Testing one of the Weirdest Water Cooling Myths - A SLOWER Flow Rate will result in BETTER Temps?

[@malcolmparkes9981]

Thanks, Roman, shut me up once and for all, I was convinced there would be a bigger difference, thanks for proving me wrong, I love science.

[@TickTock725]

We need more people like you man

[@arkie87]

anyone who knows anything about heat transfer would know it is not possible. Out of curiosity, what was your logic or reasoning for this being the case?

[@arkie87]

his measurements are not good. his results are not steady state. simple thermodynamics calculation gives 7°C temperature difference at his purported heat input and flow rate.

[@malcolmparkes9981]

@@arkie87Water coming out of a CPU/GPU waterblock should be hotter than water out of a radiator obviously, I thought the difference would be much higher than it seems. At least 10°.

[@arkie87]

@@malcolmparkes9981 well, his numbers are not correct. Either the flow rate through the block is higher, or he did not wait long enough for steady state. Steady state temperature difference should be 7°C

[@bigben2k]

Been there, studied that. A good (high) flow rate creates turbulent flow, which reduces the boundary layer, which improves heat transfer.

[@ark_knight]

FORCED CONVECTION. Don't even need boundary layer to explain things.

[@arkie87]

it also reduces the temperature rise in the fluid, which reduces the thermal resistance as well

[@arkie87]

For microchannel cold plates, though, the flow might be laminar, so increasing the flow rate does not create turbulence. It does, however, affect the boundary layer due to developing flow.

[@escalator9734]

It's not even about that, more flow = more power, that's just it. Doesn't matter if you're talking about pc watercooling or home radiators or industrial AC. more flow = more power, more dT = more power

[@Uffeful]

Did that in practice for a customer where the flow rate was set but I wanted to increse energy transfer so I reduced the inside diameter of the tubing and that increased the energy transfer because of turbulent flow. But in computer cooling it may be to small amount of energy per area to matter.

[@markvvelsen]

Thanks Roman! I was really struggling to set my pump speeds in this hot weather and what would be best. Great timing and very useful as always. Give Shiek a hug from me!!

[@JMUDoc]

It's interesting that doubling the pump speed from NORMAL only reduces CPU temps by 0.4 deg... so much noise for so little effect.

[@gamertechlive1780]

Jay did a review about this but did a fast demonstration but did not record the temps just like this. He said he doesn't care what his pump speed since he also goes for manual pump speed control and don't use PWM 4 pin pumps that's why he always forgot or have no idea what's his pump rpm or pump % is he always goes for the highest possible as long as there is no noticeable noise by manually tuning the knob of the manual pump controller. In this video 100% is useless and annoying because of noise. And slowest is not a good idea in terms of cpu/GPU temps. Normal or above normal as long there's no noticeable noise is always the sweet spot. If the loop is complicated and restricted then slow and normal is not for you need something in between normal and max.

[@idontwantacallsign]

There is something wrong with his numbers. I would think the flow rate are off by a factor of 20.

[@sarogoth]

@@idontwantacallsign I think he doesn't have it calibrated right. I have the same sensor, you must calibrate it at zero flow with the pump(s) turned off before it gives you a more reasonable number. I've had some trouble with mine trying to get it calibrated correctly and giving me a consistent number. But now I don't feel so bad since der8auer isn't using it correctly either. I say just rely on a flow indicator, as long as that thing is spinning you're good. Otherwise his test is valid and I agree with his conclusions.

[@Saddedude]

​@@idontwantacallsign Remember that pumps are rated for maximum flow with zero backpressure. Even 5psi of backpressure DRASTICALLY reduces pump output on these little d5s.

[@zpnk]

I'm running my pumps at around 60%. Because as der8auer says, you can't hear the pump. But the gain in lower CPU temps with running the pump at 100% is so minimal that there's no reason to do so

[@drspod]

This should be self-evident if you recall high-school physics! The heat transfer from block to water is dependent on the heat capacity of the liquid flowing through it. The heat capacity of the water is directly proportional to the flow rate. If you double the flow rate, the heat capacity of the water flow through will be double, so it will take twice the number of Joules of energy to raise the temperature by one degree. I remember doing a very similar experiment in school to measure the specific heat capacity of water by measuring flow rate and temperature deltas along a pipe.

[@Verpal]

Small correction ''If you double the flow rate, the heat capacity of the water flow through will be double'' is only factual in open system, rarely PC cooling are open system. My own setup is open, but that's just because there are subsidy for water bill in my city, so I decided to open up my loop.

[@broklond]

lol what? No. Water's heat capacity is constant. Heat transfer is what is better with higher flow rates and it is dependent on the Reynolds number. Also it won't be double if you double the flow rate, the correlation is much more complicated than that.

[@DarkoPetreski]

Wow you are like so cool, bro.

[@wfb.subtraktor311]

The only point where a higher flow rate wont be better is if you become so ludicrously fast that the water starts creating significant amounts of heat through friction. That speed however has to yet be reached.

[@audaxxx4202]

@@broklond Exactly this. Heat transfer is reduced at lower velocity's as the thermal boundary layer is thicker. This has an insulating effect trapping the heat next to the copper surfaces.

[@17473039]

Component order does matter. Radiators work most efficiently when the delta T is highest. 3 radiators in a row means the third rad in the loop sees a lower delta T as the previous 2 radiators have reduced it. The effect is lessened with higher coolant flow rates but not eliminated. When you are in a situation with an abundance of cooling headroom, having all the radiators in series followed by all the heat sources makes sense as you get the lowest loop coolant temps this way to minimise system temps. When you are limited with cooling capacity, alternating radiators and heat generators to maximise cooling efficiency of the radiators makes sense.

[@mikee9167]

It just doesn't matter enough to worry about, usually. The effect of loop order on overclocking or performance is likely going to be smaller than what comes from fluctuations in ambient temperature. Especially because you'll probably have high pump speeds that result in tiny temperature deltas across the radiators, if you are trying to get maximum cooling. If, on the other hand, you are trying to optimize your pump control for the lowest possible speed at any heat output, then the order will matter a bit more for the low speed end.

[@broodro0ster]

But that also means the water will warmer entering the first radiator, so the first one will a more efficient actually. In the end it's all the same. Just make sure the flowrate is high enough. More components and more heatload require a faster flowrate to stay cool.

[@17473039]

Higher flow rates mean more noise, and if you make the system as a whole more efficient rather than a single radiator, you can run your pump and fan at lower speeds for less noise. It isn't big gains granted, and if you are space limited or routing limited it isn't the end if the world. But if you have the option to make those fine tuning adjustments to your order to optimise it for your setup... Surely thats the difference between using auto voltage when overclocking and manually setting the minimum stable voltage when tuning your cpu?

[@Ephem13]

This is a matter of technical vs practical application. In a technical evaluation of it all, yes it would be more efficient to alternate heating and cooling sources. In practice, the advantage is so small that it becomes negligible. You could spend tons of time calculating the exact amount of cooling needed after every heat source, the perfect pump/fan speeds for each temp. range, the ideal tube size for every connection, and on, and on, and on. Or, you could simple figure out how much cooling the total system will need, set the pump to the highest speed you can tolerate, set a PWM fan curve that is quiet under everything but the most extreme conditions and be close enough that no real world performance advantage can be measured.

