Pumps in series vs. Parallel

[Turtle 1]

Now I really didn't want to talk about the DDC pumps at all . Flow and head is a result of both design and work. If you increase the work(flow & head) that work is done believe it or not by the motor. What is the rated horse power of DDC pump ??? anyone.
Now look at the D5 and what is the ratedHP of that motor???

When people say things like this it scares me

height of the Fort-Pak's inlet was adjusted so that it was approx. 4.5" above the water level in the reservoir that the pumps were drawing from (to negate any gravitational effects such as siphoning).

Siphoning is not a gravatational effect . Draining is however.

[montyshaw]

Now I really didn't want to talk about the DDC pumps at all . Flow and head is a result of both design and work. If you increase the work(flow & head) that work is done believe it or not by the motor. What is the rated horse power of DDC pump ??? anyone.
Now look at the D5 and what is the ratedHP of that motor???

When people say things like this it scares me

height of the Fort-Pak's inlet was adjusted so that it was approx. 4.5" above the water level in the reservoir that the pumps were drawing from (to negate any gravitational effects such as siphoning).

Siphoning is not a gravatational effect . Draining is however.
__________________

Is this where more than one person is posting as Turtle1? Because that last post made no sense at all.

]Monty[

[Turtle 1]

No its david posting and it made know sense because you didn't understand what I was saying . For the sake of not wanting to be banned lets just leave it there.

[conference]

now I'm really confused...

[Cathar]

Turtle1, people have run the physical tests, and have crunched the maths.

Even today, in series is still better most of the time. Not always though. It depends on the pumps used and it depends on the back-pressure.

You simply can't wade in and make sweeping generalisations that everyone should put their pumps in parallel. It's simply not true for everyone, and it's not even true for more than maybe 10% of users.

You have a very unique setup you're designing Turtle1, with 3/4" tubing, large fittings on low-flow-resistance PA160.1's, and so on. I don't even know what waterblocks you have on your system or what configuration you're putting them in. Series? Parallel?

Petra's tests show us that in series in best for the average user who has CPU block, maybe a GPU block, 1/2" ID tubing or smaller, a double-pass radiator, and any of the commonly used pumps.

Turtle1, in the other thread you were talking about wanting to push 8gpm flow rates. I can tell you now that while what you're discovering is relevant to you, it's not relevant to 99.9% of people who water-cool.

As much as people want to attack the maths of it, we need the maths because the answer is not so simple that it could be explained to a 2yo.

I'll try to make it simple:

When we put two pumps in parallel, we double the peak unrestricted flow rate. If our loop is so unrestrictive that with just one pump that the main limiting factor is the single pump's peak flow rate, then yes, it'll be better to put two pumps in parallel.

When we put two pumps in series, we double the pressure head. If our loop is restrictive enough AND we're not currently limited by the peak flow rate of a single pump, then having the pumps in series will be better than parallel.

Where is this "magical" crossover point? That's where the need the maths to find out. I tried to work out a rule of thumb for it in the past, and it roughly worked out to 70% of the pump's peak flow rate.

- If with a single pump the system's flow-rate is greater than 75% of the pump's rated peak flow rate, then having the pumps in parallel with be better.
- If with a single pump the system's flow-rate is less than 70% of the pump's rated peak flow rate, then having the pump in series will be better
- Between 70-75% is where the cross-over point usually will lie, and you can go with either.

If you don't care for the maths of it, follow the above 3 rules, and you'll likely never go wrong.

[Turtle 1]

Ok . You do undestand how a motor works correct. You can load a motor only so far before the failer rate goes sky high. In a dc motor that uses brushes the brushes usally go bad in a well made dc motor with brushes these bruses are replaceable.

Now there are more and more brushless motors appearing every day ,

Now in a normal DC motor if the brushes don't go bad than its normally the armature were the windings are located the flow of electrity threw this windings that react to magnits increase the flow of electrity threw the windings and more heat is created this flow is given in watts, The increased work load can be measured in amps.

when you see the wattage rise in a motor there is more electricty flowing threw those windings. if the work load rises above the design of that motor = hp those windings will burn out.

[Cathar]

Agreed. Which is in fact an argument for lower-flow. Pumps draw more power and operate under a higher load at higher flow rates.

