Radiator stacking

[manny1222]

Not sure whether to post this here or in the case section.

I'm going to do a refresh of my system. I'm looking to use a Corsair Air 240 for the build. My issue is I'll be using a hybrid gtx 1080ti and a H100iV2 and all the reviews and builds I've seen say this is impossible because of clearance with the motherboard/ram.

So I was wondering if can stack the two radiators together on the front, with fans in between (Radiator sanwich). Will this greatly impact performance?

Thanks for your replies

 

[Orddie]

blowing hot air from RAD 1 into RAD 2. does not sound like the best idea to me

 

[Jorona]

Can you? Yes. Should you? Probably not. Unless you're moving the air very quickly through them both, you will get diminished performance out of the second rad. It might still be perfectly acceptable temps though.

 

[Tsumi]

Try it out and see what temps you get under full load. As long as your airflow is high enough, the first radiator shouldn't significantly impact the second radiator, but it'll be noisier for the same level of performance.

 

[LuxTerra]

Yes, you can do it, but you must understand and plan accordingly. Sorry for the long post (please forgive any errors), but I hope this helps.

Water cooling loops operate best at a few C over ambient (e.g. 10C delta is common planning goal). They also operate best with relatively high flow rates (e.g. 1GPM is also a common planning goal). You can tailor all the parameters to your needs. What this means is that the loop heats up more or less uniformly. That is to say, there's not 50C delta water temp after the GPU block and then it cools down to 10C delta in the radiators; these are not like compressor units. Rather, the whole coolant system is more or less at 10C delta steady state with only minor temperature variations around the loop. If this is not true, your loop is broken and likely doesn't have sufficient flow. You probably have excessively high temps (maybe approaching boiling in the water blocks). You can prove this analytically given all the thermal coefficients, etc. You can also empirically test it (use Google). If our PC water loops invalidated this premise, the order of components would start to matter. Since it is valid, pump -> component -> component -> component -> radiator -> radiator -> pump is just fine. If it wasn't you'd want a radiator after every component or even dedicated loops per component.

I think too many people assume the water temps go (assume 25C ambient): 25C rad_out -> 26C pump_out -> 35C waterblock_out -> 25C rad_out. This is valid for some types of cooling loops, but not for standard PC water loops. There is a small delta C around the loop, but it's not a lot. Rather the entire loop basically heats up until it reaches a steady state delta C sufficient to reach equilibrium with the air exchange (via radiator) and then there are only small deltas around the loop.

The job of the radiator is to exchange the heat in the water with the air. Thick radiators *may* do this better because the air has a longer contact time with the fins; as does higher fin counts, etc. However, consider an infinitely thick radiator. At some point it stops exchanging heat at steady state because the air inside it has already thermally equalized to the radiator/water temp. E.g. if the air enters at 25C and the rad/water is at 35C (10C delta), thermal energy is exchanged until the air reaches 35C in this infinitely thick radiator. However, after the air reaches 35C the energy transfer stops no matter how thick the radiator. FYI, the rate of energy transfer is dependent on the delta C and thus the rate slows down as the air approaches 35C. For these simplified reasons, a radiator twice as thick won't perform twice as well.

You can combat this by pushing the air through the radiator faster or allowing a higher delta C. Faster air stays in contact with the radiator for less time. The air exiting a 30mm radiator (illustration only) may not have sufficient time to reach the full delta C of 10C (25C to 35C for example). Thus, while moving more air through all radiators helps their performance, thicker radiators tend to scale better, but all radiators have a limit. At some point it stops scaling because you've hit the thermal coefficient limits of water to copper to air exchange for the area your radiator has. Same reason your die/cores are hotter than the heat spreader which is also hotter than your water temps.

Now you can understand what will happen to your dual radiator setup...