[@Gainn]

I had issues with temps on a top-mounted 360mm rad and ended up splitting the flow right after the CPU to a pair of 120's in parallel. Those ended up giving me much more consistent (and lower) temps than the single 360 (same fans, same speed, just with the third fan as a direct extractor in-between the rads). I'm tempted to try bodging in another 120 after the flow consolidation to see how much of a difference that makes. I originally thought the 360mm rad might have been faulty, but it ended up in another PC and it gave results well within the rated spec.

[@witchsorrowful1918]

In the next episode of Mythbusters : der8auer, ....

[@mwnciboo]

*...does Ryzen make gaming smoother!* XD XD XD It should be "Der8auer Bullshit killer"".

[@montec6113]

TLDR: Slower flow rate will result in higher temps, in this case 5c.

[@brudel001]

Not just "slower" but basically a standstill compared to the normal and max flow.

[@lagynas]

Thanks, I fell asleep watching it ;D

[@Azerkeux]

This myth likely has roots in reality but only on a scale not appreciable by either the flow rate or temperature differential in a PC. It's just not possible to reach the point of diminishing returns given how low the flow rate already is in a stock setup, to see this scale it might be worth it to ramp up the flow rate artificially with much more volumetrically efficient pumps Edit* I actually recall while I was in automotive technology classes in community college that it was suggested to us that a typical automobile air conditioner is most efficient at around 80% fan speed through the evaporator

[@mekko1413]

@@Azerkeux You are correct and this is even the case in certain scenarios with in data center cooling particularly with free cooling heat ex changers which are more or less like an AIO cooler. During your transitional season time periods slowing down your flow will increase the time you can stay on heat ex changers. You can run your pumps and fans so fast that they do not do their job. I would have never thought someone would apply this to a home PC where the pumps and fans just are not able to reach these thresh holds.

[@lkuzmanov]

Another possible summary: minimal difference between 2000 and 4800 RPM pump speed...

[@OlettaLiano]

Yep, I agree. Loop order does not matter. I've been water cooling for over seven years and have tried every configuration possible. There's never more than one or two degrees difference. My current loop has the water flowing from the radiator to the GPU block, then the CPU block, then to the pump. This configuration has not noticeably increased the CPU temps at all.

[@esra_erimez]

Gamer Nexus put windows into the cpu block and radiator. And now this. You guys are awesome.

[@gio9789]

The timing of this video is perfect, just bought the parts for my custom loop

[@hyakinthos_0902]

lol me too fist bump

[@noreng9333]

The video won't make a difference to the performance of your custom loop, and failed to impact your purchase because you saw the video after you made the purchase

[@hyakinthos_0902]

@@noreng9333 ok boomer

[@gio9789]

@@noreng9333 the video didn't fail, I already studied thermodynamics

[@lewderoge7386]

Giovanni Lizzi 😂😂😂

[@user-bf5sc8pn8x]

You mention two mostly-unrelated hypotheses but only test one of them. I'd be interested in the results of multi-radiator loops at different component loads, speeds, and loop configurations.

[@der8auer]

yaaaa 2nd video is coming :D

[@filonin2]

@@der8auer I'd be interested as well. Since the blocks benefit from high flow rate and the radiators benefit from slow flow rate, maybe the cpu and gpu in series with the radiators in parallel would be ideal? Having 3 120 radiators in parallel would lead to high residence time for the coolant compared to in a single 360 and you should still be able to get the benefits of high flow rates through the fins on the blocks. There's gonna be a lot of tubes though lol.

[@riffsthatkill2180]

@@filonin2 I hear flow rate equalizes to a degree, so you might not see that much of a difference where you can say the blocks are getting a much faster flow rate than the rads. Doing rads in parallel might split the flow, but it would lower the flow overall for the system too, meaning the flow over the blocks would also get lower than in you had the rads rigged in series.

[@aaardvaaark]

lowering the temp of 10 litres of water by 1 degree is more cooling than lowering the temp of 1 litre of water by 5 degrees. The volume of water being cooled is what I think people are missing.

[@arkie87]

probably

[@TheZoenGaming]

Any car mechanic would be able to tell you that faster coolant flow equals better cooling. I'm confused how this myth ever started...

[@pippifpv]

Makes sense.. It's all about the thermal capacity of the setup..

[@CuttingEdgeRetro]

This argument dates back to the earliest years of PC enthusiast watercooling. I still remember having this argument with a guy at a lan party back in 1999 when he was convinced low flow rate and a rad between each part was king. haha XD

[@emperorSbraz]

same.. it's a neverending meme.

[@Thunder-wd7ti]

If you go for very low flow rate, you must put a rad between each part. It is just better to have higher flow rate and than it doesn't matter where you put the radiators. Maybe if you have >1000W power consumption in your loop it could matter a few degree.

[@syx3s]

yeah. if you want a cooler system: add more rad.

[@eric4946]

Earliest years of PC cooling ?? I hear stories about people from the 70s having this argument about engine cooling.

[@joe2d]

@@Thunder-wd7ti Agreed on just better to have a higher flow rate. But even if a rad between each part is better for low flow rate, it still doesn't matter in the real world simply cause low flow rate is gonna suck regardless. Radiators are optimal at a certain range of liquid flow and airflow. Same goes for the pump and waterblock, they are optimize for a certain flow rate. Too low a flow the parts are not working optimally, too high it reaches a point of diminishing return ie more noise, less cooling. If some one would choose low flow water cooling for whatever reason, better off save the money and headache, just stick with air cooling probably get better results too.

[@Sqtgdog]

Well you proved the flow rate theory, but I don't really see that you proved the alternate radiator placement theory.

[@cadetsparklez3300]

yeah the one in the title doesnt make sense but I can see a reason for the alternating radiator and heat source if you are water watercooling something like ram which took very little power, the energy from the cpu or gpu will just be heating up the ram to a point further than it would have heated up outside the loop. I see this a lot and dont understand where people get the idea that energy only conducts one way the only way it couldn't make a difference is if the heat gets perfectly distributed throughout the loop as if water is a superconductor, which if so then the flow rate wouldn't matter...

[@waldolemmer]

I'm a GN viewer and I'm here to pester you!

[@charlieodom9107]

That is a DP cell. The plate inside is called an Orifice plate. These types of meters are used in industrial applications. They can be used for level as well as other applications.

[@Michael-OBrien]

By unit of mass, water has 4x the heat capacity of air. In other words, if you need 1W of power to raise 1 kilogram of 1 °C (not actual reality, but bear with me) then you’ll need 4W of power to raise 1 kg of water the same, 1 °C temperature. Thanks for putting some metrics to this factoid, Roman.

[@HardwarePerformance]

HW Legends EVGA SR2 Classified pls

[@rtyzxc]

It's like saying lower fan speed gives better temps because the air can spend more time inside the heatsink to cool the cpu :D Obviously higher flow gives better cooling.

[@arkie87]

lol who says that

[@christophervanzetta]

Too high a flow can be negative though as it won't be cooled down efficiently

[@arkie87]

@@christophervanzetta This makes no sense from a thermodynamic/heat transfer perspective. More flow might experience less temperature drop as it passes through the radiator, but it will also heat up less as it goes through the cpu block. More flow will always result in lower CPU temp.

[@blackmennewstyle]

I remember JayTwoCents from 2018 was basically saying this almost in all his custom water-cooling builds because people wouldn't stop bringing that myth on the table lol He even made a video dedicated to loop order but people, like Sarah Connor said, "they never learn!"...

[@diezgp]

That guy always gave me a vibe that he doesn't know what he is talking about.

