Dual-pump parallel loop setups - has anyone seen one?

[Duff-Man]

Sorry for another random watercooling-question thread, but anyway: I'm wondering if anyone has seen a setup like I describe? It seems like a fairly obvious thing to try, but I haven't come across any real-world examples yet.


Okay, suppose I have two pumps, a few radiators, and a bunch of components to cool (lets say a CPU and two GPUs for arguments sake). The traditional logic says I would set up either:

a) A dual loop setup, with the CPU (say) connected to one pump and a some radiators, and the GPUs connected to the other pump and the remaining radiators. Each would have their own reservoir.

b) A single loop in serial, using the second pump to boost the system-wide flowrate by increasing head-pressure.


To my mind the more efficient way would be to set up one loop for the components (no radiators), and a separate loop circulating water only around the radiators. Both loops would feed into a common reservoir, and the water would be mixed thoroughly before it feeds back out. In essence, you have one loop to transport heat from the components to the reservoir, and a separate loop to keep the water in the reservoir cool.


My thoughts on this are that, due to radiators being fairly low restriction, a good flow rate could be maintained through the radiators even with a fairly weak pump - after all it's only pumping water between the reservoir and the radiators.

As for the other loop - the one pumping water over the components - well we don't really mind if the flow rate drops a little do we? Even at a fairly low flow rate (say 0.5GPM) the water temperature will rise by only a degree or two as it does a single pass through the system. The water it draws in will be constantly cooled by the other loop, so the intake water temperature will always be good...


Flow rate is far more important through radiators than it is through the components, so why not separate the two? Surely this would maximize the cooling capacity of the radiators? I understand that by doing this you are required to use the same coolant in both loops and so you lose the ability to match colours etc, but from a performance point of view it seems like the way to go.

Has anyone ever seen a system like this? Or even better, a performance comparison with an equivalent traditional setup?




EDIT: Added picture to explain better what I'm rambling about:

 

[Danm54]

It'd work for sure, you'd need a big res though to see decent temps.

In effect your restricting your cooling to the res only and the water from your CPU and GPU loop will pass through this very quickly so will hardly be cooled.

You'd need to opposite of what you say, a very high flow through the rads and a lower or equal flow through the blocks so the heated water from the blocks is drawn out of the res and replaced by cooler water to then be drawn through the blocks again.

Not really worth it and I dont think you'd see equal or improved temps over the traditional single loop.

You could run 2 loops off the same res and they'd stay the same temp, not one loop hotter than the other. Only reason for that is if a single pump wasn't up to the job of running a single loop, the D5 will eat most loops so not really an issue!

 

[Duff-Man]

Well the system reaches equilibrium fairly quickly, so it's just a case of removing heat from the system. So long as the water is well mixed that shouldn't be an issue. You can even set up a system to prioritise the warmer water, eg:




Remember, the difference between the warmest and coolest water in the loop is only likely to be a couple of degrees. What drives the cooling is the difference in temperature between the ambient air and the water passing through the radiator. That will be maintained, so long as the water drawn in is at equilibrium temperature, and since the flow rate is high through the radiators, their cooling performance will be maximised.


You do have a point about the pump though... Above about 1GPM there are significantly diminishing returns on cooling performance from the radiators, so it would only really be of benefit if the full system couldn't maintain that kind of flow rate.

 

[Danm54]

Well the system reaches equilibrium fairly quickly, so it's just a case of removing heat from the system. So long as the water is well mixed that shouldn't be an issue.

True, as long as you dont have a tiny res like a 150 multioption was all I was saying, you'd be recirculating too much of the same water around the blocks else.

Not sure how big a res you'd need though, trial and error I guess!

 

[lrlcr]

You can do this with the danger den monsoon d5 premium and only need one pump.

 

[Biffa]

I did something similar (not exactly) with weebeastie.

Its a 970X o/c + 4 x GTX470's with cpu/gpu's/mobo all water cooled using the orignal XSPC Dual DDC pump 5.25 res. (The one with a single res)

But I didn't split it the way you say, but if I ever pull it apart I might be tempted

I split it so one pump serviced the CPU/mobo and the other the gpu's.

 

[PhillyDee]

You basically want to use the reservoir as a type of heat exchanger. As long as the coolant is getting mixed, then it should work, but you could end up with thermal flows and a reduction ion cooling performance. What you need to do is try both!

 

[Duff-Man]

That's a very interesting setup Biffa - I hadn't seen your thread before! I'll have a closer look now. The principle looks pretty much the same though (assuming that's a single reservoir that they share?). As you say, the main difference is that your loop isn't entirely split.

The justification for splitting it entirely would be to maximise the flow-rate through the radiators; although with four GPUs, a CPU and a chipset block running a single loop could restrict flow rate to the point where the water would begin warming significantly (if you're dumping close to 1000W into the loop and your flow rate drops to 0.5GPM, you'd be looking at 7-8C temperature rise from a single pass over the components(!) ). So I suppose it's not so cut and dry either way...

You basically want to use the reservoir as a type of heat exchanger. As long as the coolant is getting mixed, then it should work, but you could end up with thermal flows and a reduction ion cooling performance. What you need to do is try both!

Yep The one loop acts as a heat-transport mechanism [transporting heat from components to res], while the other loop concentrates on removing the heat from the water. So, separate, dedicated functions for each loop.

