Watercooling Loop Order... Does it REALLY matter?? Here's Proof 

Hi Jay, physics major here. First of all, thank you for actually conducting the experiment. I don't doubt your results, but I have to correct you on your explanation. You seemed to be suggesting that there's a fundamental law of thermodynamics that implies loop order shouldn't matter. There isn't. In fact, the basic argument in favor of loop order mattering makes sense, though it depends on several parameters whose effects are hard to quantify. It is absolutely true that as the coolant fluid passes through a component which introduces heat into the loop, the fluid will get warmer and be less effective at cooling the next component it comes into contact with (and the opposite is true for fluid that passes through a component which extracts heat). The only question here is the effect size. That is to say, "will the difference actually be noticeable?" As you've shown here, the difference isn't noticeable, but it's not due to the laws of thermodynamics. It's due to the fact that the coolant fluid is passing through components quickly enough that the exchange of heat doesn't cause the temperature of the fluid at any specific location to deviate much from the average coolant temperature (though it does still deviate by some small amount). If the coolant were to be pumped through the tubes slowly enough that the exchange of heat resulted in a sizable change in coolant temperature, the difference in loop order would be noticeable. Bottom line is that the combination of fluid dynamics and thermodynamics required to fully understand cooling solutions is not at all trivial, even for people who understand physics, let alone armchair physicists who try to make proclamations without actually doing experiments. The fact that you actually did the experiment is more helpful toward answering this question than my physics degree.

When I was building my custom loop PC I used to lie in bed thinking about different loop orders to send myself to sleep. True story. Not sad at all.

You guys are badass. I search every question I have about PC building and your videos always pop up with answers presented in an entertaining way.

Being someone thats about to do his first loop this saved me a massive headache thank you.

*our bodies are water cooled and no one moans about the order of that*

Jay with the Linus ad transition

When your psu is so powerful it doesn't spin under load 😂

those quick disconnect things are cool

This was a great video. Being a car mechanic myself I've always wondered how a water cooling loop on a computer would be different compare to a car. I know the temperatures on a car are more extreme, but the same concept still applies. Great video here!

I appreciate this a lot. I'm slowly expanding my loop and I'm wanting to add the GPU block without adding a second radiator temporarily. It will only be a couple of weeks and I don't really game or do any major overclocking anyway. My loop is purely for aesthetics. I just get the bonus of lower temps.

I have run multiple loops in different configurations. Bottom line, it does matter, but not the way you think. No matter what order I ran, CPU temp never varied more then +/-4 degrees Celsius. What DOES matter is running the cooled water into your pump first! This is the main source of cooling for your pump. This can GREATLY extend pump life. I have seen pumps fail in less than 6 months because of this.

To be honest, I'd love to have the differences in running parallel vs serial(like when you have more then one graphics card) explained/tested and why you might go with one over the other.

Now THIS is a JayzTwoCents video, wooo!

Getting your computer temps in Celsius and your ambient in Fahrenheit .. NICE!

Still one of my favorite videos, and I share it with friends whenever they ask me questions about how to determine loop order or how I determined my loop order lol.

Thanks for dispelling another myth that people seem to never understand Jay. Nice work! Keep schooling the noobs!

Love your content, Jay. What I really like is that it is always helpful & informational (instead of just trying to be funny, showboating, etc.). Over time yours has become my favorite tech/hardware channel. Specifically, your videos were super helpful as I was getting started with water cooling. Thanks for all you do!

wow those GPU temps are really impressive . I know it's not in a closed case but nonetheless . One day I will maybe do it . Good work Jay

This maybe the best jayz vid I've seen. Learned a lot here. Thanks.

I was wondering about this. Very helpful to know order does not in fact matter (working toward first water cooled)

Do parallel loop testing!

0:56 Okay Linus, thanks for the uncanny transition from your topic into a sponsor xd

Thanks for this, I’ve never even thought about this but I’m new to the pc building community and it’s nice to pick up the right knowledge and useful tips straight away

Don't know much about physics, but that part about the radiator being the cooling part of the loop is what made this whole thing make sense before he went to demonstrate it empirically

It's literally thermodynamics, specifically the Zeroth Law of Thermodynamics, which is if two systems are each in thermal equilibrium with a third, they are also in thermal equilibrium with each other. At the end of the day, physically, the loop order literally does not matter. As Jay said, having the reservoir feed the pump, and having a radiator with enough thermal mass to dissipate the heat, will ensure that the system will find thermal equilibrium at idle and load. It's simple and demonstrable physics. Great video, Jay!

It is called Equilibrium.

You learned a lot from Linus, especially the way of introducing a sponsor

OCCT is rendering the Graphs @JayzTwoCents Thats the spiking

Great video, i was racking my brain about this. This will simplify my life during my current build for sure.