Let's assume you have a 10C delta loop on the first radiator (CPU) and the ambient is 25C. Let's also assume you've done a good job optimizing the radiator/airflow for this loop. The air enters at 25C and should ideally leave fairly close to 35C (let's just say 35C for simplicity). At this point that 35C air will enter the second radiator, but if that loop is also at 10C delta (35C water), there's no heat exchange just like the infinitely thick radiator! This is why stacking radiators in the same loop doesn't work well unless they're already too thin or your airflow is too low; essentially forms an infinitely thick radiator. In separate loops though, your second radiator/loop won't be able to exchange heat at only a 10C delta because the air has already reached the same 10C delta (35C). Thus, the steady state temp will rise until you reach equilibrium (just like the first loop). If you think through the first two paragraphs, this should make sense. Let's assume the two loops are identical to keep things simple. You should expect the second loop to stabilize at roughly 10C delta over the incoming air temp as well. Thus, the ambient air will be 25C, the first loop water temp is 35C, but the second will be 45C. Basically, first loop has a 10C delta and the second has a 20C delta relative to ambient. However, each loop only has a 10C delta relative to the air it's exchanging with (i.e. radiator watts exchanged is only 10C on both loops). Thus, you should expect the component temps in the second loop to be much higher than the first.

This is just an illustrative example, but the underlying concept will carry through. You'll need at a minimum of push/pull fans for a double radiator and probably a third set between them. You need lots of airflow and pressure to push through two thick radiators. Even a single thick radiator performs much better with push/pull fans due to the high airflow resistance. I'd think you'd have better luck running stacked thin/medium radiators vs. thick unless you put powerful fans on it (e.g. Noctua Industrial 3k). Also, put the loop which can tolerate higher temps on the second radiator. This likely means, CPU on first radiator, GPU on second.

Edit: You can see this concept at work in various two pass radiators (front to back, not U), where there's an optimal air flow direction through them. The hot side is behind the cool side; this should tell you that for the designed to airflows, these radiators are approaching maximum usable thickness. In the end, it helps, but it's minimal.

 

[thesmokingman]

It's like throwing money away.

 

[LuxTerra]

It's like throwing money away.

I wouldn't go that far. It can be made to work (particularly with separate loops), but the outcome for PCs usually isn't worth it unless you must have that packaging. For example, automobiles stack different cooling loops and radiators all the time. Obviously, the delta C of the second radiator is higher, but with as long as you understand the compromises on temps and airflow, it can work. I do agree, the better solution for the money is to use a different case and feed both radiators fresh air.

 

[thesmokingman]

I wouldn't go that far. It can be made to work (particularly with separate loops), but the outcome for PCs usually isn't worth it unless you must have that packaging. For example, automobiles stack different cooling loops and radiators all the time. Obviously, the delta C of the second radiator is higher, but with as long as you understand the compromises on temps and airflow, it can work. I do agree, the better solution for the money is to use a different case and feed both radiators fresh air.

Can't compare cars to pc, the applications are vastly different. PC loops run at very low pressures both in the loop and airflow. For computers it is throwing money away. Tests old as dirt by HESmelaugh show this so it should be common knowledge.

 

[LuxTerra]

Can't compare cars to pc, the applications are vastly different. PC loops run at very low pressures both in the loop and airflow. For computers it is throwing money away. Tests old as dirt by HESmelaugh show this so it should be common knowledge.

The laws of thermodynamics don't change based on application.

I only brought the cars up because it's a good example of engineers choosing packaging and other requirements over absolute thermal performance, which was my disclaimer from the previous sentence. The fact that the OP is even asking this, suggests packaging is at a high premium relative to absolute performance. Engineering is a study of compromises, rarely is it one requirement above all else.

Most tests with stacked radiators focused on a single loop, which is of extremely limited value since thick, two pass radiators are essentially optimized to the max and practically are acting as infinitely thick radiators. You'd have to push stupid amounts of air through them to get any benefit when stacked. i.e your radiator air exit temp must not reach equilibrium with the water delta C inside the radiator.

It's a bit more worth it when you switch to independent loops on stacked radiators; it's still sub-optimal, but if the second loop can deal with higher absolute temps you can make it work. Without pushing stupid amounts of air through this system, the OP is unlike to see the second loop perform better than high-end air, possibly worse. However, that may be worth it when you consider other requirements besides just thermal performance. There are lot of variables that can be tweaked to make this setup worth it.