Pumps in Series Vs. Parallel Graph

I know this has been talked about and explained before, but I got bored and wanted to make a new one to hopefully explain this a little better to newer folks in the forum.

This is why you want to run pumps in series over parallel. Most water cooling systems are fairly restrictive (and getting worse as designs continue to get more restrictive to squeeze every last drop out of the waterblock), so that's where series and the stacking of pressure comes in favor.

Parallel is only really better when you're talking extremely low restriction setups, pretty hard to do actually. And in that case, you're already moving plenty of flow with the single pump.

Anyhow, hope this helps someone...I was bored...:D

The single pump curve was one that I tested with an XSPC regular top. The two series/parallel curves are theoretical. Parallel = 2X Flow rate numbers. Series = 2X pressure numbers. I can't remember the exact gear for the High and low pressure drop curves, but they were pretty extreme examples if I remember right. Regardless, it gives you the idea. We have high restriction systems and that's where series is the better option...:up:

http://img14.imageshack.us/img14/643...sparallel2.png

thx martin for the graph setup.....very very much appreciative.

Awesome information Martin, thank you.

Waterlogged had given me a rough explanation in the other thread, but the graph helps understand.

Martin, I think this just goes to show that your real talent is being able to clearly convey concepts through the use of illustration (more specifically, the use of Excel).

I'm fine with explaining things to people and occasionally drawing something freehand to illustrate my point, but I'm an utter failure when it comes to the use of spreadsheet software. As they say, a picture is worth a thousand words... and this one certainly saves some of us a lot of typing :up:

Nice Martin. In 'normal computer restriction loops' you gain about 33% flow rate with 2x pumps in series. Obviously the more restrictive the loop the better serial pumps scale.

So I was right and it does improve the flow and pressure performance then.

Ha, smoke that Vapor! You know I am joking....right?

***EDIT*** Actually now that I go back and read it again you agreed with me right?!? Damn it I am confused again!

Sweet write-up Martin :up:

Clear explanation and a GREAT chart! Also probably worth a mention that increases in flow with high-restriction parts *usually* net more performance than gains with low restriction parts. Not always the case of course :p:

Regardless, it's awesome to see laid out so plainly. :)

Quote:

---

Originally Posted by Sadasius

So I was right and it does improve the flow and pressure performance then.

Ha, smoke that Vapor! You know I am joking....right?

***EDIT*** Actually now that I go back and read it again you agreed with me right?!? Damn it I am confused again!

---

I'm really not sure what you're alluding to :wth:

The loop order debate? That has very little to do with this :shrug:

Quote:

---

Originally Posted by nikhsub1

Nice Martin. In 'normal computer restriction loops' you gain about 33% flow rate with 2x pumps in series. Obviously the more restrictive the loop the better serial pumps scale.

---

Hey nikhsub1,

IS there a way you can get a hold of the deltaT "interactive chart" from procooling of back-in-the-day?

I would love to see someone with current test-methodologies that are on-point like martin's use that to make the FAQ section as robust as can be.
Maybe it can even be combined with data like this in a way.

Quote:

---

Originally Posted by Vapor

I'm really not sure what you're alluding to :wth:

The loop order debate? That has very little to do with this :shrug:

---

Nah not that...This-> http://www.xtremesystems.org/forums/...d.php?t=224966 Would have explained myself better but was pretty damn busy at work.

Now I just want to see the XSPC duel ddc res (the new thing) tested with the same conditions :p and see how well(bad) it does, but very nice info.

Very nice, thanks for the info, Martin.

martin,
did the switch to series everything is better now vs what I had in parrell. Better as in one pump does not seem to overpower the other pump and send things back into the res from the Y splitter. I can now feel the flow of the water in my tubes vs before.

Quote:

---

Originally Posted by JOCKTHEGLIDE

martin,
did the switch to series everything is better now vs what I had in parrell. Better as in one pump does not seem to overpower the other pump and send things back into the res from the Y splitter. I can now feel the flow of the water in my tubes vs before.