I have been using Corsair 280 AiOs for 10 years and they worked just fine top mounted. Last week, my i7 6950x decided its time to go to digital heaven. I upgraded to i9 12900k and that AiO was no good, so did research on custom loops. This video was helpful as I thought order of things were critical and you have proved that not to be the case. Thank you. I have never cooled my graphics card and never found it necessary on my system. Today I installed all Corsair Hydro components into my 10 year old box which is a full tower, but now used 3 radiators all with push and pull fans for each rad and Corsair Hydro pump reservoir combo XD5. BTW I also used the water block JTZ edition and apologies as I had no clue it was related to your channel. I only got it as it's the latest in the series and read reviews that earlier blocks didn't help 12900k very much. My loop is currently in drip test mode, so have no clue yet as to performance. My loop order is pump, 140 rad, cpu block, 140 rad, 280 rad, then back to pump. I'm looking forward to seeing this work later once no leak system is ensured. Thank you again, your videos have been instrumental in getting my custom loop in place.

What about multi rad setups? gpu -> Rad ->cpu ->rad -> pump/resivior ->gpu

I had thermal sensors in my loop back when I had a loop. Going through two GPUs and the CPU raised water temperature by 3 degrees under heavy load. It was good proof that in a loop with proper flow, the water doesn't heat up significantly at each block.

What would be more interesting to know is the difference that radiator fan orientations can make in a case. For example if a front intake radiator feeds hot air to an exhaust top radiator, compared to both radiators blowing air in or out. Practically all Jay's builds I've seen have one radiator pushing hot air to another, which I've heard actually makes a noticeable difference in overall water temperature. Would love to see more tests on this.

Agreed.Thermal gradient in most liquids is minimal so steady state temps are what matter. This is dictated mostly by the cooling performance of the radiator.There is a prerequisite. You need sufficient rate of flow to push the fluid around the loop for the rad to do it’s job.Then the sequence, (to all intents and purposes), does not matter.

Thanks for this, my take away lesson is to strategize tube trajectory to avoid bends as much as possible whether it be from CPU to GPU or vice versa.. Removes a little stress!

I understand loop order in a single rad setup does not matter however, it follows the laws of thermodynamics to come to the conclusion a greater temperature difference will increase the cooling effect. (More heat can be dissipated with a larger temperature differential). This is because there are two heat generators (A B) and one heat dissipation element (C). No matter how you order these ABC, ACB, BAC, BCA, CAB, CBA hot water from one component will flow into the radiator be cooled then be warmed by another component which then flows to the first component again. The following assumptions apply (CPU is 80deg C, GPU is 60deg C, ambient is 20deg C, at all times, heat dissipation causes temperatures to balance in exactly the center of the two temperatures eg 60 deg coolant with 20 deg ambient through rad => 40 deg coolant as a result ). What may be interesting to test is the following: CPU, RAD, GPU, RAD vs CPU, GPU, RAD, RAD. In the latter setup the CPU may heat the coolant up to 50 deg C, that then flows to the GPU where the coolant balances between 50deg (coolant) and 60deg (gpu) at 55 deg. This coolant then passes through two rads to reach back down to 20deg. VS In the former however, the coolant cools the CPU down to 50deg (20deg, vs 80deg) then passes through rad to become 35deg (20deg vs 50deg). The coolant then flows to the GPU and cools the GPU to 47.5 deg (35deg vs 60deg). The coolant finally passes through the final rad and reduces to 33.75deg this then repeats until the coolant settles at ~34.5deg with the same effect. This means the former setup reduces the temperature of the GPU by over 7deg. A considerable amount (~15%) difference

I wanna see a loop with a rad before and after cpu,thus having a rad for each component in a single loop. Just for comparison.

"You know what does matter? Protecting your internet experience!" Wow he's really turning into Linus

Thanks Jay. Your series of water cooling loop videos has been very helpful.

Thanks for the info. Doing a AMD 50th Anniversary Build with dual Radiators and was concerned about the order of the loop using the EKWB for the Anniversary Edition 5700XT and a R9 3900. Your the best Jay. BTW, I love 15 mins away from PPCS! They are the best!

0:56 doing it the cool linus way ! =)

at normal pressure: 0°C: Water starts to freeze. 100°C: Water starts to boil. 1°C difference is 1 Kelvin difference. Very easy if you compare it to fahrenheit ;)

Jay remember this. Room temp usually 72°F ~ 22°C Great vid comparison with data. Thanks! Hope this helps. Cheers* *added tip: °C=(°Fx2)+28 Little trick to do in head and is pretty accurate. I know ppl, correct formula is C=(Fx1.8)+32, but multiplying 1.8 in head is something "don't wanna think about" even though can do, lol.

