I’ve been water cooling wrong for YEARS - $H!T Manufacturers Say

[Ceremony64]

The reason why using the same air for multiple radiators does work is simple and has likely less to do with those additional case holes: When cool air passes through a hot radiator, it only carries a portion of the heat away, slightly decreasing the radiator/water temperature while slightly increasing the air temperature. Unless the air moves really really slow, air and water/rad will not reach equilibrium. So while using a 2 stack radiator doesn't double cooling performance, it certainly does still improve it over a single rad. Keep adding rads for less and less improved cooling in return.

 

The slower the air moves through each radiator, the lower the additional cooling performance per radiator stack will be. The higher the temperature delta, the more effective cooling (transfer of heat) is.

 

What might be interesting, is whether these additional holes in the case are actually hurting or helping cooling performance.

These holes might cause warm air that has just recently been released from inside the case to be sucked right back, essentially forming some small invisible "air loops". It might be better to seal off the holes near the exhaust of the air to make sure there is no suction happening near it, pulling in that warmer air again. Tho, it might be even too negligible for any kind of measurement? Who knows...

 

You up for some duct tape (test)cases, LTT?

[geo3]

13 minutes ago, AnonymousGuy said:

Parallel vs. serial probably wouldn't have made a significant difference.

I would disagree there. I am going to logically assume that a parallel flow through 2 thin rads will preform similar to that single thick rad. Since if you think about it a single thick rad is essentially like a parallel flow in 2 thin rads just all within one fatter rad.

 

This is assuming that the only difference between the tested rads is the thickness and not material or fin density or some other design factor. 

[Nystemy]

13 minutes ago, AnonymousGuy said:

I can also answer this because I have temperature sensors on the inlet and outlet of everything.  Playing Fortnite (which fairly taxes the system), the biggest delta between the water going in the CPU and exiting the GPU is about 3 degrees C.  This is with a D5 pump at full speed and the water is cooled with an enormous radiator where it never "heat soaks" over time....whatever the water temperature was at the start of gaming is what it is at the end of gaming.

 

Desktops just don't have these enormous heat loads where you need tons of flow rate.

 

 

 

(The reason GPU IN is lower than CPU OUT is because there's a radiator between them).  I'm using that radiator to cool the air in the case...unusual configuration.

3 degrees is fairly huge.

With a rather slow pump moving only 10 liters a minute, we wouldn't expect a temperature difference of more than 0.71 degrees C for every 500 watts.
With a EK-D5 pump, with a "max flow" of 25 L/m, we could expect a temperature rise of about 0.287 degrees C for every 500 watts.
(one can also buy an aquarium pump for like 50$ and get something that moves 50+ L/m, but that is overkill.)

At a 3 degree delta, I would worry about gunk in the lines, or a pump struggling to just keep the water moving.
 

[AnonymousGuy]

1 minute ago, geo3 said:

I would disagree there. I am going to logically assume that a parallel flow through 2 thin rads will preform similar to that single thick rad. Since if you think about it a single thick rad is essentially like a parallel flow in 2 thin rads just all within one fatter rad.

 

This is assuming that the only difference between the tested rads is the thickness and not material or fin density or some other design factor. 

Yes, two thin rads is roughly equivalent to one thick rad (although thin rads usually have higher fin density than thick rads so it somewhat cancels out at the expense of a lot more pressure needed).  But whether you run the water through them in parallel or series is where it really won't matter much.

 

30C -> Rad 1 -> 30.5C -> Rad 2 -> 31C    vs.   

30C -> Rad 1 -> 30.5C   +   30C -> Rad2 -> 30.5C.

[AnonymousGuy]

1 minute ago, Nystemy said:

3 degrees is fairly huge.

With a rather slow pump moving only 10 liters a minute, we wouldn't expect a temperature difference of more than 0.71 degrees C for every 500 watts.
With a EK-D5 pump, with a "max flow" of 25 L/m, we could expect a temperature rise of about 0.287 degrees C for every 500 watts.
(one can also buy an aquarium pump for like 50$ and get something that moves 50+ L/m, but that is overkill.)

At a 3 degree delta, I would worry about gunk in the lines, or a pump struggling to just keep the water moving.
 

