[discussion] Two loop flawed concept

I see here in the forum lots of people advising dual loop imho its a flawed concept.

I get the idea behind the dual loop but to me it just dose not make sense.
Dual ore triple rad setup on the other hand make's perfect sense to me.

For argument sake we have a setup whit 1 CPU 150W and 2 GPU's whit 2x 150W heat dissipating and we are using 2x 240mm rads.

(And yes you can use different size rads but the argument still holds that you always bound by eider GPU ore CPU depending on what your doing and that part need the cooling the most)

Yes your CPU will be a little cooler whit a dual loop but at the same time your GPUs will be hotter, because your only loading your CPU rad to 50% ware your GPU rad has a 100% load on him.
(and yes this is not totally correct way of using load Nr's, but it keep's the discussion simpler)

So even do you can OC your CPU a little higher your GPU OC will be lower if you have load on both your CPU and GPUs.

If you on the other hand use a dual rad setup you are going to load both rads up to 75% giving you 50% extra cooling on the GPUs and 25% less on your CPU imho that is a big win and a little loss.

Also when i measure the intake and the outlet on my current rad i only have a 1.7 degree Celsius different's in temp (*)
(*)i7 920 @ 3.9GHz and a 5870 @ 920Mhz on a magic cool slimline 240 and Swiftech MCP655 pump
So your CPU temps are not going to spike suddenly.

So if you have a pump > rad > CPU > (pump >) rad > GPU > GPU > pump setup you still have your CPU cooler then your GPU even do CPU is gone run a little warmer your GPUs are gone run a lot cooler because your getting cooler coolant in to your rad.

Also your CPU rad is going to have a higher delta because its warmer and will have a higher temp difference whit the intake air so you that one gets more efficient, on the other hand your GPU rad is doing the opposite, but overall i think you still win a littel.

but overall i think you gone win a bit, on games that are not really CPU/GPU bound and have a good balance, your win is not extreem, but if our a Folder ore encode video, that is eider GPU ore CPU bound you going to win big because you get cooling help from the second rad.

*** games max out the GPU and/ore CPU they don't max them out thermal, ware Folding and video encoding like programs do max out a possessor but only one of them, not both CPU and GPU

So i think in all whit a dual rad you can have a higher OC then whit a dual loop, in all a win win situation.

My 2 cents

u are saying that a single loop with 2 triple is better than a dual loop that seperate 2 triple rad (1in one other in the second)?
Delta T will be lower on each loop. and temp will be better.

The Delta wont really change, but your setup will be more efficient, because your using the spare cooling headroom of the least loaded rad.

Um, two separate loops, although may be an uneven heat distribution with CPU vs 2x GPU's in different loops, will always give you better performance than one big loop with a couple of radiators in it.

I don't know what you've been on or how long you've been watercooling, but don't try to jump in and tell people (like me) that have been doing this for more than 5 years that what we've been doing all along has been totally wrong, because there's a good number of extremely intelligent people in the water cooling communitythat have tested, tried, trued, and proven the methods that people follow today

I`d love to read it but as sneaky says intelligent people and even myself have switched to dual loops and found improvement.

depends on how many blocks `n what order as to the need or not

in fairness now im over radded its one loop again, too many varables to say one or the other is deffinative maybe.

Sneaky said:

Um, two separate loops, although may be an uneven heat distribution with CPU vs 2x GPU's in different loops, will always give you better performance than one big loop with a couple of radiators in it.

I don't know what you've been on or how long you've been watercooling, but don't try to jump in and tell people (like me) that have been doing this for more than 5 years that what we've been doing all along has been totally wrong, because there's a good number of extremely intelligent people in the water cooling communitythat have tested, tried, trued, and proven the methods that people follow today

Click to expand...

I have bin watercooling for 10 years and about 8 years I made my first waterblock that had the same concept as all the new blocks have now, when they ware still using circular blocks. (to bad OC forums delete ore lost the pics)

I immigrated to a other country and had less time to be active in the forums, but never stop using WC and reading the forum from time to time.