Still, so long as you're operating at, or below, the BEP of the pump, only an extremely stupidly poorly designed pump's motor/electronics will be unable to handle that. That'd be a fault of a specific pump, not a general case.

[Turtle 1]

Turtle1, people have run the physical tests, and have crunched the maths.

Even today, in series is still better most of the time. Not always though. It depends on the pumps used and it depends on the back-pressure.

You simply can't wade in and make sweeping generalisations that everyone should put their pumps in parallel. It's simply not true for everyone, and it's not even true for more than maybe 10% of users.

You have a very unique setup you're designing Turtle1, with 3/4" tubing, large fittings on low-flow-resistance PA160.1's, and so on. I don't even know what waterblocks you have on your system or what configuration you're putting them in. Series? Parallel?

Petra's tests show us that in series in best for the average user who has CPU block, maybe a GPU block, 1/2" ID tubing or smaller, a double-pass radiator, and any of the commonly used pumps.

Turtle1, in the other thread you were talking about wanting to push 8gpm flow rates. I can tell you now that while what you're discovering is relevant to you, it's not relevant to 99.9% of people who water-cool.

As much as people want to attack the maths of it, we need the maths because the answer is not so simple that it could be explained to a 2yo.

I'll try to make it simple:

When we put two pumps in parallel, we double the peak unrestricted flow rate. If our loop is so unrestrictive that with just one pump that the main limiting factor is the single pump's peak flow rate, then yes, it'll be better to put two pumps in parallel.

When we put two pumps in series, we double the pressure head. If our loop is restrictive enough AND we're not currently limited by the peak flow rate of a single pump, then having the pumps in series will be better than parallel.

Where is this "magical" crossover point? That's where the need the maths to find out. I tried to work out a rule of thumb for it in the past, and it roughly worked out to 70% of the pump's peak flow rate.

- If with a single pump the system's flow-rate is greater than 75% of the pump's rated peak flow rate, then having the pumps in parallel with be better.
- If with a single pump the system's flow-rate is less than 70% of the pump's rated peak flow rate, then having the pump in series will be better
- Between 70-75% is where the cross-over point usually will lie, and you can go with either.

If you don't care for the maths of it, follow the above 3 rules, and you'll likely never go wrong.

True Cathar In a single loop system were many blocks are used I agree .

What I am saying tho is this. Todays watercooled systems are a stop gap . There is nothing new about water cooling been around along time.
The evolution of water cooling in Pc's is still very early in its stages.

And befor it ever matures will likely be replaced by something better.

Thats all I am saying . I am not trying to sell anyone anything here or online . I am a neutral on this . But people who do profit from water cooling have put out a lot of BS. I know what a properly designed system cost. and the amount of room it requires. Its expensive . All I am saying Sir is this isn't the correct way it should be done. It works this is true . But thats all it does. Soon enough all will see the perfect storm . Until than your right . I can't just generalize. But I won't and can't conform to something that years of experiance of taught me is just plain wrong. I understand that were all working with products that are available presently.
I still have the same problem today as I did 4 years ago . THe DAM pumps.

But we do have good enough pumps now that can fill my needs but their still not were we need them to be.

[Quazi]

Besides that, I don't think anyone is interested in running any Iwaki pump in a series or parallel. Because like you asked, "Where exactly are you going to mount these pumps in the basement...". Now one of the reasons some here decided to run pumps in a series, and we're talking DDC-2's here, is to at least try to equal a particular Iwaki pumps' performance (not sure I remember which one that was mentioned so often). I know I am running pumps in a series to try to increase head pressures given the Storm is so restrictive.

[Cathar]

I see it in a more wholistic sense.

The problem is not the pumps. The problem is far more removed from water-cooling.

The problem is ATX/BTX and case sizes.

It is from ATX/BTX, and the corresponding space constraints of GPU/CPU tube routing that dictates the current mass-market PC water-cooling regime.

If people are profiting from water-cooling today, it is because they are selling people what they want, and what people want are solid in-case solutions. There are those on the fringe who will take it further, but if you want to speed $1000's on a water-cooling system, go with phase-change and do it properly.