[@detmer87]

@@diezgp Well there is a big gap between an American and a German Engineer... ;)

[@mmmdawe]

i think those 2 guys above didn't understand your comment lmao

[@diezgp]

@@mmmdawe Yeah, later on i realized what he was saying was the completely opposite of what I understood. My bad, just woke up when i saw that comment 😂

[@OlettaLiano]

I love Jay. I think he's one of the coolest guys on KZbin.

[@donh8833]

Der8auer, Im going to give this video a very cautious approval but you are making blanket statement which may not be true. High flow rate does help insure that you always maintain maximum heat transfer up to a point. But heat transfer thermodynamics is not a simple topic and it involves a lot of complicated models differential math simulations. There is a maximum thermal conductivity, and no matter how much faster you pump that fluid you arent going to get improved temps. You can have something blazing hot on one side and cool on the other. (Ie coals or shuttle tiles). Once you hit that conductivity limit you can't do better. This is why improvements exponentially decay with pimp speed increases.. Where I work, we work with over 30 inputs for a coil design. I work with this code as an engineer on a very regular basis and have over 10 years of direct experience with coils and heat exchangers. However if a fluid is too viscous and you create too high a head pressure you are wearing out your components faster and possibly even generating more heat through friction and vibration.

[@galxieranger8277]

So try this - pump flows to GPU/CPU (either order), goes into T- type fitting, splits off to 2 or 3 radiator inlets, radiator outlets to another T-type fitting, single line returning to reservoir and onto pump to continue the cycle. Full flow would pass CPU/GPU, but the flow through the rads would be slowed to provide longer contact. Edit: seeing other comments about auto radiators. This method would be similar to why a dual-core radiator is better than a single-core radiator. It doubles the surface area for the coolant and slows the flow rate through each core, thus increasing the time the coolant is in "contact" with the air

[@taiiat0]

so, the Radiators in Parallel? unfortunately, the Water will prefer the path of least resistance, and whichever Radiator line offers even the slighest bit less resistance, will get substantially more flow than the others. additionally if you put all of the Radiators in Parallel, how exactly is the Water spending more time in the Radiators? even if it was a perfect 33% split, the best case scenario would be the same effective course length, would it not? while in reality it would effectively be markably shorter.

[@galxieranger8277]

It's a closed loop, so the pressure will remain constant. But by dividing between 2 radiators, the flow through each will be halved. Since each radiator needs to handle half the volume of coolant, at that pressure, the flow will slow down, thus leaving the coolant in contact with the radiator for longer, giving it more time for heat exchange. The "path of least resistance" concerns are not how fluids in a closed loop work. The fluid will still travel through both radiators at almost entirely the exact same pressure and volume - unless it has a serious blockage - but that's not a fault of running the radiators in parallel - that would cause a series run of radiators to fail too. If you watched the tests, he got a large delta T between inlet and outlet of radiator by letting the coolant flow slowly, which this parallel radiator system would do. But the better delta T on the CPU/GPU was with greater flow through the system, which this parallel system would do.

[@taiiat0]

@@galxieranger8277 and how Parallel do you expect to need to go to get a markable difference in time spent in the Radiator? you're going to need to go to pretty extreme lengths to overcome the greatly reduced volume moving through most of the Radiators, mainly the ones that aren't as straight routed from the Pump as the others. unless you're going to presume you can bump the Loop up to a Pump that can handle real flow and pressure like around 350L/m or otherwise something just silly. might i remind while we're here that the Liquid Loops we use are not pressurized to any significant degree. the pressure is... quite low, and always with extra space in it. but then you don't really need hardly any of that extra Radiator space anymore and you're back to the sort of ratio you started with, just with more of everything.

[@galxieranger8277]

@@taiiat0 the coolant flow is still technically pressure. If the pump is running 20 GPM, that's the flow through the CPU/GPU blocks and through a single radiator. If you add a second radiator in parallel, the flow through each radiator will then be 10 GPM - half the flow rate. This slows the coolant down so it remains in contact with the air for a longer period of time. This is the same principle behind dual-core radiators in cars. It divides the flow through 2 cores, thus slowing the coolant down so it remains in contact with a heat-exchange surface for much longer. To see this principle in action, albeit in reverse, turn on the water in your garden hose. Now put your thumb over part of the end of the hose. The water pressure doesn't change, the volume of water doesn't change, but the speed at which the water flows changes. In this case, it speeds up, because it needs to go through a smaller flow path. Remove your thumb, and the water speed slows down. It's derived from Bernoulli's Principle. And in the video, we saw that by slowing the coolant flow down, there was a greater drop in temperature from radiator inlet to outlet than with the faster flow; however, that slow flow caused the coolant temp in the CPU/GPU to rise. So we need fast flow through the CPU/GPU, and slow flow through the radiators, and we get that slow flow by running the radiators in parallel. Because the same volume of water is now technically passing through a pipe twice the surface area of the original pipe, which means its flow rate is halved. We get fast flow through CPU/GPU to best pick up the heat from it, slow flow through the radiators to best exchange the heat to the air.

[@V3n0m151]

I'm about to pick up a MPS 200 because of this video 👍 thank you very much!

[@fermitupoupon1754]

For some reason I now want him to compare having 2 120mm radiators parallel versus a single 240mm U-flow and a 240mm X-flow radiator.

[@magman687]

I second this

[@Snow.Drifter]

It amazes me how many people don't understand basic thermodynamics. Thanks for the testing! Mind doing a follow up where you have a multi-part loop? Something like a CPU+GPU and note the effects on various components

[@idontwantacallsign]

Yes because the numbers in this test simply don't add up. Most of the 170 watt should go to the coolingwater but the numbers don't add up.

[@aarcaneorg]

What about having a high flow rate over the CPU and GPU, then splitting to three parallel flows across three radiators. Your o11 dynamic with three 360 radiators as intakes would create a great positive pressure case, and splitting the flow three ways would create three slow flows across the radiators with fast flow for the CPU and GPU.

[@JakusJacobsen]

It might at first glance to be a good idea for the water to be as cold as possible at the radiator outlet, and low flow rates do achieve that. Imagine how low the outlet temp would be with zero flow though. That wouldn't be a good idea. A radiator dissipates heat energy along its whole length. If the end of the rad is only half the degrees above ambient that the start is then the end is dissipating only half the energy of the front. Its counter intuitive but in actuality for best efficiency and maximum heat dissipation we want the coolant temp to change as little as possible between inlet and outlet. The only way to do that is create as much flow as possible.

[@tylerdurden3722]

Where the water and the radiator makes contact (i.e.where the heat transfer is happening), is called the temperature boundary. There are various temperature boundary conditions that influences heat transfer. Increasing flow, improves these conditions at the boundary, increasing heat transfer at the boundary. So, slowing down the flow rate in the radiators will reduce the temperature boundary conditions where the water makes contact inside the radiators. Hence why a higher flow rate is good (but this is just one of several other factors to consider when slowing down flow) In some ways, (in a scenario where something is cooled with a heatsink), increasing flow is equivalent to applying more pressure to on the heatink, thus improving heat transfer at the boundary. So lowering the flow rate is equivalent to loosening the heatsink screws (except, imagine this happening at the radiator of a water loop) So, it would be better to have those rads in series, to increase flow, and thus improve the temperature boundary conditions inside the radiators. Plus, since the flow rate in and out of both serial and parallel setups should be the same, the water should spend the same amount of time in the radiators, in either configuration

[@BlazeFirereign]

Looking at the inlet/outlet temperature in a test like this, and saying that the lower outlet temperature = better cooling, is a misunderstanding of thermodynamics. What matters is the amount of heat energy that the radiator can dissipate, in watts. That power dissipation could cool a lot of water by a small amount (meaning a small temperature difference with a high flow rate), or a small amount of water by a large amount (meaning a large temperature difference with a low flow rate), but the thing you care about for the radiator is the actual power dissipation, not the inlet/outlet temperature. I have not seen a single radiator that dissipates more power at a lower flow rate. Every watercooling radiator will dissipate more heat at a higher flow rate, and less heat at a lower flow rate. You can't look at the inlet/outlet temperature in this video and extrapolate the result to apply to 3 radiators, because if those radiators are in series, the outlet temperature will be increasingly lower after each radiator. If you split the flow over 3 radiators, the difference between inlet/outlet might be 4C instead of 2C; but if the 3 radiators are in series, the *total* inlet/outlet difference will be more like 5-6C. This is why you need to look at power dissipation, not at the inlet/outlet temperatures.