Further more i am a piping foreman in the offshore and do there what we do here on a totally other scale.

So i do know a little ware i am talking about, further more cant you read the first word in the topic title "[discussion]", damn people gotten rude here in the OC forums.

It depends on the resistance of the loops. One larger loop might make the pumps work harder than the 2 smaller, resulting in more pump noise.

Also, there is a flaw in the OPs logic...

havor said:

For argument sake we have a setup whit 1 CPU 150W and 2 GPU's whit 2x 150W heat dissipating and we are using 2x 240mm rads.

(And yes you can use different size rads but the argument still holds that you always bound by eider GPU ore CPU depending on what your doing and that part need the cooling the most)

Yes your CPU will be a little cooler whit a dual loop but at the same time your GPUs will be hotter, because your only loading your CPU rad to 50% ware your GPU rad has a 100% load on him.
(and yes this is not totally correct way of using load Nr's, but it keep's the discussion simpler)

So even do you can OC your CPU a little higher your GPU OC will be lower if you have load on both your CPU and GPUs.

If you on the other hand use a dual rad setup you are going to load both rads up to 75% giving you 50% extra cooling on the GPUs and 25% less on your CPU imho that is a big win and a little loss.

Click to expand...

In the above mentioned set up, yes you would get more cooling on the GPUs, but you wouldn't get 25% less cooling on the CPU, you would actually be heating up the CPU.

The water temp normalizes after a period of time because of the specific heat of water, etc. Basically instead of getting 40 c on your CPU and 80 c on the GPUs of a dual loop, you would get 60 c on all components in a large loop.

It isn't correct to think that the parts cooled just after the rad are somehow cooled to a lower temp than the parts just before the rad. Hence the flaw in the OPs thought process. It isnt a straight 150w of dissapation directly connected to the 300w rated rad, you are forgetting about how the heat gets to the rad.

Water can just as easily supply warm water to heat something the same way it can supply cool water to cool something. Its like a water radiator in a house, its not hotter at recieving end vs the return end. That part might warm up first but after it gets up to temp it all runs the same.

But I do agree that it would be better to have 1 medium sized loop vs 2 tiny loops. For example, if you are just doing the CPU and a single GPU, do it in one loop. But if you have multiple GPUs, I would suggest putting the GPUs on one loop and cooling the CPU and motherboard/chipset on another vs one large loop.

I reallly dont think theres a yes or no to this.

I can rig the five rads im using right now and provide situations that promote either.

I could move just one rad in the main loop now and cause the ram to fail on my full cover GPU block because it cools the GPU ram and there threshold is low.

expect I could further switch the order and provide more CPU threshold.

I could thurther touch none of the above and just mess with the fan controller and pump speeds and impact the system.

currently im running pump>pump>CPU>NB>rad>rad>GPU>rad>res

not even the order im intending on keeping but the flow rate and un-banking of rads means my offensive clocking holds up 24/7

the other loops pump>rad>ram>ram>rad>pump>res (because I was bored and had spares #Laughs#

thats dual loop because one loops all alu and steps to 6mm tubing.

im fairly sure I dont want to merge into one loop in my senario

I'm not arguing the case either way because I don't know enough. I run a single loop with the components in my sig and although I'm aware the water temperature normalises after a time I don't believe the GPU and CPU temps normalise in the same way. I can run Prime95 for 24 hours and the GPU temp only rises a couple of degrees at most. I'm assuming localised heat load on the CPU or GPU is transferred to the water and then quickly (relatively) dissipated through the rad without substantially affecting the other components.

havor said:

So even do you can OC your CPU a little higher your GPU OC will be lower if you have load on both your CPU and GPUs.

Click to expand...

I think this is the whole point of having dual loops. Almost everybody will OC the CPU while fewer people will OC the GPU. Given that as a starting point it makes sense for people who are more concerned about the CPU overclock to keep CPU cooling as good as possible.

GPUs also have a much higher tolerance for heat than CPU's. Having a GPU 10°C hotter (for example) than the CPU is not as much of an issue and it's no issue at all if the GPU isn't seeing an extreme overclock.


If the owner is willing to let the CPU overclock suffer a little for GPU speeds then I'd agree it's better to run one massive loop than two loops ...

Who would water cool a gpu without overclocking it? I would like to see a review on how the two different methods compare, using the same components and seeing the impact on cpu / gpu temperatures.

I think if you have room in the case or outside it, two loops would be better, as they would be shorter than one long loop for a start (higher flow rates), and two pumps in series does not double the effectiveness I believe.

The point i trying to make is this that you never load both GPU and CPU to 100% thermal load unless you running both furmark and OCCT like programs at the same time in a real world scenario. ( i cant see the point of doing that)

In all other cases you properly load the processors up to something like 50~75 and 50~75% thermal load during gaming, your system is still stable

But when you running 100% load during furmark ore OCCT on one of them one rad is going to help the other, and even do the the temps of the water going true the other block is high enough to make it fail during full thermal load it no problem because its not generating heat and there for is still under the thermal threshold for failing.

The only real-word scenario ware you could fail is maybe and i say maybe during the physics CPU test on nVidea systems as it offloads some of the calculations to the GPU

But then again i cant see the point of doing that all day unless you going for records, but i see lots of people/WC guru's here on the forum advising new people to WC to get a dual loop setups, and imho specially for those people imho is two loops just a flawed concept

Badbonji said:

Who would water cool a gpu without overclocking it?

Click to expand...

I think the main reason is noise. Since GPUs only have a single fan they can get pretty loud. With water cooling you can use two quieter fans on a dual rad for the same job ...

Badbonji said:

and two pumps in series does not double the effectiveness I believe.

Click to expand...

Your right about that, if you set the pumps in strait after one other. because your building up massif pressure one the first blocks, but if you have the second pump halfway in the loop, it would be the same effect as having dual loops.

I talked a bout it a little more here

im no guru but i`d like to think people are given advice on a case by case basis.

unless its set in stone in a sticky here somewhere

not everyone wants to find a home for a D5 capable of driving long restrictive loops for example and may opt for something more compact and less capable to drive two loops, just as much as some dont want the pleasure of paying for a couple

synthetic tests do only indicate as to how things will hold up. only real world senarios give the bottom line on the balance.

im not so sure what you want from us, present some research that clearly shows its all questionable or enlighten us to some situation were someones gone away clearly miss informed maybe

I totally agree that two loops one for CPU/whatever and one for dual GPUs both having a 2x120mm rad would not be enough for the GPUs

Having the proper amount of rad for the components you are cooling though changes everything.

Someone already touched on the point of dual loops for restriction/flow rate reasons.

But there is another benefit.

If you want to swap cards out or whatever, with dual lop you can kill the GPU loop and leave your CPU loop running.

I swap HW TOO much. I have not watercooled my GPU yet simply for that reason. (I have a FC block for my 4890. ITs been sitting for weeks because I dont feel like breaking down hte loop again).

Just my n00b experience

Dual loops single res, you would still have issues with of course.

on the pump placement I use two together on one loop for convenience and because two D5`s will provide ridiculous flow.
and the other loop half way round because its restrictive and the res/rad provides a unhandy column of back preassure its an open loop.

friend of mine was a civil engineer and in many a back room fishery chat he always said the other should be half way round if possible but closed loop makes that less negligible (my half way round loop isnt closed and full of poison... im going 50/50 with stupid/brave #Laughs#)

As someone who does not OC, and just to throw in a different view, my reasoning for single loop with 2 pumps is in case one breaks. I am part of the FaH team so my CPU is pegged all the time as is my GPU. I don't want failure. Currently I only have a rad and one pump but this weekend I will be upgrading to 2 rads (triple and a double) and 2 pumps. Just waiting on parts.

I agree this is a depends on the setup situation.

Badbonji said:

Who would water cool a gpu without overclocking it?

Click to expand...

People that don't like fan noise.