Eh, the water temperature directly/linearly goes towards the component temperature, where 3C water temperature increase on a component with a TJmax of 105C is really nothing.    The GPU never goes above 60C anyways and the CPU even at max load never goes above 90C so I wouldn't gain anything really chasing a lower delta T.

 

You'd also have to know the volume of liquid in the loop to calculate how "long"  it takes for water to circulate through the entire loop and exit based on the flowrate.   Rough estimate with 3 feet of tubing is about 0.1L of water in the loop between the start and exit.  At 25L/m that's about 250mS for the water to move through the entire loop.  I dunno what's the math on how much heat it takes to heat 0.1L of water by 3 degrees C in 250mS?  (which honestly seems a little fast, it probably takes a full second for an airbubble to move through my cpu + gpu, 25L/m is probably assuming zero resistance)

 

What volume did you use to get 0.71 degrees for 500W?

[Nystemy]

36 minutes ago, AnonymousGuy said:

Eh, the water temperature directly/linearly goes towards the component temperature, where 3C water temperature increase on a component with a TJmax of 105C is really nothing.    The GPU never goes above 60C anyways and the CPU even at max load never goes above 90C so I wouldn't gain anything really chasing a lower delta T.

 

You'd also have to know the volume of liquid in the loop to calculate how "long"  it takes for water to circulate through the entire loop and exit based on the flowrate.   Rough estimate with 3 feet of tubing is about 0.1L of water in the loop between the start and exit.  At 25L/m that's about 250mS for the water to move through the entire loop.  I dunno what's the math on how much heat it takes to heat 0.1L of water by 3 degrees C in 250mS?  (which honestly seems a little fast, it probably takes a full second for an airbubble to move through my cpu + gpu, 25L/m is probably assuming zero resistance)

 

What volume did you use to get 0.71 degrees for 500W?

The total water volume in the loop is actually not important when calculating how much temperature will rise for a given load at a given flow rate.
Ie, we can connect one side of the cooler to the ocean and it will heat up just the same amount if the flow speed remains the same.

(even inlet temperature won't matter from the point of our calculation, in real life it would matter due to lower die temperature leading to lower power leakage and thereby a cooler chip. But then we have changed the amount of power we are trying to cool. As long as power remains the same, inlet temperature doesn't matter.)

The specific heat capacity of water is 4186 Joules per liter.
1 watt is 1 Joule per second.

If we have a load of 500 watts, then with no flow, we will heat 1 liter of water with 0.1194 degrees C per second.
If we have a flow speed of 1 L/s then we exchange that 1 liter volume every second. Ie, the temperature rise for the water is thereby 0.1194 degrees C at a flow rate of 60 L/m. Regardless of the water volume in our cooling block. (I can add in the very marginal addition due to thermal conduction through the water, but this is so small that you won't notice, and likely on par with resistive heating due to flow impedance from the cooling fins themselves.)

Also, the flow speed of a pump can be dependent on a few factors, so the flow speed over flow resistance graph can have a non linear relationship.
In my own experience, most pumps don't see much difference between pumping water through a 1 meter long hose, or a 5 meter long one. Flow speed will be largely the same.

In short, I would expect a D5 pump to give more than 10 L/m in most computer loops, a long loop might get down a bit lower, but I still wouldn't expect to see a 3 degree delta over a GTX 1080TI. (it might work, but it is indicative of something not being right. It's a bit like a car that squeals every time one breaks, yes it works, but soon you won't have breaks...)

Edit:
Though, actually looking into the graph you provided, the GPU doesn't have a 3 degree temperature delta over it. It has about a 1.25 degree delta. Much more manageable and realistic, still a bit on the high side to be fair, but at a low pump speed it is within reason.

[Zberg]

if thats truly what corsair said to them, they were at least kind of right on the stacked scenario (but not completely) but nowhere close on the other scenario (based everyones water cooling setup, lol)

 

In the stacked scenario, I look at it like SLI on 2 GPU.  You will probably get some benefit, but nowhere near scale to double the initial rad because it it directly receiving ONLY the hot air right from RAD 1.  I also dont think Rads are truly as efficient as they think they are and that the air is probably lower temp than they say.