Ensuring reasonable mixing wouldn't be too difficult - IF you were designing a custom reservoir, but it might require some modification with a standard dual-bay res. I suppose ideally you'd want to put a couple of pieces of acrylic with small holes drilled in them to separate the "warm" and "cool" areas of the res (see above). This would ensure that the radiators are primarily fed with water from the component exit, and gradual flow between the two sections would prevent cool fluid from recirculating back into the radiators.

So long as water isn't stagnant, it mixes pretty rapidly at these length scales. But as you say, you need to remove the possibility of re-circulation of cool water back into the reservoir.


EDIT: something a bit like this:

 

[rjk]

I had two pumps in a loop once. Changed to a d5 and didn't notice a blind bit of difference.

My loop order used to be like this:
Res
Pump
480rad
240rad
Pump
4890 block
4890 block
CPU
Res

Both pumps were ddc 10w that I modded to 18w.
My flow rate was excellent but the difference between two pumps and the single d5 is minimal.

 

[Duff-Man]

I had two pumps in a loop once. Changed to a d5 and didn't notice a blind bit of difference.

My loop order used to be like this:
Res
Pump
480rad
240rad
Pump
4890 block
4890 block
CPU
Res

Both pumps were ddc 10w that I modded to 18w.
My flow rate was excellent but the difference between two pumps and the single d5 is minimal.

The key difference would have been if you want back to the res after the 240 rad. This way the restriction on the radiator-only loop would have been minimal, and flow rate would remain high in that loop, no matter what restriction you find in the components. ... but it's still true, with a sufficiently powerful pump it shouldn't make much difference, due to diminishing returns with high flowrate.



Perhaps it's better to look at this idea in terms of specialisation of equipment. If I know that I have two loops, one of which I have to overcome a high restriction (the component loop), and the other where I don't (the radiator-loop), then I can choose my pumps accordingly. Perhaps a D5 for the components, and a smaller pump for the rads?

Similarly it could allow better selection of heatkillers - for example since your component loop doesn't affect your cooling potential (since it doesn't affect the flowrate through the radiators) you might be more inclined to choose more restrictive and better performing blocks.

... just thinking out-loud here really!

 

[Biffa]

That's a very interesting setup Biffa - I hadn't seen your thread before! I'll have a closer look now. The principle looks pretty much the same though (assuming that's a single reservoir that they share?). As you say, the main difference is that your loop isn't entirely split.

Single res, not split. This was before they made the split version. Mines even further removed really as you can see the return from the cpu/mb side goes back to the gpu side of the res and vice versa

 

[Sayso]

I used a dual pump single resevoir XSPC EOL bay for a bit basically running dual loops with a single resevoir, worked well but i did have a 480 and a thin 360 cooling CPU, Chipset and single card.

Before that i just run the lot in series with a dual pump top, the temps were about the same, never checked flow rate though.

 

[lrlcr]

What you've posted duff-man is essentially one loop with two reservoirs, that happen to share a common wall.

 

[Tefal]

i suppose if you had a large vertical tubular ress, had the inputs coming in at the top and angled slightly you'd get a very turbulent flow of water down the reservoir making sure they were mixed incredibly well and the temp would be homogenised pretty much fully, then both outlets at the bottom.

 

[Duff-Man]

What you've posted duff-man is essentially one loop with two reservoirs, that happen to share a common wall.

The principle is to separate off the radiator loop from the components, so that high flow-rate can be maintained through the (low-restriction) radiator loop. The reservoir design is only to facilitate mixing - or in that case, to prioritise the feeding of "warm" water into the rads.

The "common wall" in the reservoir is only there to stop recirculation of hot water through the components (i.e. to prevent streamlines that exit from the components, and then re-enter the component loop without passing through the radiators). So yes, it's like a double res that shares a common wall, but the common wall HAS to allow the fluid to pass through, otherwise the left-hand portion of the res would quickly run dry.

i suppose if you had a large vertical tubular ress, had the inputs coming in at the top and angled slightly you'd get a very turbulent flow of water down the reservoir making sure they were mixed incredibly well and the temp would be homogenised pretty much fully, then both outlets at the bottom.

That would work nicely

 

[craarc]

my loop with an internal res being cooled by airflow prior to cooling from the rad in an external loop to the pump header

 

[andybird123]

so you are deliberately dumping heat from your watercooling loop inside your case when you already have an external rad?

doesn't that defeat half the point of having an external rad?

@the op, I looked in to doing something like this, but to be honest having something like 2 EK DCP 4.0 18W 800LPH pumps in series is just a helluva lot simpler and will keep the flow rate up in the whole loop, it doesn't matter where in the series you have the pumps either, stacking them right after each other still basically doubles the pressure and flow

a dual pump series loop also has the advantage that if one of the pumps fails you will still have flow and cooling giving the system time to shutdown safely

a single D5 has something like 3.5-4m head (depending on top), but dual EK DCP's should be the same cost but over 7m head

 

[craarc]

yes it will dump a small amount into the case granted because the airflow thru the case is taking some heat from the resevoir but not that much tbh. most is done externally. but if the rad was inside the case i would also get some heat dump from that through the sides of the rad. so i am not getting any of that. i wanted the main resevoir inside the case as is donne in most peoples system anyway everyone withan internal resevoir is getting the same effect. i am minimising my internal heat transfer in a way by having rad and pump externally and tbh when running my i7 950 at 4.2ghz my water temp and cpu temps hover around 27-31 degrees which i am happy with case temp is only a couple degress below those

im sure if i had the rad and the pump internally my heat dump inside the case would be greater