPU. The only question is, what does the temperature difference have to be (between water and cpu) before it does this? The CPU will get hotter until the water absorbs all of its heat (discounting the somewhat minor secondary heat paths of the pins to socket and mobo). PERIOD.
Water flow can improve a blocks C/W, sure - and reduce CPU temps accordingly. However, that has absolutely nothing to do with how much warmer the WATER is...........it absorbs the same amount of heat regardless (again, neglecting secondary cooling paths).

For the Hard disk statement, it's hard to find that one back.... :shrug:
I saw it some time ago on a hard drive vendor page, MTBF times compared to operating temperature. A hard drive running at 80 degrees had half the mtbf as the one running at 40 degrees.

80 degrees is quite hot. I doubt your HD gets that hot (i have to think in F, so thats 176 degrees F, thats hot enough to burn your hand).

[Althornin]

Originally Posted by fhorst

As I said... If your CPU block is restrictive, and your GPU/NB is not.... what will that do to the flow (and pressure gradient) in the CPU block?

You are right about "pressure gradient" (I guess, I'm not native speaking English)
For mee its all the same. The force that you need tho push the water through the tubes, rad and waterblocks
If you split the hose, thsi force will be split also. This force is needed to create a good jet. A better jet will give better temps!

simply use restrictive GPU blocks...........................

You dont really understand flow, eh?
Its not like splitting the flow through two blocks means each block gets half the flow it got before. Each block (assuming the blocks offer the same flow resistance) gets MORE than half of what a single block would get before, and depending on the flow resistances, more than the total flow through two blocks in series!

[Gooserider]

Skulemate:
[Quote: Originally Posted by Gooserider ... 3. In a series system, the flow will be NO MORE than what is allowed by the most restrictive block. 4. In a parallel system the flow will be the sum of what can flow through each block, (up to the input flow)

These points are perhaps a bit misleading. The resistance from blocks, fittings etc. is cumulative, so even though you may have a very restrictive CPU block, that GPU block will contribute to the system resistance, and will alter flow rate. In a parallel setup the flow rate will balance itself across the multiple paths such that the pressure drop along every path is equal. And again, the resistance of each path is cumulative, meaning that if there is both a northbridge and GPU block on a line, both will contribute to the resistance.[/quote]
Technically you are right, however I have found that they are good working approximations. In a series system, the most restrictive block will dominate, with only minor effects from the less restrictive blocks.

In my own setup, either of my CPU blocks would flow just over 3 GPM by itself, I got 3GPM with them in series, or almost no additional pressure drop from the second block. When I put them in parallel, I got 2.5GPM through each block, or 5GPM total for the two.

fhorst:
As I said... If your CPU block is restrictive, and your GPU/NB is not.... what will that do to the flow (and pressure gradient) in the CPU block? You are right about "pressure gradient" (I guess, I'm not native speaking English) For mee its all the same. The force that you need tho push the water through the tubes, rad and waterblocks If you split the hose, thsi force will be split also. This force is needed to create a good jet. A better jet will give better temps

Note that I do NOT reccomend splitting flows prior to the CPU block in most cases. I think it is usually best if all the flow goes through the CPU block(s) first, THEN split the flow to the other blocks which don't need as much cooling. This maximizes the flow through the entire system, and the pressure drop across the CPU block.

The ONLY time I reccomend splitting the flow before the CPU block is if the blocks on the other branches are so restrictive that having them in series reduces the flow through the CPU block significantly (which can be found by the bucket test)

If, when you split the flow after the CPU block, there is a great difference in the flow through the parallel branches, and the low flow branch is not cooling adequately I would reccomend re-arranging the branches to get a more even flow, and / or putting a restriction in the high flow branch in order to force more into the low flow branch. If all the branches are cooling adequately, then don't do anything that would reduce the total flow.

Perhaps I need to add another guide:

8. The ultimate objective should be to maximize the flow through the CPU block(s), and have good flow through the GPU block, with adequate flow through the remainder. Remember that except for possibly the GPU, the other blocks don't have to get rid of that much heat, so any flow at all is probably enough.

Gooserider

[Vector86]

Thanks for all the info guys, this forum is always good for a friendly WC debate

[Thiago_brazil]

Originally Posted by Vector86

Thanks for all the info guys, this forum is always good for a friendly WC debate

yes!! i like too this forum