 

In the other scenario they talk about (big intake rad, smaller exhaust rad on other side of case)...I mean literally 1000's of people set up their rads this way and have seen the results.  The air is mixing around, this isnt vacuum sealed, and the rads arent as efficient as they think.  My current set up I went from only intake 360 rad at front, to adding a 280 at the back/top that is "just getting hot air" and "blocking airflow" from leaving case...and surprise temps dropped hugely.  I mean how many cases are set up this way that they would think this wouldn't work? 

[Ubersonic]

I love that Corsair are trying to peddle a myth that was debunked by the water-cooling enthusiast scene almost FIFTEEN years ago, the balls on their PR dept

[Trik'Stari]

I bought a Caselabs SMA8 Magnum, and I'll be goddamned if I fail to have a least two radiators.

 

I don't care if it adds anything, it looks kewl.

[Cedric from la Reunion]

I everybody!

In my opinion Linus is right! You do with the place you have in your case and also with the price you are paying for. 

Also you didn't take into account the air around you. At that time for me in La Réunion island ( a french island near Madagascar) it's 5am and it's already 24C degrees, in the day it will be around 30C degrees. 

So every solution to decrease the temperature is good even if it's not the must do one.

 

[Tamesh16]

I assume this should all break down to a Cross vs counter current heat exchanger design, i assume that is what Corsair is trying to say, 

because by using counter current heat exchange you maximize the delta T and thus heat being exchanged, obviously more surface area is better but if you dont use that surface area effectively you wont get the best perfomance it can offer, as such try to run your radiators in a counter current design and it should give more performance over a cross current layout 

I understand why the people at corsair have been showing CFDs as well, because as someone who designs CFDs, they do work, they have to inorder to design better heat exchangers, otherwise the word of linus isnt enough to precisely manufacture radiators at its maximum performance with minimum cost, 

also heat exchange depends on many fluid properties other than delta T thats why they run CFDs and thats why they are so important to heat exchanger design

[caincha]

I do agree with the 'you gotta work with the space you've got' however as many videos as I've seen here there is always a better way to do it and the chosen one is usually the one that can be done quicker.

In the computer used as testing bench for instance the rads should be placed in the front and in the back both pushing air inwards and fans on top pushing the air out. Maybe even fans at the bottom pushing more air inside to help direct the airflow upwards but that could blow A LOT of dust inside so… Anyway I'd bet that would get even better results.

I do appreciate the fact that they tested the same theory in the Minecraft server with stacked rads. For that alone I clicked like on the video.

[LmnSour]

I mod'ed a used NZXT H630 Ultra Tower case that I got for check with x2 140mm holes on the front panel and three on the top panel.  I have a 280mm radiator on the front (intake), a 240mm radiator on the bottom of the case (intake), a 360mm radiator on the top (intake), and a 140mm radiator on the back vent (intake). I cut a 250mm hole on the side panel and mounted an old Thermaltake fan (from the old Armor series case) to run as exhaust.  Best ambient cooling performance I've seen. 

 

An older picture.  I've since cleaned it up and removed some of the extra fans.  It started as a gimmick, became a test platform to help pick out a new case, but I don't think there is a case out there that will perform better than this one so I decided to keep it (but under my desk :D)

 

[Nystemy]

So re-watching the video and pausing more often than a sane person would, to note down the various statements made is a rather useful exercise.

Then the whole video starts off with a total misunderstanding of what Corsair states.
Corsair did indeed state that "adding a second smaller radiator for the exhaust would increase air resistance and likely negatively impact cooling performance in the Hack pro."

Mainly since the output radiator were smaller than the input one in a fairly poorly ventilated case.

 

Linus and the team then tests a case where they start with a small radiator on the front, then tests adding in a second much larger radiator at the top. (In a more well ventilated case as well...) Ie, the test doesn't even remotely test the scenario Corsair commented about.

Yes, Linus and the team added a second "exhaust" radiator. And Corsair's comment were that the second "exhaust" radiator in the Hack Pro build were likely more restrictive then the much larger intake one. That it were an exhaust radiator weren't important. It were important that it were more restrictive to air flow.

 

The test would have been more fair, if Linus and the team started with the large exhaust radiator, and then added the smaller intake one. But since the case were better ventilated then the system Corasir commented about, then there were other ways for air to easily enter the case, ie not creating as large of a pressure difference compared to ambient. (In the Hack pro, it would have lead to a positive pressure, in the test case it would lead to a negative pressure, but in both cases, the larger radiator would see less airflow.)

