The 'pump after res' rule

[rioja]

There is some sort of the rule, that pump must be after res. But in my case this setup is uncomfortable a little, i would get more simple setup if pump was after rad or waterblock. So how important this, may I ignore this rule at this time? Or you guys highly don't recommend it?

[Petra]

The reason for the "rule" is because the pumps most commonly used in PC water cooling aren't self-priming pumps, they're gravity fed and must be fully primed prior to turning on. Plumbing the reservoir's outlet to the pump's inlet is the easiest way to insure a good, proper prime. Can things be done in a different order? Well, yeah... but it's usually not something that people will recommend because it complicates the filling and bleeding process.

[millertime359]

If you want to, you can do a t-line. It is a line that takes the place of the res and you would follow the same rule for the reason Petra described.

The reason why T-lines are not as good as a res is that the can be a bit hard to bleed and don't hold as much extra fluid. The extra fluid really only helps if you get a slow leak or some of your water happens to evaporate over time.

[Manicdan]

theres also the effect of pressure and vacuums, before the pump there will be a vacuum, after will be pressured, and it changes between them as the water flows through each restriction (cpu block, radiator and reservoir). to test it you can punch a needle sized hole in the tubing at each area and check if water comes out, or air goes in. with the way a res is built, its generally designed to work best as a vacuum and the pump pulls water out, rather than pushes water in. although i wont test it, it would be cool to see the effect of water flow if you put the pump before it, i cant imagine it would make too much of a difference, but that greatly depends on how much air is in the system, and how flexible the tubing used is. (i am not an engineer and i may have a few things wrong, do not hold me to these details as if i got it from a college textbook, lol)

[Kurz]

It also effects the amount of pressure the individual blocks get.
Since a Res is empty space it eats up the head pressure that gets sent to it.
When its the end of the loop it doesn't matter how little pressure it gets.

[Petra]

It also effects the amount of pressure the individual blocks get.
Since a Res is empty space it eats up the head pressure that gets sent to it.
When its the end of the loop it doesn't matter how little pressure it gets.

That's it... I give up.

[Waterlogged]

theres also the effect of pressure and vacuums, before the pump there will be a vacuum, after will be pressured, and it changes between them as the water flows through each restriction (cpu block, radiator and reservoir). to test it you can punch a needle sized hole in the tubing at each area and check if water comes out, or air goes in. with the way a res is built, its generally designed to work best as a vacuum and the pump pulls water out, rather than pushes water in. although i wont test it, it would be cool to see the effect of water flow if you put the pump before it, i cant imagine it would make too much of a difference, but that greatly depends on how much air is in the system, and how flexible the tubing used is. (i am not an engineer and i may have a few things wrong, do not hold me to these details as if i got it from a college textbook, lol)

In short, it's called cavitation, which pumps don't like very well.

That's it... I give up.

[millertime359]

That's it... I give up.

It's ok Petra.

[[XC] Oj101]

Have a cookie

[OldChap]

I just love reading explanations of scientific principles in layman's terms

Me? I know nothing about what is best but to date whether I've put the res before the rad or the pump or even after the pump ...the temps stay the same. So with no flow rate measurement etc and only air temps monitored and an external sensor on the pump outlet I can't tell because the difference seems so small

[Soulwind]

I just love reading explanations of scientific principles in layman's terms

Me? I know nothing about what is best but to date whether I've put the res before the rad or the pump or even after the pump ...the temps stay the same. So with no flow rate measurement etc and only air temps monitored and an external sensor on the pump outlet I can't tell because the difference seems so small

It's a closed loop.

Actual positioning isn't going to make a diddly squat of difference in that regards.

The reason for res leading into the pump is simply for ease of filling and bleeding, and to ensure that you don't run your pump dry.

Trying to gravity pour enough liquid into the loop to ensure that the pump is fully supplied with the res in a different place is a pain. Sure it can be done, but it's just a lot more difficult. Hence the recommended res into pump into rest of loop.

[shazza]

quoted from Soulwind: The reason for res leading into the pump is simply for ease of filling and bleeding, and to ensure that you don't run your pump dry.

That about says it all.

Still chuckling at the image Petra posted.

[rioja]

so as i understood, it is higly recommended to have a res before the pump
ok, will try organize such setup

[Shocker003]

That's it... I give up.