This is where water-cooling has an upper bound. It is a quiet and effective high-powered cooling solution for system components, but once it gets too expensive, you may as well go with phase-change.

The biggest problem with pumps nowadays is their heat-dump. We're pretty close to the balance point of where greater flow rates just means more pump heat. The only way to compensate for that nowadays is to add more radiator capacity, and we're talking like 2 x PA120.3's if we want to keep it quiet as well.

Once we factor in the wholistic aspects of cost, space, & noise, we are pretty close to the limits today. Yes, we can do a fair bit better, but only by dropping one or more of the cost/space/noise trio of constraints.

[virtualrain]

lol,

Sounds like the golden rule of Network Engineering and Design. Fast, cost effective, robust. At any given time you can have two, but never all three. There's always a trade off

andyc

That's true of any project... building a bridge or a water cooling system!

This thread has been very enlightening.

I believe that the number of loops and pumps and selection of equipment must be desgined as a system with the needs and requirements of the project dictating these decisions. Not everyone has the same requirements when it comes to cooling. There's a time and place for a Zalman Reserator and there's a time and place for multiple loops each with redundant pumps. There is no single "best" way to water cool. There is a "best" way to cool a particular system with particular goals.

Because of this need to design in relation to the requirements, informative and educational threads like this are critical to folks making the right design decisions. This is great stuff! In particular, this is an extremely valueable guideline anyone can use...

- If with a single pump the system's flow-rate is greater than 75% of the pump's rated peak flow rate, then having the pumps in parallel with be better.
- If with a single pump the system's flow-rate is less than 70% of the pump's rated peak flow rate, then having the pump in series will be better
- Between 70-75% is where the cross-over point usually will lie, and you can go with either.

[Cathar]

Sounds like the golden rule of Network Engineering and Design. Fast, cost effective, robust. At any given time you can have two, but never all three. There's always a trade off

For PC water-cooling it's:

Lowest Noise
Highest Performance
Smallest Footprint

Pick any two. Anything in-between is a trade-off.

Cost is a real factor too, but only if you want it to be. Can be as ghetto or as refined as you want, but the above trade-offs will always apply.

[virtualrain]

For PC water-cooling it's:

Lowest Noise
Highest Performance
Smallest Footprint

Pick any two. Anything in-between is a trade-off.

Cost is a real factor too, but only if you want it to be. Can be as ghetto or as refined as you want, but the above trade-offs will always apply.

True!... and if you are going Xtreme on any one of those, you can forget trying to constrain it with the other two!

[Martinm210]

Pumps in series will still perform better than parallel from what I can see. I've been working on some graphs here:


Series:


Parallel:


You have to be in an extremely low restriction condition before parallel becomes the better option.

[Turtle 1]

As has already been pointed out by cathar and myself . Inorder to see the benefits of Pumps in parallel you have to have a PROPERLY setup system .

I noticed in your charts that you have head listed as PSI. I thought head was rise given in feet or meters and not PSI . It really doesnt make a lot of sense to try and push double the flow threw 1 3/8 tube .

However use a 2 port manifold and the improvements will be seen . Thats what I am saying here. More and more watercooling guys are going to multi loops. You can do this 2 ways the right way and the wrong way. The use of a manifold along with shutoff valve to control flow is the right way .

[IanY]

Wouldn't multi-loops still be better than a parallel setup?

[Martinm210]

As has already been pointed out by cathar and myself . Inorder to see the benefits of Pumps in parallel you have to have a PROPERLY setup system .

I noticed in your charts that you have head listed as PSI. I thought head was rise given in feet or meters and not PSI . It really doesnt make a lot of sense to try and push double the flow threw 1 3/8 tube .

However use a 2 port manifold and the improvements will be seen . Thats what I am saying here. More and more watercooling guys are going to multi loops. You can do this 2 ways the right way and the wrong way. The use of a manifold along with shutoff valve to control flow is the right way .

My graphs are PSI simply because that's the unit I prefer to look at. The Y axis of a P/Q chart on any pump is pressure vs flow rate. Pressure can be any sort of unit. Seems like alot of manufactures for water blocks are presenting pressure drop vs flow rate charts in PSI vs GPM, so those are the units I've selected here.