[@dandel351]

I think some people forget why water cooling is even used in the first place. It's either to get the most performance possible with high fan and pump speeds and the noise that goes with that. Or its for keeping the system cooler than air can at a low noise level. The more important question is , how much flow is required for the waterblocks used to function correctly and how much radiator capacity is needed to cool the components in the loop. I would be interested to see the same test used on a cross-flow style of radiator , just to see if the temp drop is higher or flow restriction is less. Thick vs thin rads would be a cool test as well.

[@THESHADOW97139]

I've been asking myself the same questions for several days now.

[@marshghoul]

debatable two problems: 1. higher flow benefits CPU cooling - more water passing through CPU block, more heat you can extract from it (difference between Tcpu and Twater much higher than difference between Twater and Tenvironment) 2. you have only one cooling loop - or would rather say that radiator not big enough to dissipate heat into environment what to do? - bigger radiators? - resulting in massive construction and bigger pump to push bigger volumes of water thus making it much noisier - double loop? make big reservoir for water and have 2 separate loops: one to take water for CPU cooling, second to cool down water from it - resulting in even more massive construction and more element thus making it more unreliable and less ergonomic.

[@JuxZeil]

It's all about balance. As you say the block type will matter(pin/fin, jet-plate/high-flow, how thick the coldplate is, etc.), the CPU design will matter(transistor density, heatspreader, etc..), the pump quality will matter(head pressure), and most of all the ambient room temperature. I myself had had a lot of success with specific loop orders, The best loop order say for my X99+5930K+1070 AMP system is res>pump(Delphi DDC type(the same 18W pump used in the Mac G5) with body+top and heatsink kit from Barrow)>rad>coolest component(HDD's(4X 3TB)+M.2 SSD's)>next coolest(mine is my RAM)>next coolest(mine is the VRM's) >rad>GPU>rad>CPU>back to res. There's a scientific reason for this, and although it only gains you a measly 5 or so degrees, for a good OC every little helps. 1)the res itself can have a good cooling effect as it's usually within the airflow path IN THE CASE. 2)followed with a rad you get extra cooling as that rad doesn't have to deal with the same "nominal" temperature equalization the loop has. 3)the "heat-soaking" effect from running the loop on a high OC or just in high use(eg:video editing/encoding) is drastically reduced by ordering the coolest parts through to hottest. And uncontrolled heat-soak can become a "run-a-way" situation... The enemy of heat dissipation. Linus did a little vid on the subject too recently, I think it was one of the tech manufacturers coolers he was checking out. Also, if the "coolest rad first" loop order didn't have any benefits(however you implement it, my X570+3950X+1080TI+Lian Li TU150 system has every fan on a rad so coolest are used for intake), they wouldn't use it in server farms with the way they have to stack the rads in-line within a U chassis. 👍

[@WouterVerbruggen]

It's a classic qualitative vs quantitative thing. One could argue a slower flowrate could be better, as the water has more time to take up heat at a block and more time to lose it in the convector heat exchanger (they are NOT radiators!). Indeed, this effect is not significant enough to matter, compared to a faster refresh of water. The reason is simply that, comparing the two cases, the change of the flow rate term of convective heat transfer equation is significantly higher than the change in the temperature term.

[@chesterfield900]

Thank you for testing this. It solidified the reason why i always run my pump at full speed // the aquasuite seems to be an amazing tool // i think i made the wrong decision with the corsair commander pro

[@MastaSquidge]

There's diminishing returns on pump speed mind you. You may be able to see nearly the same temps with much less pump speed and wear. The reason is once turbulent water flow is reached, any additional flow beyond this gives increasingly tiny improvement. Diminishing returns. If you want to take the time to test this, load up the system and drop the pump speed 5% at a time and run at load for an hour. You'll see very small increases in cpu temp until you get a very sudden increase. The last step before the sudden increase is the optimal pump speed, but you may wish to go one step further back up just to give a margin of error. At this point anything higher might give you a fraction of a degree lower temps at the cost of much higher wear.

[@3800S1]

Another misconception I hear a lot is that having too fast of a water flow won't allow the heat to transfer to the water and cause over heating which is obviously bs.

[@Omaricon]

Right. because heat transfer is like a handshake, you need time to do the handshake to transfer heat. /s

[@JMUDoc]

You've just burnt your hand - do you run it under a slow tap, or a fast one?

[@emperorSbraz]

@@JMUDoc i swear so violently physics optimizes the flow for me for fear of retaliation. :3

[@Omaricon]

@@JMUDoc Haha wash your hans with fast enough moving water and you won't feel that its cold or hot... /s

[@noreng9333]

@@Omaricon That's technically true, but you would probably experience problems with heat buildup due to restriction in the waterblock at such a high flow rate (if you could even create a pump capable of such a ridiculously high water flow), and/or the water block literally exploding.

[@clawhammer7]

Great experiment and testing. As a conclusion, make sure our pump runs at least 2K rpm which produces good enough flow rates.

[@simondale3980]

Good video, having experimented with loop order, I can confirm, As long as your flow rate is high, loop order makes very little to no difference. But when you have multi GPU and CPU in the same loop, flow rate can be an issue, ( avoiding the expense of large tubes fittings, multi pumps etc) loop order can have a small benifit.

[@arkie87]

his conclusion is correct, but his measurements are not: A simple thermodynamics calculation gives 1.5[K]*4.2 [J/g-K]*5.5 [g/s] = 35 W. He is putting 170 W into the radiator, and his measurements only explain 35W. The temperature difference across his CPU or radiator at steady state should be ~7 degrees C.

[@jeroenvanos9653]

You could still get a slower flow rate in the radiator while having high flow rate in the cpu block. If you increase the diameter in the radiator, the water flows slower. But that's just a hypothetecal experiment

[@austinshaner]

I know this video is a couple years old but I'm curious. Knowing that longer time in the Rad= lower water temps, but you sacrifice flow rate at the cpu block.... what if you ran multiple cross flow radiators in parallel. You'd slow down in the radiators because the flow is split between multiple paths, but increase flow rate in the loop as a whole by decreasing the restrictions from the radiators. Best of both worlds?

[@Teatime4Tom]

I can understand the misconception. There is not a simple statement that explains why. But it is just a few simple concepts. 3 statements is the best i can do. 1. Assume the total amount of energy never changes. 2. You can only move it around. 3. Your only goal is to spread the energy around evenly.