---

Awesome, I hoped that would take care of the problem, glad it did..:up:

Great illustration! I can imagine this will be a much linked thread when questions about dual pumps pop up. :up:

sticky this somebody

Awesome, thanks for that info :)

Quote:

---

Originally Posted by defect9

sticky this somebody

---

I agree alot f martions OPings should be stickied in general instead of doing a damn search for them all the time....STICKIE THIS NOW PLEASE :)

top notch as usual. Thanks for the info Martin.

What about the whole "pump downflow will have its impeller burnt out due to pressure from pump 1" talk ?

that is if you have mismatching pumps iirc.

Quote:

---

Originally Posted by Boogerlad

that is if you have mismatching pumps iirc.

---

Would the pumps be considered mismatched if you used a xpsc Res top for one and a regular xpsc acrylic top for the other?

For two pumps in series you should place the weaker pump feeding the stronger pump for maximum performance, otherwise the weaker pump would restrict the flow of the stronger pump. But as for a powerful pump "burning out" or overpowering a weaker second pump, I am not really sure that is possible, because the extra pressure will pass through the second pump (albeit reduced by the restriction of the pump). However, I suppose I can imagine a scenario, where if you had some kind of a super-powerful water-jet type pump, maybe you could do some physical damage to a weaker pump, but even then, I don't think so. To give you a real world analogy of what is happening when pumps are in series, suppose you have three pumps with 10 feet of head pressure each, and you need to get water out of a hole in the ground that is 25 feet deep. Now this hole is so deep that if you try to use any one of your pumps, there will not be enough head pressure to lift the water out of the hole (because each pump will lift the water only 10 feet). So what do you do? Put the pumps in parallel? No way, in parallel you will simply pump the water to a height of 10 feet, but three times faster. You need to add the pumps in series, thereby adding the head pressure of each additional pump, until you get enough pressure to get the water to a height of 25 feet in order to reach the top of the hole. With three pumps with 10 feet of head pressure each, in series, you should be able to pump the water out of a 25 foot deep hole. Imagine the first pump is at the bottom of the hole and, that each additional pump is mounted 10 feet above the other. The same principle will apply inside your computer case, except instead of overcoming gravity, you are overcoming restriction in your water blocks and the friction resistance of your tubing. And as Martin pointed out in his graph, the more restriction in your loop, the more you will benefit from pumps in series.

Guess it's time to change my loop around and put pumps in series. Oh well, it's time for a cleaning.

Quote:

---

Originally Posted by Martinm210

The two series/parallel curves are theoretical. Parallel = 2X Flow rate numbers. Series = 2X pressure numbers.

---

At first this sentence looked really bad. However, the "theoretical" GPM gains from adding a second pump in series are in line with the real world observations that have been posted, so I guess it can be as simple as multiplying by two...

Glad you guys brought this back up. I'll do some actual testing onb this and see how real world compares with some stand alone pumps.

I'll also look at pump redundancy.

OMG Martin is on a review frenzy... :lol:

:up:

Martin, I read in one your old threads that you recommended 4" tubing run between the pumps. Does this still hold true?

Quote:

---

Originally Posted by DexNfX

Guess it's time to change my loop around and put pumps in series. Oh well, it's time for a cleaning.

---

WOW

NECRO'd a thread from 5-2009.

80% of the people posting here do not even date back to that time. :rofl:

Quote:

---

Originally Posted by Martinm210

Glad you guys brought this back up. I'll do some actual testing onb this and see how real world compares with some stand alone pumps.

I'll also look at pump redundancy.

---

You'd have to start from scratch martin. Not unless u can recreate at least one of the old data you had in that bench, which im thinking is going to be very very very difficult.

Quote:

---

Originally Posted by JoeBar

OMG Martin is on a review frenzy... :lol:

:up:

---

I get an itch now and then...gotta scratch..:D

Quote:

---

Originally Posted by theseeker

Martin, I read in one your old threads that you recommended 4" tubing run between the pumps. Does this still hold true?

---

I've never done any testing on this, that was purely based on some rules of thumb I've read about installation of industrial pumps. I've got a water guy in my office that tells me 4XID or four times the pipe diameter is a minimum recommended elbow to pump inlet spacing. This probably depends somewhat on flow rate and turbulence of the stream itself, etc.