Good to hear you use Fahrenheit not C I'm English and still use Fahrenheit. Great videos I get a lot of information thanks

This was really helpful for me. After 10+ years I want to build a water loop and these type of video helps take care of some of my questions.

I've done tests with loop order with a 2600k and a HD 5970. the GPU being first or last didn't make any difference. and an overclocked 5970 is possibly the HOTTEST GPU ever made.

Yoooo that nordVPN transition !!! He just did it linus style !!!!!!

I'm here from the future, Jay... and NordVPN is a no go!

This was a great explanation Jay! I believed before that Loop order was true but when I replaced my flexible hose with glass pipe and changed the order i noticed that it didnt matter...... Great info for the newcomers on liquid cooling though!

Infrared imaging camera can provide a pretty nifty visual representation of where the hot spots are in various setups.

A better experiment would have used two machines and two loops, each loop with one GPU and one CPU from each system. That way you can put load on one system and leave the other completely idle, and see how much each component in the idle machine increases in temp. If one increases more than the other, your claim is false. If they both increase by about the same amount, then your claim is true.

I totally agree with you on this Jay. I got a Core P5 and my loop runs from my CPU to GPU to Pump to Rad. I have zero issues with heat, period. Loop order does NOT matter. 6600k delided @ 4.8 GHz (1.3975 V) running 60°C max and Strix 1080TI O8G oc'ed further running max temps of 44°C. Thanks for all you do. Your the Real Tech Jesus....

insta-liked for the smugface in the thumbnail!

The Celsius scale is very easy to navigate, Jay: 0 deg. - water freezes, 100 deg. - water is boiling. This is convenient for figurative perception. But on the Fahrenheit scale it is already difficult, so I have never seen either frozen or boiling alcohol in my life. 🤔🧐🙂

Very valuable info - even 4 years later. I have always said the same thing, order doesn't matter. But I had no stats to back that up (I am an engineer - I knew I was right without stats). This is a very good confirmation that I will share with the haters that occasionally cross my path.

Thanks for the vid Jay! Can you do a video on whether running a loop in parallel is no worse than in serial as long as flow is ok? So I have something to send people when the say my loop sucks.

Loop order >implying each component in my rig doesnt have a 360mm rad

I don't think Jay recognizes the magnitude of the pun he made when he said "it's *cool* and all".

I love this plug n play connections

Jay I love your videos man!!!! You're the shit. A loyal fan.

Private internet access vs Nord VPN ? I can't decide .

This is even more confusing than the time I locked myself in my car.

Can't stop staring at his amazing hair.

12:22 If you have no patience, like Jay..watch from here. Error of margin is usually 1-2c, he says room increases as SoCal is HOT AF. It doesnt truly matter your loop order...just res feeding pump. Thank for all your hard work Jay and Nic!

Good video. I have wondering if mattered if you have both on the same loop. Well done.

Fahrenheit really is the worst unit on earth. I mean imperial length and weight also are very "unscientific", but Fahrenheit is just a complete mess. Anyway, nice video!

0:57 Man, that was smooth, but you've still got a long way to go to catch up to Linus on the $hilling department, lol.

With only 1 small radiator u are correct. With 2 big and a high end setup I bet the sound would come down if mounted correctly. And vice versa.

I'm making my own out of oak, aluminum and plexiglass. I'm parting out a Koolance ERM and Exos2 for the controller and startup system and some of the trimmings and using al Alphacool XT45 1080 radiator, an EKWB 400mm reservoir and a Koolance PMP-600 pump. 9 Aerocool fans and 4 mcpctech fan grills for blowholes in the plexiglass on each side. I'm taking a lot of pictures of my current ERM system setup so I can repurpose the startup circuitry and controller in my homebuilt system.

Jay 70 F is about 20 C.

Hey Jay, is there a cooling difference when water cooling multiple gpus in series vs parallel?

It does matter... when you have small water flow. I tested my loop with 25 litres / hour (from 500l/h) and then it really mattered. Because you have so small body of water to heat up per second, so the water is actually hotter aftter exiting the block. I got 5°C higher at the exit of a loop cooling 300W.

You are absolutely correct. As long as the loop is flowing, it WILL cool everything. The fluid will not be there liong enough to really heat up is the radiator is working.

Good to know. Just received my CL SMA8-A. ONE DAY..... I too will build a badass custom loop.

The real question is, why do people state their opinions as factual when that's not how it works?