If you want feet of water, simply multiply the PSI by 2.307.
http://www.geagolf.com/GEA%20Pressure%20and%20Flow.html

You could take advantage of a parallel pump system if you can reduce the pressure drop vs. flow rate enough, but enough is an extreme amount. By the time you add even the most free flowing of water blocks and radiator to any loop you added enough resistance that series appears to always be the better option.

I'm planning some flow rate tests with my D5s in the next week. Do you want me to try anything out?

I have:
D5 x2
D-tek Fuzion with nozzle
MCW60
MCR320 x2
T-lines x2

[Quazi]

http://www.xtremesystems.org/forums/...d.php?t=148668

Look at this thread where I discovered that the difference between the DDC2 and the D5 was .29GPM in a fairly restrictive loop with 2 MCR320s in series. I can partially agree with the OP on account that in the end of the tests, the DDC2 has more head left over to go through more things than the D5. However, the D5's increased flow seems to make up for its lack of head. If the flow of the DDC2 w/petra's top was 100GPH less, then I believe it would have lost to the D5.

I hope that helps.

I didn't know that there was a battle, much less that you won it.

Sorry Sick. I couldn't help myself. HEHE

[Turtle 1]

My graphs are PSI simply because that's the unit I prefer to look at. The Y axis of a P/Q chart on any pump is pressure vs flow rate. Pressure can be any sort of unit. Seems like alot of manufactures for water blocks are presenting pressure drop vs flow rate charts in PSI vs GPM, so those are the units I've selected here.

If you want feet of water, simply multiply the PSI by 2.307.
http://www.geagolf.com/GEA%20Pressure%20and%20Flow.html

You could take advantage of a parallel pump system if you can reduce the pressure drop vs. flow rate enough, but enough is an extreme amount. By the time you add even the most free flowing of water blocks and radiator to any loop you added enough resistance that series appears to always be the better option.

I'm planning some flow rate tests with my D5s in the next week. Do you want me to try anything out?

I have:
D5 x2
D-tek Fuzion with nozzle
MCW60
MCR320 x2
T-lines x2

Yes I would If you have the equipment.

Run 2 in parallel to a 4 port manifold with shut off valves useing 3/8 tubes. The pumps should connect to the ends of the manifolds. 1 pump to each end of the manifold . 2 manifolds are required 1 return and 1 supply .

use 2 cpu water blocks and 2 gpu water blocks measuring the preasure drop of each circuit. The flow threw each circuit.(loop) THe shut off valves are used to tune the flow to each circuit .

Than report the pressure drop . And flow for both serial and parallel

If you can do this than perform same test running in serial but this time only 1 line goes to the manifold. The results I am sure you will find very interesting . Use THE DDC also than show us the plot chart.

[Turtle 1]

Use something like this its cheap but will allow you to do the test. There are much nicer ones done in stainless but for testing 2 of thes with 6 ports will do nicely

Any plumbing shop should have these

http://spapartsnet.com/Spa-Replumbin...bing-Manifolds

Shut off valves are required only on the supply lines to the water blocks


Rememder use 3/8 line or 5/16 no 1/2 . 5/16 will show the most interesting results .

Its ok to use 1/2 from the pumps to the manifold we wouldn't want that to sqeu the results same on the return to the manifolds the loop should go like this .

1/2 from pump to rad from rad to manifold 1/2 from supply manifold to water blocks 3/8 or 5/16 (preferred) from water block 3/8 or 5/16 to second manifold. 1/2 back to pumps

[Cathar]

Than report the pressure drop . And flow for both serial and parallel

If you can do this than perform same test running in serial but this time only 1 line goes to the manifold. The results I am sure you will find very interesting . Use THE DDC also than show us the plot chart.

The obvious third test would be to then put everything in serial (pumps leading to blocks) and observe the flow rate.

With the manifold method Turtle1, are you actually measuring the flow-rate through each block/channel?

Is the manifold straight? It would seem to me that two streams coming in from each end and colliding in the middle would be adding a lot of unneeded resistance.

There is a real advantage to Turtle1's method though, and that is you can easily tune the system such that the pumps are operating at their Best Efficiency Point (BEP), or "duty point", meaning that you're getting the most hydraulic power that the pumps have to offer (provided that this occurs at some point less than having all of the shut-off valves wide-open).

