Looking for [relatively] detailed calculators for multiple radiator performance

[Thunderdolt]

With Titan Rig doing a 20% off sale, I think it's finally time for me to finish designing and building my rackmount external water cooling setup. To that end, I was hoping someone could point me towards a calculator or xtremerigs-quality review to help guide my design.

Is there a decent calculator that would help me get an idea of performance when running multiple radiators where:

• Both air and coolant are in parallel
• Both air and coolant are in series
• The air is in series and the coolant is in parallel
• The air is in parallel and the coolant is in series

Obviously, nothing would be very precise and any setup would require testing to confirm performance. I'm really just looking for gross calculation of where each combination falls apart in terms of performance and maybe some gross estimate (+/- 30%) of acheivable cooling performance. I know LTT did a video on the topic once, but, uh... it's was LTT - I don't have enough grains of salt to take that seriously.

I'm aiming to run the coolant quite a bit hotter (40C) than Xtremerigs did for their radiator testing (30-31C), so I don't know that simple linear extrapolation from their data would prove to be useful.

 

[Zarathustra[H]]

Sorry, don't know of any good calculators. Most of what I have had to do is trial and error, and report the data myself.

Thanks for the heads up about the 20% off sale. Unfortunately it is not on any brands I am interested in at this point (I have been thinking about picking up a couple of Watercool MORA radiators for a while...)

 

[Tsumi]

Definitely no calculator for that kind of stuff, but you can use physics principals to get an estimate in ranking.

Air and coolant in parallel- the highest performance setup. The primary benefit is higher flow rates due to lower effective restriction. This assumes you don't have an air intake restriction.

Air parallel and coolant series- second highest performance setup. Every radiator is getting fresh air. Once again assuming no air intake restriction.

Air series and coolant series- not the most ideal setup due to one radiator getting the exhaust of the other, but can be countered somewhat by ensuring a counterflow setup- coolant goes from exhaust radiator to intake radiator to ensure it is always warmer than incoming air. Also mitigated by high airflow. Arguably the best setup for cases with top and bottom radiators and limited/no front and side air intake.

Air series coolant parallel- the worst performing setup due to the lack of use of the counterflow principles. Don't bother using unless restriction is actually an issue. I actually have this setup currently but it is due to my reservoir not supporting series flow. I mitigated this to a degree by routing the GPU loop to the top radiator and CPU loop to the bottom. I will eventually replace the reservoir for a series setup.

 

[cdabc123]

If you plan on running the coolant hotter, you want the rads originated so the hot air can freely flow through them. Ideally at the top of the case with just push fans. Sequential rads will drop the coolant temp more. The hotter one should be the one that air can flow upwards through.

Concerning the loop, parallel works pretty well.

I personally like to design loops to passively cool as best as they can then set the fans super low. As honestly in most cases temp doesnt matter as it's always low enough.

 

[Thunderdolt]

Definitely no calculator for that kind of stuff, but you can use physics principals to get an estimate in ranking.

Air and coolant in parallel- the highest performance setup. The primary benefit is higher flow rates due to lower effective restriction. This assumes you don't have an air intake restriction.

Air parallel and coolant series- second highest performance setup. Every radiator is getting fresh air. Once again assuming no air intake restriction.

Air series and coolant series- not the most ideal setup due to one radiator getting the exhaust of the other, but can be countered somewhat by ensuring a counterflow setup- coolant goes from exhaust radiator to intake radiator to ensure it is always warmer than incoming air. Also mitigated by high airflow. Arguably the best setup for cases with top and bottom radiators and limited/no front and side air intake.

Air series coolant parallel- the worst performing setup due to the lack of use of the counterflow principles. Don't bother using unless restriction is actually an issue. I actually have this setup currently but it is due to my reservoir not supporting series flow. I mitigated this to a degree by routing the GPU loop to the top radiator and CPU loop to the bottom. I will eventually replace the reservoir for a series setup.

Interesting idea with the counter flow. That makes sense. Perhaps a hybrid setup with parallel sets of series radiators (ie, 3 air-parallel branches with 3 coolant-parallel radiators in each branch).

One thing about the different flow arrangements is that the air flow is largely fixed in the design, but all combinations of coolant flow can be tested out once the cooler is built simply by swapping connections around.

