Jump to content
Corsair Community

Need suggestions for first loop


Recommended Posts

Hi all, complete watercooling noob here working on my first loop. I'm building in the Lian-Li O11D Mini and am using two XD3 pumps, pic attached. My first choice would be separate loops for CPU and GPU. However, is it possible to link the XD3 pumps, connecting the bottom (temp sensor) port of one to the top (fill port) of the other? If so, would there be any benefits or just a redundancy thing? Or is it just a terrible idea altogether?



ROG Strix Z270i Mini-ITX

i7-6700K CPU

ROG Strix RTX 3060 Ti

Corsair XR5 360mm x2


Edited by adrian5683
Link to comment
Share on other sites

OK, you have parts so we are somewhat past theoretical.


Combined vs Separate loops - Separate loops can be useful where you have either massive amounts of space or too much distance/components for one single pump. The other possible use is when you have one component that has a very high wattage (GPU) where you do not overly care about the exact temperature and then a lower watt device (CPU) that is temperature critical. Someone running a 3090 who does not care if the GPU temp is 45 vs 50C combined with a 3700X AMD chip that is overclocked as high as it can go and sitting on the CPU temp limit. Since they share the same baseline temperature (coolant temp), the CPU temp is adversely affected by the GPUs heat in the pool. This is more important when one component is close to the user's temperature limit or goal for that device.


The other side of this is our components do not often operate maximally at the same time, except for something like Folding or select professional applications. So if you interested in keeping the lowest possible temperatures on both components, having them share multiple radiators can wind up being less watts per radiator. I'll use the hardware I have on this machine as an example. My 2080 Ti will produce up to 350W in some games. My 10900K@5.2x10 will hold between 80-100W in most of those same games. If each has it's own 360mm radiator, the expected coolant temp rise with fans at 1300 rpm would be +3C on the CPU and around +12C on the GPU. Doesn't sound bad. This results in CPU temps ranging from 45-60C while gaming and GPU temps around 50-55C. However, by combing loops I reduce the wattage per radiator down from the 350W/1 on the GPU to 225W/2. Now the two components share the same coolant delta (about +7C). So the CPU got 4C worse, but the GPU is now 5C better. It seems like an meaningless trade, except I don't care if the CPU is 45, 50, 55, or 60C. It changes frequently and there is little consequence or long term effect (no longevity issues with an average temp of 60 vs 55C). However, if you have a Turing or Ampere GPU, there are consequences at specific GPU temp points in the form a clock drops. You can't run your 2080Ti at 2145 MHz at 55C. It will not allow you and starts dropping -15 Mhz at various points on the other side of 42-45C. Now this is minor and we can have another discussion about where there is any perceptible performance loss at 2130 vs 2100 MHz, but that is the difference. The more obvious difference in using separate loops is the extra cost and noise from the pump.


The above example works because I have a ton of headroom on my CPU and want to prioritize the GPU. The heavier watt component (almost always the GPU) is the beneficiary and the lower watt component takes a small increase in temperature by sharing the pool. The more disparate your wattages (6600K vs 980 Ti SLI), the more separate loops make sense. The more even your components (10900K=250W and 2080Ti 330W officially), the more a combined loop makes sense so that cooling potential is not wasted.


All that said, I would have tried to warn you off if you were still in the planning phase about squeezing two XD3s into a mini. I was having trouble picturing it in my XL. Now that I see it, it's rather interesting and definitely doable. However, there are a couple things you will want to account for.


1) At present you have the bottom radiator as intake and the top as exhaust. In that configuration, the waste heat from the lower radiator becomes the intake air temp for the top radiator. That will increase the temp of the top radiator by the same amount as the bottom, making it little more than a passenger. The only way it works at all is if the the lower watt CPU goes to the bottom radiator and the higher watt GPU goes to the top as exhaust. Yes, the criss-cross of tubes there would be a total nightmare not to mention any OCD feelings about the top component going to the bottom and the bottom to the top.