For those of you that still believe this myth here is some science why it doesn't make any sense. Lets say your GPU the moment it gets under load goes to 100C. The fluid will then take the thermal energy because it is at a lower temperature. thermal energy goes from High to Low. (this is also why heat death is inevitable). the GPU then cools down to 70C. that fluid is not 100C, it is far less. The fluid has a high specific heat. This means it take more energy to raise the temperature. So the fluid lets say is 50C. it then goes tot he CPU. the CPU is temperature is HIGHER than fluid temp so it actually gives the fluid more energy. but if the CPU is lower than the fluid temp in theory the cpu will heat up but in practice it doesn't. this is because the block is design to facilitate thermal energy transfer in 1 direction. there is an imbalance of surface area between he cpu and the fins in the block. This on top of the fact that the fluid is moving so fast that a section of water doesn't have time to heat up the cpu.

Haha.....he said “huge load” haha haha 🤪🤣

I watched the first minute, believed that I can't mess up the loop direction, build a custom loop for a 4090 in 7h (case mill&drill incl.) and booted Cyberpunk (4k/ultra) for an hour: Water temp max. 48°C, GPU max. 62°C. I can finally play at full load while hearing absolutely NOTHING from my PC. I love it😎👌

True story, Jay! The most impact on a system would NOT be component order, but indeed the routing and bends. Keep it more straight, simple, it will optimize temps.

You moved the Rad, tests are invalid.

Just making my first loop… thanks for this video 👍

i like the chill music in the backround jay

Ran my custom loop in reverse order for six years with no issues with temps. Gonna switch things up and go reverse reverse order after I’ve finished deep cleaning it.

First thing, excellent video and valid information; hopefully people listen. Secondly, to further reinforce Jay's points...order doesn't matter. The key here is transfer of BTU through materials and turbulent flow. As loop temperature rises, the radiator expels this heat to atmosphere and heat transfer continues at the rate of materials used (most common cases this is copper). As long as there is a delta between loop temperature and components, transfer will happen.

If using 2 Radiators, there will probably be a difference where you place them Res>GPU>Rad>CPU>Rad is one of the best options i guess... If you put the radiators after each other the last component wil be warmer, if you split the rads you basicly have somewhat 2loops in the same loop...

The best order for your loop is whatever is most convenient for your configuration. The differences between order is so minimal that it doesn't matter, and if you can get by with shorter runs (or more fancy runs) or more direct runs, go for that. I ran pump -> cpu -> rad -> res -> rad -> gpu -> pump for over a year with no problems.

What about 2 loops. Pump>gpu>cpu>rad and back to pump. Pump>gpu>rad1>cpu>rad2 and back to pump. Will 2 radiators will make a big difference on hard load?

Thank you sir. I did see. a lot of general criticism as to how the loop should be set up, however looks like science proves the point. Get a 360-420 radiator and feed the liquid through as you indicated.

I shared this video with a guy on reddit and all he could do was insult me and say he did not care he knew different from personal experience. I share even more things that proved him wrong and he just started trying to reword what i was saying to make it sound like i thought the rads did not even cool the loop. I kept coming at it with proof and he kept coming at it with insults to me. Got to love people.

Did my first one today, did it by pump to rad, to cpu block back to pump and repeat... I thought of a car cooling system and went by that and then found I had a instruction manual telling me (pump to cpu, cpu to rad, rad to pump and repeat???) Thought it was odd but meh it still goes... and now I've seen you talk about it lol no worries at all and for my first crack nor bad... set up I got is a c360 thermaltake pacific soft tube kit, goes hard 👌

Shame this was 6 years ago, but thank you anyway! Good empirical demonstration that the ordering is unimportant.

"it's not necessary but it's fun", probably the most honest water cooling comment ever. the maintenance is what kills it for me. with air cooling, all you have to do is blow it out once in a while and occasionally replace a fan, but many fans still last a decade+

I’ve found putting your radiator after your pump does increase temperatures quite a bit... (but you have to run the stress test for a long time)

Hi Jay! The order would matter only within first few minutes of the loop running when the thermal energy produced by the components and the energy dissipation in the radiator reaches the equilibrium. That means the second component will be first to reach the max idle temperature by a few seconds. Very interesting case for the precise measurements but completely negligible in a real life scenario.

The main thing you want is that the radiator is before the liquid goes into the reservoir, so the reservoir doesn't get damaged, and as you said, from the reservoir right into the pump. Nice vid Jay.

Thank God for this video. I had been searching like hell fpr a little insight on which way the loop is supposed to go. The only constant i see is that the radiator must be connected to the pump / resevoir. Other than that the order doesnt matter.

this helps beginners like myself

Great Video Jay thanks for putting this one issue to bed

Drinking game of death : take a shot every time Jay says CPU or GPU

So my SO saw me wathing YT videos instead of reading for my entrance exams (engineering) and got mad but lucky for me this video was about thermodynamics so she took the excuse that I was actually learning from your video.. I should go back to reading.