[@ericthedesigner]

I run a Hygger pump at 300 gallons an hour, CPU and GPU are pinned at 28c at all times. I'm scared of the water blocks leaking if I set the pump to any higher speeds. The pump is capable to 450 gallons an hour. I use Hygger Aquarium and fountain pumps in my computers. I also run a 95,000 watt 6 pass, water to air heat exchanger as my radiator. Edit: just upgraded my pump last night to 800 gallons (3000 liters) an hour. Now my temps only spike to 43c while benchmarking.

[@icanrunat3200mhz]

This should be pretty obvious to anyone with any amount of engineering experience or education. It's a closed system. Loop order cannot make a significant difference if all of the components are in the same system and you have enough head pressure to move the coolant effectively. If you were to have each component with its own loop, pump, and rad, you _might_ be able to see a statistically significant, although still pretty small, difference versus the same flow and rad size all in one loop. More importantly, though, you would be able to fine tune how big of a rad each component needs to be cooled effectively. Which could be helpful if space to mount rads is at a premium.

[@FOXCRF450RIDER]

I think the English word you were looking for @3:12 is "calibrated orifice" or just orifice. Is cavitation ever an issue with high flow rates? Great video!

[@UNSCPILOT]

I would expect the probably set the max speed on (most) pumps to be below the threshold where cavitation is a serious issue, though it probably still happens in at least some CPU blocks, but I don't have the knowledge to say for certain

[@GMargrove]

This video forgets one element which is important for the whole argument. Radiators are more efficient at removing heat the hotter the water is and the greater the differential with the air temp. So theoretically hotter water in the loop means you can use lower fan speeds on the rad to exchange the same amount of heat. So while your temps may be higher with hotter water, you can run the fans slower. So using a medium pump speed with slow fans is most efficient and you can have a bit higher temps but for disproportionately less noise from fans.

[@Lucaspc99]

Dude is literally a Guru in the Cooling world, and the "Keyboard experts" wanna argument. Lol

[@arkie87]

he knows a lot about computers, overclocking and stuff, but does he have any knowledge of the fundamentals of thermodynamics and heat transfer?

[@SteelBlueVision]

What this video doesn't answer is whether or not there is a maximum flow rate or maximum reasonably achievable flow rate, given the restrictive nature of the waterblock, beyond which it would make sense to interleave radiators and waterblocks, because (the then miniscule) increases in flow rate, even with more pumps or more powerful pumps, becomes impossible or begins to add heat into the system stemming from the pumps themselves. At this maximal flow rate, which is clearly a function of all of the restrictive parts of the loop, the radiator inlet/outlet temp difference becomes the only remaining factor that influences CPU temperature.

[@Johny666EU]

hi, nice test, i like it, there is only one think, the mensurments liter per hour are like coma is 1 zero to the right. the minimum should be 15 l/h and on 2000 rpm should be 210 L/h. as its hard to belive that this pump for and hour with 2000rpm could transfer only 21 liters or water.

[@pudicus2]

You proved that one cpu temperature is affected greatly by flowrate, but extrapolated to a different scenario. This did not prove that a radiator in front/back add between versus at the end does not affect the temperature of the 2nd CPU.

[@arkie87]

it is simple thermodynamics and heat transfer. If the inlet temperature to the cooling block of the CPU is 5°C hotter than the first, than the 2nd CPU will be 5°C hotter than the first.

[@Niklby12]

He was attempting to show that the difference entering or exiting any element are so small that it simply doesn't matter what you do as long as there is high flow rate through the whole system Its simply a negligible difference so the words used often of "it simply doesn't matter" are used (correctly) to sum up the conclusion. For the record the conclusion was that yes having all the blocks in series is actually worse than having it go block rad, block rad (for the second block) just that the difference (for the second block) is so negligible its not worth considering. Keeping a loop simple which reduces run length is likely more important as reduced runs would likely increase flow rate. Then there's aesthetics to take into account ;)

[@arkie87]

@@Niklby12 That presumes you do in fact get enough flow for this effect to be negligible. However, his own flowrate measurements show this is not the case. So, it is true as long as your pump can supply 36,000 L/hr per 40 W PC power (assuming you want to keep temp difference < 1°C)

[@taiiat0]

the Loop will always equalize. that's the simple answer to where you put stuff. if you put some Part immediately after another high Power part, sure it would average another degree or so higher than if it was put farther away in the Loop from it. but the important part here, is that the Water throughout the Loop will all be within a few degrees of each other, because all of the Water is in contact with itself.

[@arkie87]

@@taiiat0 your explanation is absolutely wrong. Just because the water is in contact with itself, does not mean the water temperature will always be close. For that to be the case, the flow rate must be sufficiently high and the heat sources sufficiently low. It is a simple thermodynamic calculation.

[@Sikorsky66]

hear me out, pumps are cooled by the fluid they pump, if you get a big enough pump on a small enough system, it could happen. this is obviously not going to happen in a typical water cooling build.

[@MrWoowootila]

I would like this experiment done with radiator fan speed. I would expect at a lower flow or pump speed you would see the largest difference in radiator inlet and outlet temperatures with higher radiator fan speeds. My question is the pump speed or radiator fan speed have a greater impact on dropping loop or component temperatures. I would be curious to find out if there is a limit in radiator fan speed or pump speed that noise increases dramatically and cooling effectiveness drops. You could use this information to run both the pump and radiator fans at a barely audible range under light system loads and then configure the max pump and radiator fan speeds for higher system loads.

[@NakushitaNamida]

People assume that the cpu metal has the same thermal capacity than water. The cpu is basically a bottle cap of thermal capacity and the water in the loop is a bucket of thermal capacity. Off loading a cap into a bucket doesn't really change the total thermal energy in the bucket. As long as you offload the cap often enough it won't fill up. Water flow rate is the speed at which you offload the bottlecap. That's why you can basically say that the water doesn't eat much after going through the cpu waterblock. Only flow rate matter

[@Micro-Motive]

The one variable you may have overlooked is the total liquid volume of the cooling system. One might think that having a larger liquid volume in the cooling system might also have a noticeable difference in system temperatures. Thermal mass is a well documented property used to advantage in science, engineering and construction to both retain and dissipate heat.

[@PatricksDIY]

@der8auer I guess my view would actually be what you first were speaking about, using the same loop for the CPU and GPU, if you go into the CPU first, it heats up the water, and then goes to to the GPU which would then result in higher GPU temps. I agree with you on the information about the water flow in the example you were showing, but you did not go over what was actually spoken about (using the same Rads or loop for the CPU and GPU, VS using 2 different Rads or loop for CPU and GPU. This is what I would like to understand personally. Thanks for the Video, I always enjoy them! Edit: The Flow rate as you were saying I agree on, Just not so sure about it " not mattering if its on 1 or 2 loops or rads).

[@BlazeFirereign]

People overestimate how much the water actually heats up. In a typical block, at a typical flow rate, you're adding 1-2C to the coolant as it passes through the block. Yes, that will result in 1-2C higher temperatures if you then go straight into another block instead of going through a radiator first - but does that 1-2C matter to you? It matters if you're looking for the absolute best performance possible, but for most people, going for a cleaner loop setup is preferable instead of trying to route a radiator in between each block.

[@KenneMCL]

Need to spread this video's wonderful info!

[@brudel001]

It would have been interesting to see how low can you go until it starts seriously affect the temperature. Still it seems that the most reasonable way to go is just around normal speed.

[@BRUXXUS]

Now that would be interesting!