2) Better plan for separates or combined loop - Turn the bottom fans to exhaust. Now the component to radiator path does not matter and you are expelling heat directly out. That is something much more important in smaller cases. Also, I am assuming the grill on the bottom of the mini is the same as on the XL. For whatever reason, I get really weak results using a 360mm radiator on the bottom as intake in the O11. It would be 40% less effective than the identical top radiator, which should be slightly worse because of the slightly higher temperature in the top of the case. After flipping the bottom fans to exhaust, whatever restriction there was went away and the top and bottom radiators now show the 1C drop per pass I was expecting (prior max was 0.6C per pass on the bottom). Air is passively drawn into the case through the rear and also through the gaps around the XD3s/rear chamber path. This is a viable set-up.


3) XD3 pumps - These are DDC which are smaller, but like anything small and high speed, it will make some serious noise at maximum. The XD3 creates quite a whirl at maximum and you will never need to run that except when trying to fill the loop. At desktop levels, you will want to keep the XD3 below 1700 rpm. I found that to be the detection range for 3-4 ft from ear to pump. So two XD3 at 3500 rpm is going to be quite noticeable, two at 1500 rpm should not. 3500 rpm was the most I could take for gaming with headphones and it was noticeable if I take them off. However, it is possible you could keep the pump speed down at those low levels all the time with 1 pump per radiator/block. While it is possible for flow rate (pump speed) to be too low and affect performance, it is not something where you can crank it way it up and watch temps come down. If you find any radiator review, you will see most no longer even test at various flow rates (0.5 to 1.5 gallons per min) because the differences in temperature are measure in tenths of a degree. You do not want to combine XD3s into a dual loop. This is too much pressure for a 2 radiator, 2 block, short length system. Without some kind of large reservoir or other means of separating them, I would be concerned with the affect one would have on the other. You never connect two pump directly in line with each other. You always need something in between. Besides, with 1 XD3 and those same components, it was like a firehose going through my CPU block at high speed.


You've got my interest and I am already imagining what 2 XD3s would have looked like in my XL. Curious about your CPU and GPU models as that may have some bearing on the decision.

Edited by c-attack
Link to comment
Share on other sites

Thank you for the detailed reply, much appreciated. I've updated the original post with some components. I'll take the advice under consideration and update.


EDIT: So based on my components it seems separate loops is somewhat redundant... How would the order for a combined loop work in this scenario?

Edited by adrian5683
Link to comment
Share on other sites

  • 4 weeks later...

Sorry it took so long to update, the parts and tools I needed to start bending tool forever to get delivered, I'm still waiting on a few extension fittings... It wasn't easy figuring out the runs, the fact that the XD3 pumps are recessed in the case makes it a bit tricky. Based on the original pic I've planned the following run: Bottom XD3 -> GPU Block -> Top XD3 -> CPU Block -> Top Rad -> Bottom Rad -> Bottom XD3.


Looking for suggestions on whether it would work or if I should try something different.

Link to comment
Share on other sites

As it was said, two pumps is a bit overkill and redundant :) I'd remove one, also less tubing to bend.

It's a very simple setup, so you really don't need a dual pump. If you had some quad SLI with radiators everywhere, it would have made sense ^^


Also, since you have an Intel block, you don't have to necesarely install it with inlet and outlet at horizontal. You can flip it sideways if it makes your tube runs easier or more pleasing to the eye.


As for loop order, same story, whatever looks best.

For looks you can also consider a parallel loop, it looks really good with hard tubing.

An example (with the bottom pump only) would be :

pump out - (GPU in passing through to CPU in) - (CPU out passing through to GPU out) - bottom rad - vertical run to top rad - pump in


Waterblocks are fed water in parallel, so the GPU doesn't dump "hot" water to the CPU (hot and cold being very relative.. 1 or 2°C difference), but overall it's pretty much equivalent to a traditional series loop.


Your GPU is a 200W TDP, the CPU isn't going to boil water anytime soon either so you can really go to town with loop order and tubing runs.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Create New...