how to setup 2 pumps
 

Hi All,

I have a simple question, about the setup of pumps.
I have 2 L20 pumps.
1 Dtek pro, 1 Black ice extream, 1 astek radiator
Wite water CPU block, 875 NB block, GPU, HDD.

What is the best way to setup my pumps?

Several options are:
- Let one pump push the water trough the WW, go via one rad (Black ice) and
let the restricted flow being sucked out by the second pump.
This pump will push the water to the NB, GPU and HDD, and rad. (Dtek and Astek)
(I think that the flow on the WW will be good.)

- Let 2 pumps push the water to the WW (outlets of the pumps combined)
From the WW the 2 flows (as it has 2 outlets) rerouted trough the system.
(I think that the flow on the WW will be better.)

- Place the pumps in serial. Let the one pump feed the other,
and then go to the WW, and trough thwe rest of my system
(I don't know if this will perform better)

Please advise

Simple: put everything in series.

everything in series? wouldn't that totally suck compared to putting things in parallel?

Am i reading you wrong? Are you saying it would be better to send the water to the CPU then strait to the GPU then to the north bridge then rad. then res.?

as apposed to having a manifold that split up to flow, or since he has multiple pumps already dedicate one pump to CPU, one to GPU, and one to "other". Wouldn't that be more efficient?

Yes.

Everytime that you split the flow, you reduce the flow rate to each component. The only plus to it is that it allows the pump to provide a little bit more flow, but in the end, it's usually not worth it, compared to a complete serial support.

So you recommend running only one continuous loop? Like running water strait from the CPU to cool the GPU?

Would it be better to put another radiator in between the CPU and GPU? But that would also significantly reduce flow rates.

I'm just concerned that the efficiency would be greatly diminished since delta T isn't as high as it could be, even with the otherwise higher flow rates.

Yep, everything in a single continous loop.

You don't need to add another rad: you don't have the heatload to justify it. (i.e. you'd be wasting your money).

If you did add another rad, you'd set it up in parallel with the other rad, but only for those cases where the flow is so high, that you need the split to get as close to 1.5 gpm through each rad (point of "highest efficiency"): there's also no sense in dealing with the pressure drop of a rad above 1.5 gpm.


The delta T that you refer to otherwise is so small, that you may not even be able to measure it. I'll refer you to an article by a colleague of mine, which includes a formulae to calculate the temp difference, in and out of a waterblock:
http://www.nordichardware.com/articl...vattenkylning/

superart,

It is very unlikely that there would be more than a couple of degrees difference in coolant temperature either side of the rad, provided flow rates are adequate.

If you want to have separate loops, then don't forget you will need a much larger pump in order to get significatn flow rates through each component, and this larger pump will add a lot more heat to the system.

8-ball

What If you run a setup like this?

Here your only losing flowrate to the radiator, which will not negatively effect the performance of the entire system.

Each pump draws cool water from the res. independently, thus giving the same effect as just having one pump.

If you think about it, you wouldn't be losing much flowrate even at the radiator, since there are 2 pumps feeding the manifold, the flow rate to the rad. will be approximately the same as through the blocks.

Also, the flowrate through the GPU will actually be higher than it would have been if using one pump and putting it in series, since the water is not losing flow by going through the CPU block first.



What do you guys think of this type of setup? It gives you the best of both worlds. High flow rate and better efficiency since the CPU isn't heating up the water that would be used for the GPU

Superart, I'm not sure you are listening.

You will have lower flow through the cpu block and the gpu block in the setup you have described than if the whole lot were plumbed in series.

Also, please take note of the point I made about the water not really heating up as it goes round the loop. If it's flowing relatively quickly, it will only be heated by around 1 degree by going through the blocks.

Let me put another argument to you.

Most of the pumps used in watercooling are not really ideally suited to the job.

An ideal pump for watercooling, would have a lower rated max flow rate than the eheim 1250 for example, but would be capable of producing a much greater pressure.

Since we are operating at the high pressure, low flow rate end of the operating range of most pumps, the L20 included, we would gain significant benefits in flow by increasing the pressure produced by the pump rather than the freeflow rating.

Pumps in series increase the overall pressure produced by the pumps, while pumps in parallel increase the freeflow rating (pretty useless).