But again, never tested it myself, but it has been one of those nagging questions.

Quote:

---

Originally Posted by NaeKuh

WOW

NECRO'd a thread from 5-2009.

80% of the people posting here do not even date back to that time. :rofl:



You'd have to start from scratch martin. Not unless u can recreate at least one of the old data you had in that bench, which im thinking is going to be very very very difficult.

---

Are you admitting to being an old timer, you ole fart you!..:D

Regarding the parallel vs series. The recent Koolance bay res reviews was the first I've done much of any series type testing myself, and it does appear to double pressure head.

To test the theory, I just planning on something like this:

Test PQ

• Single Pump of a PMP450
  
• Two PMP450s in series with a short piece of tubing in between the two
  
• Two PMP450s in parallel using two Y fittings

Compare.

I'm sure there are some minor losses in the interconnecting tubing, but it should be roughly doubling the pressure direction for series and doubling the flow direction for parallel.

It shouldn't be all that hard, all I need is to make myself another copper Y that's very low restriction.:up:

You mean Y-s like these or some plastic ones?

Quote:

---

Originally Posted by churchy

You mean Y-s like these or some plastic ones?

---

No, my own custom made type like this:
http://img152.imageshack.us/img152/9...mcoppercc5.jpg

It's a fairly a gradual merge and made of 1/2" ID copper. It's not real pretty, but probably really low on restriction which should help it more closely match theoretical. I still have this one in my parts drawer, just need to make another.

The "mother of all" Y fittings..:D

that is a job very well done, looks great :)
my only question is: will the results using that equipment be truly indicative of that most users get in a typical usage scenario?
also, are the PMP450 going to be tested at variable rates, so that a comparison could be gleaned for low flow pumps such as the PMP400?

Quote:

---

Originally Posted by Computurd

that is a job very well done, looks great :)
my only question is: will the results using that equipment be truly indicative of that most users get in a typical usage scenario?
also, are the PMP450 going to be tested at variable rates, so that a comparison could be gleaned for low flow pumps such as the PMP400?

---

Well, that's debatable I suppose. I just wanted to give parallel the best chance possible since it does have to deal with extra fitting restriction over series. If it looses in the perfect low restriction fitting scenario than it surely would with added restriction.:up:

I'll give it a run, we'll take a look...then see where we to take it from there.

I've got the pump testing bench all set up, so these sort of things are pretty quick to check out..:up:

i have a parallel setup *of sorts* so this is very interesting to me. one loop is one block for cpu and low restriction rads, but the other has some fairly restrictive blocks, but they all empty into same res.

so i guess the point is, cant wait for your results:up:

Quote:

---

Originally Posted by Computurd

i have a parallel setup *of sorts* so this is very interesting to me. one loop is one block for cpu and low restriction rads, but the other has some fairly restrictive blocks, but they all empty into same res.

so i guess the point is, cant wait for your results:up:

---

Ahh...I see. Parallel pumps and parallel loops are sort of two separate things though.

This was talking about parallel pumps (Splitting before two pumps, then merging back together.).

In your case, parallel loops(Splitting after and merging before the pump), that's another topic all together. The trick is as you've mentioned the restriction of each loop. The lower restriction loop might receive 60% of the flow where the high restriction loop would only receive 40%..ect.

Nothing wrong with parallel loops, it's just it's a give and take emphasis of flow rate depending on the restriction of each loop. If you know the other blocks in your setup are higher in restriction, you're probably doing good.:up:

The hard part of parallel loops is the lack of pressure drop info. You could always do a bucket test yourself I suppose, but without some sort of info or quick flow test, it may not be apparent where the emphasis in flow is going. Pretty safe on SLI GPU blocks or dual rads etc though.

In most cases the differences are probably too small to measure, but parallel loops of mixed components should be taken with a bit more planning than single loop series. Series is just hard to mess up, you only have one flow rate to deal with so it's very easy for your average user.

well i learned something! thx for that :) now i get it :)