I guess one other bit is that I'm planning to run cooled air through the radiators. I guess this will allow me to run more rads in series before the efficiency goes to zero. For 6-8mos of the year, the air will be coming from an air conditioner that's in the window. For the other months, the air will simply be ducted in from outside via a similar arrangement (hellooo sub-0F January overclocks!). For the purpose of my early size estimation, I'm assuming 20C air. This would give a 20deg delta T. Based on the numbers from ExtremeRigs and extrapolating for the different rad size and air temp, this should put the first radiator at around 800W capacity. If the air rose to 30C after exiting the first radiator, that would leave the second one at around 400W. I think I could live with that.

In reality, the air should be appreciably cooler than that, so the numbers should be much better than the first pass guess.

 

[Tsumi]

Interesting idea with the counter flow. That makes sense. Perhaps a hybrid setup with parallel sets of series radiators (ie, 3 air-parallel branches with 3 coolant-parallel radiators in each branch).

One thing about the different flow arrangements is that the air flow is largely fixed in the design, but all combinations of coolant flow can be tested out once the cooler is built simply by swapping connections around.

I guess one other bit is that I'm planning to run cooled air through the radiators. I guess this will allow me to run more rads in series before the efficiency goes to zero. For 6-8mos of the year, the air will be coming from an air conditioner that's in the window. For the other months, the air will simply be ducted in from outside via a similar arrangement (hellooo sub-0F January overclocks!). For the purpose of my early size estimation, I'm assuming 20C air. This would give a 20deg delta T. Based on the numbers from ExtremeRigs and extrapolating for the different rad size and air temp, this should put the first radiator at around 800W capacity. If the air rose to 30C after exiting the first radiator, that would leave the second one at around 400W. I think I could live with that.

In reality, the air should be appreciably cooler than that, so the numbers should be much better than the first pass guess.

Cooled air or not, the counterflow setup will always be more efficient when running air in series, it's just physics. How much more efficient is dependent on both airflow rates and coolant flow rates. The higher the flows are (whether air, coolant, or both), the lower the efficiency penalty when not using counterflow.

If you look at the HWLabs GTX radiators, they're designed with counterflow in mind. Unlike most radiators where the coolant goes down one half the width of the radiator, makes a U at the end, and returns at the other side, GTX radiators split the flow along the thickness. GTX radiators therefore have an optimal port configuration depending on the direction of airflow.

 

[Thunderdolt]

I got distracted and fired off that reply without a proof read.

It makes sense that counterflow would work within a certain operational envelope. With any arrangement involving series airflow, there is a point where the air becomes too warm to provide meaningful cooling even in a counterflow setup. Air that begins extra cold should be able to pass through more radiators before it is heated to the point where it is no longer effective.

Given this, it may make sense for me to design for two air-parallel branches of 3 radiators each rather than one long chain of 6 radiators where the air is presented in series. The individual branches could then be plumbed for counterflow coolant.

 

[Tsumi]

I got distracted and fired off that reply without a proof read.

It makes sense that counterflow would work within a certain operational envelope. With any arrangement involving series airflow, there is a point where the air becomes too warm to provide meaningful even in a counterflow setup. Air that begins extra cold should be able to pass through more radiators before it is heated to the point where it is no longer effective.

Given this, it may make sense for me to design for two air-parallel branches of 3 radiators each rather than one long chain of 6 radiators where the air is presented in series. The individual branches could then be plumbed for counterflow coolant.

This is almost never true for a properly set up counterflow system. The only scenario where that can possibly be true is where your flow rates are so low that passive temperature diffusion takes over (virtually no flow). See the diagram below:

 

[Thunderdolt]

Looking at the data from ExtremeRigs, I don't think that assertion is true. As one example, this is what they measured with the EK CoolStream XE360:

Looking at the air temp rise and air-water temp deltas, adding a third radiator in series would provide almost zero additional cooling even when plumbed for counterflow. By the time the air reached the third radiator, there would be almost no temperature differential between the hot coolant and the hot air.

It is correct that additional airflow would overcome this. That said, I would hardly consider 3x120mm fans at 1850rpm do be "virtually no flow" like you state.

Regardless, I'm looking forward to trying it out both ways to see how big of a difference it makes in my setup.