I know it would seem that separate loops would be the best way to go, but this question has been asked MANY times before and serial is the way to go, particularly through the cpu block. Some benefits can be gained by running the gpu and northbridge blocks in parallel, but only after all of the water has been through the cpu block.

I would wager that the flow rate through the cpu block in the diagram you have drawn would be lower than the flow rate for the whole system in series with just one of the pumps.

And what happens if one of the pumps fails?

Some of the water will now flow backwards throught the other pump. Is this bad? Yes, because not all of the water will be going through the rad.

What about in series?

If one of the pumps fails with them in series, it doesn't really matter, you will still have pretty adequate cooling for as long as it takes to replace the failed pump.

Does that make sense?

8-ball

How about this setup? The pumps are in serial, but the water from the CPU is cooled before sending it to the NB/GPU and then cooled again before going to the CPU. Maybe you could leave out that last RAD, since the GPU/NB isn't going to add as much as a CPU would, but anyways, that's what my system looks like, just thought I'd share my 0.02.


Black is pump, brownish coler is waterblock and grey is RAD, sorry

I'd reccomend going Res>Pump>Pump>Rad>CPU>GPU/NB(parallel)>Res.

A res works best just before a pump as it allows the pump a relatively unrestricted source of water. The 2 pumps in series will give you the best flow overall flowrate. Place the radiator after the pumps to dissapate the heat the 2 pumps will add. I'd go with the GPU/NB in parrallel simply because neither really needs much cooling. Theres no point in cooling the water a second time inbetween the CPU and NB/GPU as the difference in coolant temp is minimal and both the GPU and NB can handle the slightly higher temp.

8-ball, yea, I think I get your point now.

waterblocks setup
 

Hi All,

Thanks for the replies!
So, the pumps should go in series.

How about the water blocks? series or paralel?
If I would setup:
water pump, split in 2, radiators, combine 2, water pump, reservoir, from reservoir a 4 split (so in the reservoir 1 hole in, 4 holes out)
go simultanesly to the CPU, North bridge, GPU and Hard disks.
(flow gets devided to 50, 15, 25, and 10%, so all items get the flow they need compared to the heat they produce)

then combined to one again (reservoir, 4 holes in, one out) and to the water pump.

on this way I should get the best flow. Right?
Will the extra flow (should go up by 30% compared to a setup where all is in one row) compensate for the loss of pressure?

some more....
 

just to get things clear for me.
we are talking about a closed water loop. my sciense teacher used to tell me that pressure is the same in all directions, that's why a boat can float. Right?

every pump has a maximum pressure it can produce. the 1250 a bit more then a L20 (head for the 1250 2.5 meters, for the L20 2 meters)

If you would restrict (or under presure) the water inlet of the pump, the pressure would drop. If you would place pressure on the water inlet, the pressure build up would increase. That's why the pumps in serial should perform better.

Would there be any diference in pressure if you would squeeze the water through a 8 mm hole or trought a 20 mm hole?

In an open system, sure. In a closed loop I'm not sure.

Sinse there goes less water through the 8 mm hole, the pump also gets less water, and therefore should perform less. The pressure biuldup will be faster, that's for sure.

With a 20 mm hole, it can move almost all of it's pressure. A cosequesnse will be that the water in a closed loop, entering the inlet of the pump would also be in a pressure state, or at least less restrictend. so, the pump would be able to give more pressure.
The presure build up would take a bit longer, but should end up equal or more to the pump who squeezes the water through the 8 mm hole.

This is only about the pressure build.

Next part is that we will be losing pressure(?) since we are using hoses, radiators and water blocks. The friction will course a pressure drop.
This will be more if we would place all blocks in a serial loop. (according to some posts, you will lose 85% flow, compared to the pumps specifications. a 1200L pump will give only 150L of flow in a water cooling setup)
If we would make a paralel loop, the friction will be less.

So the question here is:
Where is the break point?

The White Water has one inlet, and 2 outlets, giving less pressure in the waterblock, but a better flow.

If only pressure would be the key, I should turn my setup arround, make the outlets an inlet, and place my two pumps there.
pump-pump-split-2 inlets-1 outlet.
This would give the most pressure on the inner side of the WW.

Please help me out here!
:confused: