Water cooling loop design for SLI 1080Ti's

[GEEKCPA]

I'm building a new PC with the R6E MB as the core and 2 Strix 1080Ti video cards. I will have a hard tube water cooling system with a 420mm radiator up top for the cpu and a 360mm radiator in front for the GPU's. A single D5 pump and 300ml reservoir. My question concerns the routing of the water tubes between the two Phanteks gpu water blocks. Phanteks suggests it is best for the output from card #1 to go to the input of card #2, which makes sense to me, but such a design requires a complex bending of the output tube from card #1. I have seen loops that use a straight tube from card #1 output to card #2 output, which means the coolant would be flowing the opposite direction in card #2. If the coolant flow is going "backwards" is the cooling in card #2 less efficient? IOW, are the coolant paths directional? Some might suggest a coolant SLI block, which I would rather avoid because the block would block the view of the beautiful MB underneath--even if the block is plexiglass.

Is the best solution to take the extra time to bend the output tube to connect to the input tube of card #2?

[Georgia]

Hi GEEKCPA
I run 2 Strix OC 1080ti's with the Phanteks blocks and the R6E MB, only difference is i run 2x480 raditors and 2xD5 pumps just for my GPU's. I have tried my loops both ways,
in serial and parallel and the temps and flow rates are identical,so at the moment im staying with the parallel setup. (eg using 2 straight pipes to connect the GPU'S).
Hope this helps.

[GEEKCPA]

When you say "two straight pipes" do you mean one in to the intake port on top, but plugged on the bottom, and the top side of the outlet port plugged, with the bottom outlet pipe going straight in to the second card's outlet port with a plug on the bottom. The inlet port of card 2 is plugged on top with a pipe coming out the bottom and voing to the pump or radiator? The instructions that came with the blocks are confusing. They imply the intake port is the one closest to the front of the video card, but the directional arrows appear to be backwards. Am I getting something wrong?

[Georgia]

GEEKCPA wrote:
When you say "two straight pipes" do you mean one in to the intake port on top, but plugged on the bottom, and the top side of the outlet port plugged, with the bottom outlet pipe going straight in to the second card's outlet port with a plug on the bottom. The inlet port of card 2 is plugged on top with a pipe coming out the bottom and voing to the pump or radiator? The instructions that came with the blocks are confusing. They imply the intake port is the one closest to the front of the video card, but the directional arrows appear to be backwards. Am I getting something wrong?

Like This, input is on the left side of the GPU's,output on the right.

[GEEKCPA]

Georgia wrote:
Like This, input is on the left side of the GPU's,output on the right.

That is a parallel setup right? As with anything you read on the internet, there are various opinions as to which setup is superior. Generally, the consensus seems to be parallel has less flow restriction, but the jury is out as to whether that makes a significant, or any, difference in temps. Since I'm into this build around $6K so far I may have to curb my spending. It seems the parallel setup is easier to make, has fewer bends or angles, so that's the route I'm going to go. Thanks for your help. Thanks to all.

[JustinThyme]

I dont use phanteks but thats of no consequence. Two things I can tell you with absolute certainty. Putting them in serial, aka out put of one to the input of the other, will 100% of the time reduce your flow rate. parallel they both get the same flow with less restriction on your loop. If you are doing it the hard way without an SLI block (these are specific to each manufacturer) then two pipes between the cards. Ins and outs doesn't matter top or bottom just as long as you have the inlet on the correct side and the outlet correct. For my blocks (this also varies by manufacturer) The inlet is on the left and outlet on the right. The other thing that people stress over is order of componets. It doesn't matter! What you want is the least restriction possible which also means less pipe and the least amount of elbows and fittings you can get by with. Take time to think it out and plan as best you can to minimize the runs. Im running 2 480 rads with two D5s in serial. With my loop and the way its done Im getting just shy of 300L/hr at full speed which is more than enough, Im normally running the pumps at about 70% as this is where they are the most quiet. Heres my rig plimbed up for an example. Again my inlet of the GPUs is on the left and my CPU inlet is on the right. Single loop. Res>pump(this is the most important)>flowmeter>CPU>top rad>GPUs>bottom rad>Res. I am using an SLI block at its easier and most importantly it adds rigidity to the installation. The cards are not sagging and in the very solid! One more thing, the size of the tubing doesnt make a hill of beans on performance. Bottom line, every connection point is G 1/4, this defines your flow. So long as your tubing is larger you are good, the 12mm OD is plenty big enough. Anything larger is simply a matter of taste. Some must have the 16mm OD because they think it looks better. Thing is its harder to bend, especially the tight corners. You are always better with longer radius too.

[MikeAdu]

Hi.
I am not sure but think that serial connection will make good water flow through VGA waterblock instead of parallel setup.
I have two 780 EK waterblocks in serial(VGA1 out->VGA2 in) so water flow would go as waterblocks were designed for.In my opinion parallel setup will not provide good water flow.


Sent from my iPhone using Tapatalk

[Silent_Scone]

The difference in temps will be negligible, if any at all. Correct application of TIM, and plenty of radiator space will do a lot more for GPU temperatures. If it makes routing easier, then it's a no brainier regardless, though.

[Skunkfoo]

For a given pump speed (actual psi of loop):

Parallel GPU setup has less flow restriction as your pump does not have to work as hard. This does not mean more (faster) flow over your GPU block. It actually means less flow in most cases.

Serial GPU setup has more flow restriction but the water speed flowing over your GPU block will be moving faster in most cases.

In my build, like Georgia's build I setup in parallel. I too have Phantek blocks and in the manual for their blocks they have designated input and output ports and I didn't want the output of one block going in the output of the other block.

And yes, G1/4 fittings everywhere are the real limiting factor to total system flow. If your GPU block has less flow restriction then a G1/4 fitting then theoretically parallel will give your system more TOTAL flow.

Parallel GPU blocks don't break the laws of fluid dynamics and somehow turbo boost flow. For example, if your pump moves 300L per hour measured at the pump outlet then it is impossible for a 4 way SLI parallel GPU section of your loop to move more than 75L per hour through each GPU block. Parallel is exactly that, it splits the flow reducing restriction but the flow just before the split would be 300L per hour and after the split should equal 300L per hour.