[@Ephem13]

The most recommended setting for flow rate is as high as you can run without being annoyed by the sound. I run mine at 60%, keeps temps well within a rage that anyone would be comfortable with and quiet enough that it's overwhelmed by my fans that are usually running around 50% speed. Granted my system cooling gets very aggressive at 70C but it rarely ever goes into that territory under normal usage.

[@KyrreStalsberg]

Tale a look at Bernoullis equation and its pretty obvious that flowrate matter..

[@mrozu1337]

Looks like those l/h values are wrong. These must be l/min (litres per minute), but i have no experience in water cooling. 1,7l/h == 0,47ml/s == 2,3mm/s @16mm tubes or 21,4l/h == 5,9ml/s == ~30mm/s @16mm tubes, which seems very small. Then, 1,7l/min and 21,4l/min would result in 0,14m/s and 1,7m/s @16mm tube.

[@Mythricia1988]

The pump used in the video (D5) are usually good for 200-300l/h in an average system, so yeah I think something about the units is wrong. I'm sure the sensor is working, but something is getting lost in the unit conversions

[@aleloAT]

when i look at flowspeed discussion i just look at my motorcycle - the water pump is driven by the engine RPM - meaing the higher /faster it runs (the more heat) the faster the pump runs - hotter = faster better (for me)

[@KonstantinDankov]

Many people make that mistake. You can’t compare engine cooling with PC cooling. The temperatures in cars and bikes are way higher and the cooling systems are quite different. Also for engines you have a minimal safe temperature

[@looncraz]

That's far too simplistic of a view. The reality is very simple: the water only needs to flow as fast as is needed to absorb the same amount of heat and to dissipate the same amount of heat from the sources and sinks as that of a higher rate of flow, slower is worse, faster is just wasteful.

[@aleloAT]

@@KonstantinDankov the temperature is different yes, but the problem is the same, and the cooling systems are pretty much the same, a radiator that gets cooled by air (riding or fan), cooled water goes to the pump, coolant goes to throuh the hot components, back to the radiator - yes an engine has an "operation temperature" which you cant really undershoot anyway with the cooling and is only a "concern" the first 1-5 mins (depending on ambient temperature)

[@jake20479]

engines have a operating temperature. with rpm the flow increases.if it worked at 100% all the time then the water would never warm up. the job of an engine is to reach operating temperature, and open the thermostat. try takng your thermostat out. your engine will barely, or wont at all, heat up to operating temp.

[@aleloAT]

@@jake20479 thats literally what i wrote? " the water pump is driven by the engine RPM - meaing the higher /faster it runs (the more heat) the faster the pump runs - hotter = faster better" maybe the last part is the misunderstanding - as you ride faster, the engine gets hotter, but also incrases the RPM of the water pump (as it is driven by the engine RPM) which incrases the cooling efficiency

[@rossharper1983]

You know making this video is pointless, right? Internet warriors always think they're right even when proved wrong

[@der8auer]

never give up xD

[@rossharper1983]

@@der8auer I gave up a long time ago lol. You have more patience than me

[@andrew1977au]

@@rossharper1983 without stupid people we would have no one to laugh at

[@snawbadde2371]

So we want high flow rate through the components to be cooled and low flow rate through the radiators.That sounds like radiators should be in parallel instead of series?

[@fred_derf]

+snawbadde, writes _"That sounds like radiators should be in parallel instead of series?"_ They are. The single inlet is split into multiple (parallel) lanes through the radiator. The number of lanes being limited by the size of the radiator.

[@siliconartGAMING]

Awesome explanation, thank you

[@roberthelcher2884]

I always wondered which was better low flow or higher flow. Thank you for the demonstration. I love your amazing content.

[@PyroVulpes]

Very informative video, thank you! Also, can we take a moment to appreciate how you centered your microphone in that circle graphic on your shirt?

[@anub1s15]

had to recalculate the liter per hour to liter per minute because my head math was giving me impossible numbers but that flow rate is... odd.... extremely odd. in my loop (1x 360 alphacool slim, 1x 240 EK PE, EK RTX gpu block, EK Velocity cpu block with a D5 and a EK tuberes/top) 0.9 L/M is the LOWEST i can get (at 25% pump speed, aquacomputer usb controllable pump) what aquacomputer is telling you would be 0.36 to 0.6 L/m. that is freakishly low for a D5. (sensor i'm using to measure is the Aquacomputer High Flow USB which does have a ... wheel...stick.....thing in the path of the water flow.) with a minimum of 0.9 L/m i've found that minimal speed is enough speed, at 30% i start to hear my pump, and even at full blast which results in 3.7L/m my cpu temp drops by maybe 1-2 degree's (it's really hard to tell if it's actually helping, which for me is reason enough to run it at 25%). perhaps a clogged cpu block, it's microfins tend to be much finer then the gpu's fins resulting in it "filtering out" anything that might have gotten in the loop, atleast it actually helped you in emphasizing the point of the video ^^

[@Speak_Out_and_Remove_All_Doubt]

I have toyed with so many different cooling ideas over the years: My first was a massive chest freezer with about 200L of water (in 2L water bottles) so you have the specific heat capacity of the ice and water but more importantly you also have the latent heat of liquefaction when the ice turns to water. My second was submerging a system in either deionized water but this would be very tricky (impossible) to keep clean enough so it doesn't become electrically conductive, or submerge in mineral oil but that just looked too messy. What looks amazing is that Novec liquid but sadly it's crazy expensive. My third, was a 'closed system' liquid nitrogen. So a sealed room, 100% nitrogen gas, with condensers to create the liquid nitrogen (radiators for the condensers are outside the room to dissipate the heat). Then you have an endless flow of LN2 and no ice or condensation build-up as you have removed all water vapour in the sealed room. But clearly this is very expensive, a massive amount of work and highly impractical. My fourth, is either fully submerged or just a water loop but with a greatly reduced air pressure so that the water boils at ~60c as the latent heat of evaporation is so much bigger (better at absorbing heat) than simply relying on the specific heat capacity of water. This one is feasible but still pretty complex. My fifth option, purchasing a chilled water loop system but not only does this feel like its cheating as it's all off the shelf but also there are not many companies making these nowadays, same as sub-zero phase change units. Plus they can be very loud too. My finial option and favourite, a normal regular water loop but with a massive reservoir of water. I was thinking like a large metal oil barrel set up (you could paint it to make it look smart) so that you pretty much always have essentially room temperature water going into your system all of the time. I then took this idea one step on and thought rather than a giant water loop I could use a normal sized water loop but with the radiator inside the water of the barrel which then makes things much simpler in terms of draining the system but more importantly it allows you to have water flow in and out of the barrel. So you could use your cold water tap at ~10c degrees to easily cool your water loop, you could also dump ice in there if you wanted that extra little bit of cool for a certain benchmark or something. Anyone have any other ideas or comments on these?

[@solbot]

If you run your cooling system below ambient temperature you run the risk of condensation, it’s not just a problem for super cooling your system. I don’t know how you would reduce the pressure to achieve a boiling point ~60 C which is around the air pressure at 10km elevation. I’m assuming it would be rather bulky and expensive, maybe inside of a vacuum chamber. Possibly you could place your computer into a cold environment such as a refrigerator or freezer but I would be concerned about the coolant freezing if you go too low. Random thoughts, free advice is as valuable as what you pay for it.

[@Speak_Out_and_Remove_All_Doubt]

@@solbot The reduced pressure design would just need strong ridge tubing in a sealed low pressure loop, I think it would be doable.