 

 

Then we have the Minecraft server build.

There the information provided in the video about Corsair's statement doesn't even state that there were a problem with adding more radiators, but rather how one could optimize the air paths to and from the radiators. As to get more efficient utilization of the radiators.

 

Now stacking 3 radiators so that the air flows in series through them will work, and will exchange more thermal energy between the air and water.

 

But a radiator has a fairly restrictive air path compared to an open space.

Generally when it comes to radiators, we want as cold air as possible to flow through them.

Because the amount of energy they can exchange is directly proportional to the delta temperature between the radiator and the air flowing through it.
 

So here is an image of 3 different implementations for a multi radiator 2U server:

 

Now, no radiator will heat the air to the liquid temperature of the loop.

But with each radiator the delta temperature will be lowered by X %. (The exact value of X is dependent on the radiator you have, and this can be calculated if you have knowledge of air speed, loop temperature, and the air temperature before and after the radiator, loop flow speed can largely be ignored if sufficiently high. (like 5+ L/minute for most computer radiators since the temperature variation over the radiator will then be fairly marginal compared to the delta between air and rad.))

So lets say that these radiators are able to exchange 20% of the delta. Ie, if the delta between the air and rad is 10 degrees, the air will be 2 degrees warmer after passing through the radiator.

 

Then lets say we have a loop temperature of 45C and an ambient of 25C.

Then both mine and Corsair's implementations would see a 4 degree C temperature rise for the air.
LTT's implementation would see a 4 degree rise for the first radiator, a 3.2 rise for the second and a 2.56 degree rise for the third.

Since every radiator has its own set of fans, then we can largely expect air flow through each rad to be roughly the same in all implementations.
This means that LTT manages to take a volume of X air per second, and heat it with 9.76 degrees.
Corsair manages to get to 4 degrees increase for twice the volume.
And mine is also 4 degrees but for three times the volume.

Now since we need the same amount of energy to heat X volume of air by Y degrees as we need to heat half of X volume by 2x Y degrees, then we can convert both mine and Corsair's results to be comparable to the LTT solution.

Ie, Corsair gets to 8 degrees, and I get to 12.
While LTT is at 9.76
For using the same amount of radiators as I do, LTT gets only 81.3% as much heat out of their system. (Ie, they either need more fan speed, or can't run as power hungry components.)

Now, Corsair gets beaten by LTT here to be fair, but if we rerun the numbers but say that our radiators are 50% efficient.
Then LTT gets 10 + 5 + 2.5 = 17.5
While Corsair gets 10 + 10 = 20
And I get 10 + 10 + 10 = 30

If the radiators are 75% efficient.
Then LTT gets 15 + 3.75 + 0.9375 = 19.6875
Corsair gets 15 + 15 = 30
And I get 45....

So its a question of how good those extra thick radiators are at exchanging heat, if they do a splendid job at it, LTT has the worst performance.
If the radiators are shit, (something I would be surprised by) then LTT gets fairly good performance, but still not optimal.

Though, I do have to say that I have oversimplified things here, its a rule of thumb that is off by a bit. (I after all don't want to make this comment 3x longer and 20x more boring to read.)

Also, yes, removing a radiator from the LTT implementation will lower performance. (just pointing this out since it were mentioned in the video, yet another misunderstanding of what Corsair stated.)

[DJ_Stag]

23 hours ago, ColinLTT said:

Apparently stacking radiators is bad -- According to the brains over at Corsair, so we built a water cooled gaming rig to test their theory.

 

 

 

 

Buy Intel 9900KS
On Amazon (PAID LINK): https://geni.us/RyZyZQ5

 

Buy Corsair Vengeance RGB RAM
On Amazon (PAID LINK): https://geni.us/uxWIG
On Newegg (PAID LINK): https://geni.us/3HfM0Ii

 

Buy EK Reservoirs
On Amazon (PAID LINK): https://geni.us/YubShVJ
On Newegg (PAID LINK): https://geni.us/kPLy

 

Buy EK Coolstream SE 240mm
On Amazon (PAID LINK): https://geni.us/bZQBl
On Newegg (PAID LINK): https://geni.us/VUjAIu

 

Buy EK Vector RTX 2080Ti
On Amazon (PAID LINK): https://geni.us/aeSlS
On Newegg (PAID LINK): https://geni.us/axcLVv2

I think the mass of water is what make 2 rads more effective. as a test creating a cpu cooler which is just a vessel sitting on the cpu itself (some engineering would be required here), then doubling the capacity recording the temps of the two would make an interesting video, obviously there is a threshold level of 100 degrees and the first would need to be well under 100 degrees c. Your engineering videos always make for some very interesting viewing :-)

[Garns]

Okay, its 1am and I fired up MS Paint to sort this out.
You're welcome.