 

[Tsumi]

Looking at the data from ExtremeRigs, I don't think that assertion is true. As one example, this is what they measured with the EK CoolStream XE360:
View attachment 609865

Looking at the air temp rise and air-water temp deltas, adding a third radiator in series would provide almost zero additional cooling even when plumbed for counterflow. By the time the air reached the third radiator, there would be almost no temperature differential between the hot coolant and the hot air.

It is correct that additional airflow would overcome this. That said, I would hardly consider 3x120mm fans at 1850rpm do be "virtually no flow" like you state.

Regardless, I'm looking forward to trying it out both ways to see how big of a difference it makes in my setup.

You are misinterpreting the data. The correct way to intepret that is as airflow increases, the the coolant outlet approaches ambient air temp, which is to be expected. When you have a counterflow setup, the inlet on the intake radiator will be lower due to the cooling provided by the exhaust radiator while simultaneously the air between radiators is warmer than ambient, so the exhaust radiator physically can't cool down to ambient.

Of course the benefits of adding an additional radiator in airflow series decreases as airflow increases, but this is true for airflow parallel as well. Theoretically more radiator area is always better, but practically whether or not it adds a meaningful benefit is a different question entirely.

 

[NightReaver]

What is this cooling? Any specific reason for wanting to run the coolant that high, or just for slower fans/ less rad?

 

[Thunderdolt]

Right now, it will be cooling just a pair of render boxes. I'd like to expand its use to all of my builds going forward. I figure that if I start building in rack mount cases, then I can efficiently stack my hardware as it ages and keep it online for different distributed on-demand compute projects. I can be much more space efficient with this by cooling them with a centralized radiator bank while also getting the benefit of reduced fan noise.

I keep the coolant at that temp because it is far more efficient, runs the fans much more slowly, and has no impact on system stability or performance. I'd run it cooler if I were doing heavy overclocking.

Having less rad is also a consideration due to the space limitation from having to fit in a standard 19" rack.

 

[xDiVolatilX]

What are you guys meaning when you say in parallel? In series? Branches? Counter series? Counterflow? Series setup? Counterflow principles? Passive temperature diffusion?
I've been water-cooling my whole life but I've never understood the details about the wording used here. Please enlighten my vocabulary lol

 

[Thunderdolt]

What is this cooling? Any specific reason for wanting to run the coolant that high, or just for slower fans/ less rad?

FWIW, here's a point of reference:
On the build linked in my sig, I've had renders (Octane - a GPU renderer) running for the last ~40hrs with 30min frames followed by 90-120sec of downtime to load the next one. GPUs are 3090 Strix OC at original settings and they average ~350W TDP during the main render with jumps to 380W during denoising. Coolant is routed from the pump (with temp sensor) to two rads in series, to a HF Next, to the CPU, to two GPUs in series, and then back to the pump.

I'm seeing 34.5C coolant entering the CPU and 40.0C coolant returning to the pump. HF Next is showing 110 L/h flow rate. GPU hot spot temps are steady at 51C on #1 and 55C on #2.

Rereading it now, I realize that I wasn't clear about where the 40C is measured. It is at the end of the heating section, so that's the peak temp in the loop after it has gone through the components rather than before it has been heated by them.

The loop is controlled by an Aquaero using a setpoint controller with a target temp of 40C at the pump.

 

[NightReaver]

I keep the coolant at that temp because it is far more efficient

This I don't get. How is running hotter water over hot pc parts more efficient? Remember, you can always add more radiator/fan. You cannot run more waterblock. You should be maximizing the heat transfer between the waterblocks and liquid by making the coolant as cool as possible.

 

[cdabc123]

This I don't get. How is running hotter water over hot pc parts more efficient? Remember, you can always add more radiator/fan. You cannot run more waterblock. You should be maximizing the heat transfer between the waterblocks and liquid by making the coolant as cool as possible.

You get more passive dissipation out of upward facing rads and can go with slower/quieter fans.

With the exception of extreme overclocks a cpu at 40c vs 60c will make no differance. Some of my favorite loops used almost no fan and very little coolant flow allowing for a near silent build.

 

[NightReaver]

You get more passive dissipation out of upward facing rads and can go with slower/quieter fans.

With the exception of extreme overl9ckes a cou at 40c vs 60c will make no differance. Some of my favorite loops used almost no fan and very little coolant flow allowing for a near silent build.