[@jscrub6232]

How do you not have more sub's. The work you put in is amazing, thank you for everything you do.

[@JMUDoc]

Think of it this way: you are moving around a loop where half the time you are repeatedly being punched in the face, and the other half you are receiving medical treatment. The slower you move through the loop, the longer you receive medical treatment... but it's still going to HURT more after the "punching" phase!

[@der8auer]

hahahaha best metaphor

[@drspod]

Better analogy is a line of people carrying boxes from the CPU to the radiator. If you have twice the rate of people passing the CPU block, you can carry twice the number of packages per second.

[@arkie87]

@@der8auer your conclusion is correct, but his measurements are not: A simple thermodynamics calculation gives 1.5[K]*4.2 [J/g-K]*5.5 [g/s] = 35 W. You are putting 170 W into the radiator, and your measurements only explain 35W. The temperature difference across your CPU or radiator at steady state should be ~7 degrees C, assuming your flow rate measurements are correct. If they are, you needed to wait longer to reach steady state.

[@JusttheEdge]

How about testing 2 radiators in parallel to drop the radiator flow but keep the high block flow.

[@mhamma6560]

Doesn't matter 1km @ 50kmh = 2km @ 100kmh with constant time. Half speed at half distance still = full speed at full distance -- you've changed nothing. Higher flow rates are always better as it reduces the impact of the boundary layer that could prevent optimal thermal transfer as illustrated with slower speeds.

[@maoathanaric7862]

Oh, that would interest me too! serial vs. parallel radiators

[@jamesthomson8659]

@@mhamma6560 Except two in parallel might give more total water flow depending on the pressure /flow curve of the pump.

[@AntExe-ey5my]

I admit, I put a rad between my CPU and GPU but that was only because it made it easier for me to have the tubing runs the way I want them for aesthetic reasons.

[@cadetsparklez3300]

the one in the title doesnt make sense but I can see a reason for the alternating radiator and heat source if you are water watercooling something like ram which took very little power, the energy from the cpu or gpu will just be heating up the ram to a point further than it would have heated up outside the loop. I see this a lot and dont understand where people get the idea that energy only conducts one way the only way it couldn't make a difference is if the heat gets perfectly distributed throughout the loop as if water is a superconductor, which if so then the flow rate wouldn't matter...

[@Ephem13]

@@cadetsparklez3300 Alternating heating and cooling sources won't make a difference in loop performance, there are literally thousands of videos here on YT demonstrating that. Also note that even here, on a test loop with massive amounts of heating sources, the water temp doesn't break 40C. DDR4 RAM will operate in excess of that temp, so at no time are you heating the RAM. Also when you talk about water needing to be a superconductor, it seems disconnected from the rest of your statement. All parts of the water are spending equal amounts of time on all heating and cooling sources. If the flow rate is sufficient to keep operating temperatures well within a safe range, the difference in temperature across the loop will be less than 2C in most cases.

[@cadetsparklez3300]

@@Ephem13 in order for the heat to be spread evenly throughout all the water either the flow rate has to be infinitely fast or the water needs to be infinitely conductive. So if the water has not yet gone through the radiator it has not yet fully cooled, even if its only a small amount. No video can claim to disprove the math, plus there are tons of people showing the last component in the loop being hotter. Nowadays manufacturers git smart and you can buy fittings with more than 2 holes to run each block in parallel. Plus they also make it easier to connect every component in most cases, while before you would have to double back after each radiator, which was the main reason people said there's no point in doing it.

[@BlazeFirereign]

@@cadetsparklez3300 The point is that the heat capacity of water is so large that coolant will only heat up by ~1-2C as it passes through a CPU or GPU waterblock, and will only be cooled by ~1-2C as it passes through a radiator, in any practical loop with a realistic flow rate. Assume a fairly low flow rate of 100 L/h - that's realistic for 2-3 blocks, 2-3 med-high restriction radiators, and a D5 set to a quiet speed. Most coolants are mostly water, and water weighs 1kg per L, so it's about 100kg of coolant per hour, with a heat capacity of 4.2kJ per kg per C - meaning 4.2kJ of energy to raise 1kg of water by 1C. A temperature difference of 1C through a block means we're dissipating 420 kJ per hour into the coolant (4.2kJ/kgC * 1C * 100kg/hour), or 117 J per second (420,000 J over 3600 seconds) - i.e. 117 watts. That's about half of the power consumption of a heavily-overclocked CPU or a typical high-end GPU, so you're looking at about a 2C temperature difference before/after a typical CPU or GPU waterblock, and a similar decrease before/after a typical radiator. In effect, in any practical loop, unless you're running *a lot* of blocks and/or a very low flow rate, loop order does not matter. Furthermore: the typical loop isn't watercooling the AM, or any other low-power component, because doing so is pretty pointless in 99% of cases. If you're worried about RAM running hotter in your loop than outside of it...why are you watercooling them to begin with? Even if you're doing it for aesthetics, and even if they end up running a bit hotter as a result, the temperatures are going to be so low that it won't matter anyway. The typical loop's coolant is going to be in the 30-40C range under load.

[@fred_derf]

@@cadetsparklez3300, writes _"the energy from the cpu or gpu will just be heating up the ram"_ The water in the tubes is not going to be hotter than the RAM, so it won't be heating the RAM. _"Nowadays manufacturers git smart and you can buy fittings with more than 2 holes to run each block in parallel."_ That's a terrible idea unless you have a way to monitor flow through each loop and a way to balance the flow rates between the two loops. If you split the loop like that you will reduce (possible dramatically) the flow through which ever heat exchanger has the highest resistance (which is probably the one that generates the most heat).

[@mart446]

How would you rate the flow sensor ? im planning on purchasing two of those for separate systems.

[@EVLS10]

Order of heat source and heat sink does matter. Just not so much with a small amount of components. In order for rules/laws to be correct they must hold up to extremes. Take 100 CPUs and 100 radiators. Granted flow rate would be a challenge with normal size fittings and lines however if you ran all the CPUs first in the loop then temps would reach a critical level well before the end of the chain of CPUs and the temp would reach ambient temperature well before the end of the chain of radiators. Run that in a cpu, rad, cpu, rad etc configuration and you would have much better performance. Tldr. It does matter. Just not so much on most builds.

[@JoeBob79569]

I watched this video a few weeks ago when I just had a normal old Magicool pump and no temp sensors. I've since got an XD5 pump, and an additional water temp sensor and I've been messing around with it for a few days, and now this video makes a bit more sense to me. Probably because I have my own system for reference. I have a temp sensor at the inlet of my radiator, and one just after it (in the pump), and I'm seeing similar results to this. But what I find really interesting is that it seems like every system will have it's own perfect cooling configuration. Basically it's something like if you call the heat from your components (the heat that's transferred to a chunk of water) "H", then you have to remove that H by the time that water gets back to the pump. So basically your fans have to run at a particular speed to achieve this. But that also depends on how long that chunk of water is in the radiator, and so it also depends on the pump speed. So, it's like a fast pump speed is good for removing heat from components, but it's bad for removing heat from the water. And contrarily, a slow pump speed is bad at removing heat from components, but it's good for removing heat from the water. And of course when you have more radiators, and/or larger radiators, you can afford to run the pump a bit faster and your fans a bit slower, depending on the noise from each. I suppose the best way to go about it is to stress your components and equalise your temps. And then gradually drop your fans and pump speed until you hit a barely acceptable temperature, and then tweak the fans and pump from there to see how the temps change.