U just made my day

[eth0s]

Let me try to explain the reason for the reservoir before pump "rule", here goes: A Dry Pump = A Dead Pump.

So to avoid killing the pump, you must never, ever run the pump dry. So you always need water flowing into the pump. And that is why you should always mount the pump so that it is gravity fed by the reservoir. Now that is a "should" not a must. I would not recommend this for a n00b, but if you feel you are an experienced water cooler, then by all means, mount the pump anywhere you want, just make sure that you can fill the loop without running the pump dry. This can be accomplished by turning the case upside down, or on its side, or any other creative solution. Just make sure the pump is always fed with water, because contrary to what was said above, centrifugal pumps do not "suck" water, they push it. As such, there is no "vacuum" created at the input of the pump. This is some kind of superstition or fundamental misunderstanding of the nature of our universe that I cannot explain, but it persists, even in this so-called "information age". You would think that google could solve such ignorance, but alas, I am afraid it cannot.

Also head pressure is necessary to overcome restriction in the loop, not the empty space in the reservoir. To say the opposite is to speak nonsense.

[Jamesrt2004]

so as i understood, it is higly recommended to have a res before the pump
ok, will try organize such setup

do a t line?

[B_Boy]

if the pump is below the waterline of the res it wouldn't matter if its directly after it or not, it would still be gravity fed. so it cant run dry, unless your res is low on water...

correct me if im wrong.

[StAndrew]

Hahaha Petra! Priceless.

As for the question, I dont think its required. Just try to position my res as high in the loop as possible, then you can gravity feed my entire loop without even having to turn my pump on. If you cant get it high enough, placing in before the pump will probably be your next choice.

My last few builds have been with a T-Line as im running out of space. I hate T-Lines and would never recommend them as they are a pain in the @$$ to bleed. Also, if your pump is variable, when you bleed, do so on the lowest setting.

[millertime359]

if the pump is below the waterline of the res it wouldn't matter if its directly after it or not, it would still be gravity fed. so it cant run dry, unless your res is low on water...

correct me if im wrong.

You have to worry about bleeding though. You put the rad in between the res and pump, and don't get all the air out of it before you power on the pump, you will push that air bubble threw the pump and the rest of the loop before it gets to the res.

[bobo5195]

There is another effect if you want to get geeky if you put the pump after the rad your cooling blocks (CPU/GPU) will slightly warmer water and hence lower performance. It is pretty insignificant but it is there

1 litre per minute of water has a heat capacity of 70 W/k i.e. 70 W will warm it by 1 degree. So a 30W heat dump pump will increase the temp of the water by almost 0.5C.

Depends on how extreme you want to go of course and that is well within the error range of most temp sensors.

Numbers from
Water heat capacity 4180 J/g/K
1 litre of water = 1 kg
1 lps = 1/60 lpm

[justin.kerr]

yes, but who has 1 liter per min flow? 5 liter per min, makes it insignificant.

[Church]

bobo5195: Often increased restriction&lowered flow because of longer/more difficult&sharper bent tubing to implement them nullifies all possible gains of secondary importance advises like "cpu wb after rad", "pump after all wb-s, but before rad" and alikes .. that's why they are of secondary importance unlike main "res before pump"

[rge]

Vapor tested 3 pumps >> rad >> cpu block vs 3 pumps >> cpu block >> rad. The difference was negligible even with 3 pumps 0- 0.2C.

And like Justin said, what pump uses 30W with flow of 1 liter per minute? typically more like 5x that flow and with 18W, ie DDC 3.2, (and not all 18W of that is heat dump in water). Would work out < 0.1C for typical pump/setup, and vapor showed even with 3 pumps (3x heat and much less than 3x flow) barely measurable.

[nikhsub1]

It also effects the amount of pressure the individual blocks get.
Since a Res is empty space it eats up the head pressure that gets sent to it.
When its the end of the loop it doesn't matter how little pressure it gets.

That's it... I give up.

I gave up long ago... but Kurz actually has a point. If you put a res right after a pump it will create unnecessary pressure drop and I think that is what he is trying to say.

[Petra]

If you put a res right after a pump it will create unnecessary pressure drop and I think that is what he is trying to say.

Scott, pressure drop curves have been done for reservoirs and the results were pretty insignificant. My main issue with his comment was with regard to pressure and block performance implications (a pet peeve of mine that I'll be putting together a demonstration of next week).