Sure you can. Doing it for quietness is fine. But for component cooling efficiency it's backwards.

Not saying it's wrong overall if you're placing more emphasis on noise. Kind of the nice thing about custom loops, you can weigh the attributes differently for a truly custom outcome.

 

[cdabc123]

Sure you can. Doing it for quietness is fine. But for component cooling efficiency it's backwards.

Not saying it's wrong overall if you're placing more emphasis on noise.

Larger temp delta between the rad and air. Although this "efficiency" is in regards to air speed, and rad size. If you dont care about those things you are correct that the loop should be optimized for the limited disipation of the block.

 

[Thunderdolt]

This I don't get. How is running hotter water over hot pc parts more efficient? Remember, you can always add more radiator/fan. You cannot run more waterblock. You should be maximizing the heat transfer between the waterblocks and liquid by making the coolant as cool as possible.

It is far more efficient on the radiator side. Feeding them with coolant that is 20C over ambient rather than 10C allows over twice as much heat to get dumped from the same radiator. This results in a quieter system which also has significant reserve capacity in the event of something like a period of unexpected high loading or a cooling system malfunction. Alternatively, it allows for tighter packaging by allowing for less radiator.

Lowering GPU core temp from 51C to 49C, on the other hand, provides absolutely no benefit whatsoever. 51C is already over 10C cooler than it is from the factory. Aside from a negligible degree of epeen enlargement, what benefit does lowering core temp to, say, 45C provide? Not performance, not reliability, not durability, not service life. So what then? Again, these aren't builds with extreme overclocks.

 

[NightReaver]

Although this "efficiency" is in regards to air speed, and rad size.

Agreed. Was just getting a clarification on what this loop is weighted towards.

*Edit* Though lower rad efficiency can be solved through adding more radiator. Lower block efficiency cannot be compensated for. But if space/cost/ no desire is a factor, then nevermind.

 

[xDiVolatilX]

You get more passive dissipation out of upward facing rads and can go with slower/quieter fans.

With the exception of extreme overclocks a cpu at 40c vs 60c will make no differance. Some of my favorite loops used almost no fan and very little coolant flow allowing for a near silent build.

Interesting. You like your rads vertical instead of horizontal?

 

[cdabc123]

Interesting. You like your rads vertical instead of horizontal?

I like the airflow vertical so the heat can rise out of the rad. I believe this is "horizontal" ie at the top of the case.

I also still use car heatercores as rads so maybe I'm out of the loop

 

[xDiVolatilX]

I like the airflow vertical so the heat can rise out of the rad. I believe this is "horizontal" ie at the top of the case.

I also still use car heatercores as rads so maybe I'm out of the loop

Wait so your rads are sitting horizontal? Like flat with the roof of your case? Not standing up vertical like in the front? I'm confused lol

 

[cdabc123]

Wait so your rads are sitting horizontal? Like flat with the roof of your case? Not standing up vertical like in the front? I'm confused lol

Yes. So that under minimal airflow the heat naturally rises through the rad. All my cases are homemade allowing for this.

A few cases have space to mount the rad ontop, in a dual rad sequential setup it would be ideal to place the hotter rad on the top and the cool rad in the front of the case, then ensure the case has a slight positive pressure.

 

[xDiVolatilX]

Yes. So that under minimal airflow the heat naturally rises through the rad. All my cases are homemade allowing for this.

A few cases have space to mount the rad ontop, in a dual rad sequential setup it would be ideal to place the hotter rad on the top and the cool rad in the front of the case, then ensure the case has a slight positive pressure.

Question, what do you mean sequential? Like both rads in sequence before hitting a block or pump?
I understand the part about the top exhaust rad disapaiting more of the heat first that is ideal.

 

[cdabc123]

Question, what do you mean sequential? Like both rads in sequence before hitting a block or pump?
I understand the part about the top exhaust rad disapaiting more of the heat first that is ideal.

Yes the radiators are plumbed one after another. The hot water goes into rad 1 (top) where the temp delta from the rising hot air from the case doesnt matter as much, and after the cooler water goes through the front rad that intakes cool air to get the coolant temp as low as possible.

Personally I just go with one top rad and let the loop run hot. As the temp increases the thermal dissipation capacity increases so it's pretty easy to dissipate a few hundred watts with minimal fan use and keeping the parts cool enough (<70c is normally my threshold)