[@joshbooth9772]

With vs without a radiator!!! Lets see that test!!!

[@zushikatetomotoshift1575]

I never heard Slower flow rate will drop temps but I did hear it less stress on the life span if anything.

[@kenney0313]

Wouldn't slow moving water become saturated with hear faster? It seems like it'd perform better moving faster to wisk away heat for dissipation in the rad BEFORE the water get hot.

[@JakusJacobsen]

Why is the flow rate so low? It should be much higher with that loop. For starters the MPS400 only measures down to 40L/h with the standard calibration and a loop like that should be running closer to 200L/h or higher at that pump setting. Something isn't right there. It looks like possibly you have not imported the correct calibration for the flow meter. If you use the calibration from the MPS200 or 100 model it will read low like it seems to be there. Also, why is the pump only running at 2000rpm? 1800rpm is the minimum speed for most d5 pumps and 4800rpm is max. 65% doesn't seem like it should be only 2000rpm

[@djangoofyork]

Your observations are correct, looks like the incorrect calibration curve was used. I have the same flow meter in a few machines and for a single D5 you should be seeing between 150l/h and maybe as low as 80l/h on a really restrictive block CPU/GPU loop with something like a Nexxxos GPX at around 60%. On a dual D5 setup I see 380l/h at 100% with a 0.1-0.2°C delta temp for CPU in and out. At 140l/h at 60% this widens to 2-4°C depending on workload. Turbulence from the pump and not having a pressure relief value can also cause strange readings. I think I read in the manual or else where you should put the flow meter in-between two pieces of tube horizontally on the res return to get accurate readings.

[@jamesthomson8659]

@@djangoofyork The flow is obviously not correct. Using his numbers, the radiator is only removing 31.5 W in the Normal case and an even lower 21.5 W in the High case. The low flow case is a ridiculous 8.2 W.

[@jamesthomson8659]

How is it that the CPU is cooled better (ie lower temperature) in the High Flow case when the radiator is only removing 21.5 W while in the Normal Flow case the radiator is removing 31.7 W?

[@gregschaust4298]

Interesting video, even though it is all things that I already knew. I am surprised that you don't run dual cpu's in parallel. It works perfectly well, and looks much better. I run 2 8280 platinums in parallel and my temps are perfect,

[@merrychristmas3213]

I'm getting ready to build my first loop and am so glad I found this video. Thank you!

[@quazy1328]

It uses 2 pressure senors and a orifice in the center to cause a pressure differential and using a formula you can get the flow rate.

[@arkie87]

how does it get fluid density?

[@simona.5295]

Will this behaviour stay the same with bigger radiators like the MO-RA?

[@DaPoets]

Great video, no matter how much this is said though, people will still not understand though. Hey I sent you a DM on Instagram by the way. I hope you all are doing well.

[@xXxHarryMuffxXx]

I know you were using it to show flow but it kills me seeing aquaro's with 1 thing plugged in. Had my XT since back in the 4770/90 days most under utilized item I ever put in my system. Quality bit of kit that is for sure.

[@tuxr4z3r]

do u think that ur LED panels are finished now ?

[@jamegumb7298]

So I just get a larger radiator wit hthe betetr pumping pump for lower temps because longer in the radiator, or I let the pump go slow for better noise levels.

[@Ren-kei]

what about double radiators and double pumps?

[@LadBooboo]

I was just thinking this the other day, while messing around with the fan curve on the radiator. I was wondering if pump speed mattered but I got lazy and just set an agressive fan curve

[@o0Dan0o]

The delta-T for any component can actually be calculated relatively easily. Basically, it's power divided by flow rate and the specific heat of water (adjusting all units to match). There are cases where having a rad in between components would be helpful (4x 4090 + 350W Threadripper), as that's around a 10c rise across all components at a 230l/h flow rate (~1GPM). BTW, your flow rates seem really low to me, 37l/h is only 0.6l/m. TechPowerUp shows the typical D5 or DDC pump pushing more like 4-ish l/m of flow at 100% PMW. Which is why I assume more like 230l/h above.

[@emperorSbraz]

weird you didn't bring up laminar/turbolent flow because that's what makes high flow better. :)

[@arkie87]

not in the microchannel cpu cooler. flow there is laminar.

[@spencerwarren8302]

You never presented any proof of no difference in ordering parts, and I still believe that it matters. I would love a video presenting that theory separately! Good job on this one though

[@der8auer]

Yea it will be a 2nd video :)

[@cadetsparklez3300]

yeah.. if you are water watercooling something like ram which takes very little power, the energy from the cpu or gpu will just be heating up the ram to a point further than it would have heated up outside the loop. plus anyone with 2 of the same gpu has seen the one after the first reach higher temperatures.. I see this a lot and dont understand where people get the idea that energy only conducts one way the only way it couldn't make a difference is if the heat gets perfectly distributed throughout the loop as if water is a superconductor, which if so then the flow rate wouldn't matter...

[@goneinsane6088]

i went for 2 gpu water blocks, 360 rad, pump, 2 gpu water blocks, 360 rad, pump. I find it keeps my radeon vii's at a nice low temp.

[@wadeepperson6906]

Is the word you're looking for Venturi ?? For the smaller middle but flared opening and exit ?

[@blazbohinc4964]

Stop using "therefore". Don't you know any other words???????????????????????????????????

[@DanPellegrino486]

Go make your own tech empire then.

[@MTBScotland]

At what point did you start to see diminishing returns. Between 2000rpm and 4800 rpm you only gain 0.4 c better cooling. Should try this with a spec better D5 (Corsair have worse spec than say an EK D5)

[@MiliasColds]

i'd be interested in seeing what would happen if you forked the flow through 2x of those radiators, so the flow rate is only low in the radiator, most multiple radiator setups put them in series

[@nilz__]

It shouldn't matter in theory. The total radiator surface and total waterflow stays the same. If you put two radiators in parallel, the flowrate will be lower and cooling will be more efficient, but either radiator will only cool half of the water. In series the cooling is less effective but both radiators will cool all the water.

[@Arokhantos]

Less time spend in block and less time spend in radiator would only apply with small radiator, if you had triple Mora 420 radiator with enough pumps for enough presure and waterflow your temps would probably drop a lot cos even tho the water spends only less then 1 second in the block it will spend more time in the rads cooling down more, and yes i purely made this comment cos i wanna see such crazy video with triple mo-ra 420 radiator and crazy flow perhaps you could test it with a liquid devil and see how much you could bring down gpu temps but also hotspot temps.

[@henkblignault8255]

Hi Roman, Can you check your experimental data with some thermo calculations. I did some calcs and something seems off, I do not get to the typical watts heat dissipation for the cpu by means of using the water temps and flowrates... would be nice for you to verify some of these experiments with fundamental science. Furthermore, this experiment would lead met to think there would be an algorythm which could be developed to optimise the flowrate for maximum heat transfer from the cpu to the radiator (and air). Thanks

[@jamesthomson8659]

The maximum flow rate will give the maximum heat transfer.

[@oafkad]

Imagine being so confident in yourself that you walk into a der8auer video and think "I'm going to teach him how overclocking and water cooling works."

[@arkie87]

does he actually know fundamentals of thermodynamics and heat transfer?

[@hirronnez]

Were fans always running at the same speed ? room temperature equal over time ? did you make a graph of power consumption of the pump (and fans) compared to the actual temperature of the cpu to find a sweet spot ? Those would have been interesting things to present