[Turtle 1]

I really don't care how he brings the supply lines to the manifold he can use 2 across from each other . But in real world were your actually installing inside a box the manifold would be smaller so coming in from both ends would be the best its a straight shot across the manifold . to the supply ports.

Ya if he properly tunes each port with the shut off valves the result will be eye opening .

For a system like this there is one thing you must do forget what is known about a single loop to a degree. In a manifold there is 1 thing that all should know . Its very easy thing to remember this applies to fluids/ pneumatics/ electrity. FLOW will take the path of least resistance.

I think we all know this but one has to think in those terms .

[Turtle 1]

Cathar you should see what I did with your storm water block. Its your basic design but I did a small amount of tweaking with it . I can't say what because I am tring to sell the design. But I have tested against the dreaded fusion and the VORTEX (still very angry about that name.) Guess what Your modded storm wins by 1.2c Its was a great design fella. It doesn't beat my own design tho . I have it beat by .8 C . I am trying to sell the design to Koolance . I would like to see them with this design .

[Turtle 1]

Besides that, I don't think anyone is interested in running any Iwaki pump in a series or parallel. Because like you asked, "Where exactly are you going to mount these pumps in the basement...". Now one of the reasons some here decided to run pumps in a series, and we're talking DDC-2's here, is to at least try to equal a particular Iwaki pumps' performance (not sure I remember which one that was mentioned so often). I know I am running pumps in a series to try to increase head pressures given the Storm is so restrictive.

Actually in a one loop system your right. Its overkill.

But . This is importnant. In a 5 or 6 circuit system like I am working on 2 RD 20 could prove to be very very useful. I am working in 2 differant directions. 1 high flow system were I don't need that kind of head. Than there is my orginal work were I modded a Storm block The result are very good for both. The high flow system as of now is a little bit better than the High pressure system . But its so close that I keep looking back at the modded Storm . I feel that if I use my top and modd it for high pressure It will be the winner. Bob should finish machining the parts tomorrow so I should beable to test it next week . I am hoping for a 2c advantage over my modded storm . If it works I will have a top that can be used for either High flow or high preasure.

[Cathar]

Cathar you should see what I did with your storm water block. Its your basic design but I did a small amount of tweaking with it. I can't say what because I am tring to sell the design.

You're trying to sell a tweak to someone else's design as a whole design?

Let me guess. Either cutting radial channels or pins into the copper surrounding the cupped area, or drilling holes and soldering in copper pins around the cupped area?

It's a snap to boost the size of the cupped area of the Storm design to boost performance and lower pressure drop for modern IHS CPUs. It works extremely well to do so. Would even beat any small tweak/mods handily. It just wasn't done by Swiftech for commercial reasons specific to their operation.

But I have tested against the dreaded fusion and the VORTEX (still very angry about that name.) Guess what Your modded storm wins by 1.2c

Doesn't surprise me in the slightest. The design still has the single highest proven h(eff) per unit of heated contact area of anything out there, to my knowledge anyway. Jet impinged mini-pins come very close though. Modern manufacturing techniques means that it's cheaper to manufacture mini-pins than carve textured cups in copper, and on a large-scale manufacturing cost-performance benefit scale, mini-pins win. The Storm design is somewhat doomed to remain an expensive boutique design made in small quantities as a result.

I have extensions to the design as a result of Storm/G7 research that I'll resurrect one day, but not any time in the very near future.

It doesn't beat my own design tho. I have it beat by .8 C.

Heard it before a thousand times. Time doesn't stand still, and neither do I. If you want an apples-apples comparison with what the Storm design (not the specific implementation) is really capable of, stick it on a bare-die CPU. That'll let you know its ability per areal unit of heat-load. Testing the existing 3yo implementation on modern CPU's where its cooling area is at a disadvantage is like running a race against the an old champion horse after shooting two of its kneecaps out. If you're still only beating it by a small amount...

I am trying to sell the design to Koolance . I would like to see them with this design .

If you're trying to sell Koolance (or anyone else) a variation of the Storm design, I won't be happy. It's not yours to sell.

If it's your own, Good Luck!