Maybe Corsair isn't taking into account coolant flow direction, or the video wasn't in depth enough, because "Stacking radiators" is exactly how heat exchangers work, this diagram will show you why. 
Ideally i think all radiators would draw in outside air, not "used" air. This would be the best solution, which may be what they're getting at.

Edit: Arrow colour indicates temperature, in case that's not obvious.
Nystemy's post did an excellent job of delving into things. Nice one.

[Nystemy]

1 hour ago, Garns said:

Okay, its 1am and I fired up MS Paint to sort this out.
You're welcome.


Maybe Corsair isn't taking into account coolant flow direction, or the video wasn't in depth enough, because "Stacking radiators" is exactly how heat exchangers work, this diagram will show you why. 
Ideally i think all radiators would draw in outside air, not "used" air. This would be the best solution, which may be what they're getting at.

Edit: Arrow colour indicates temperature, in case that's not obvious.
Nystemy's post did an excellent job of delving into things. Nice one.

The rather immense thermal mass of water means that each rad would only see a water temperature drop of only a few tenths of a degree C at most if we are dumping hundreds of watts into each of them. Air on the other hand will quickly heat up far faster at these amounts of power.
Water has a specific heat capacity of 4186 J per degree C per liter, while air is at 1 J per degree C per liter. (So they are about 3.6 orders of magnitude away from each other.)

Counter flow heat exchange technically still does something, but shaving off the last few tenths of a degree on the delta isn't really all that spectacular when one can double or even trippel the effective temperature delta of the whole system by just moving how the rads are placed and guiding the air with some baffles.

(Yes, I know that an earlier post of mine did state that counter flow heat exchange could have en effect, and it does, but actually running some numbers on it shows that it is laughably marginal at best.... It only starts to matter if we have far higher power loads. (talking 100+ kW Think power plant cooling, not CPU.) Or if we have similar thermal mediums, like an air to air heat exchanger for an air conditioning system. (Or a water to water heat exchanger.))

[moriz1]

flow direction doesn't matter much in terms of watercooling computers - the heat sources are too small, the temperature deltas are not great enough, and the thermal mass of water is too great. what will end up happening, is that the coolant temperature will reach some equilibrium, and proceed to stay at that temperature throughout the entire loop.

 

effectively, a stacked radiator will perform pretty similarly to a single thick radiator of the same total thickness, assuming the same number of fans are used.

 

however, stacked radiators suffer from massive pressure drop caused by the air flow restrictions introduced by the radiators. you can counteract this by running the fans faster.

 

HOWEVER, doing so would also reduce the maximum cooling capacity of the loop, since since the loop's cooling performance is dependent on the delta temperature between the coolant and incoming ambient air. or in other words, the hotter the coolant, the greater the maximum cooling capacity. this is why car radiators are relatively "small" compared to the heat source that they have to cool: as long as the coolant doesn't start boiling off, allowing it to run hot increases the amount of maximum cooling the small radiator can generate.

 

in an ideal setup, Corsair's proposed air baffle setup WILL likely cause better cooling, since each radiator surface gets their own fresh air intake, and each exhaust are isolated from each other. the second radiator does have an airflow restriction due to the smaller opening, and the first radiator would have to generate additional pressure because its fans have to pump air through a smaller exhaust.

 

as for the video itself:

 

Linus basically tested the wrong metric. what he should've done, is test the coolant temperature of a stacked radiator setup against that of a serial setup, both in 1-intake-1-exhaust and 2-intakes configurations. keep the same radiators and the same fans, at the same RPMs. the best solution is the one with the lowest coolant temperature after reaching steady state. test at multiple RPMs as well.

 

the test shown in the video is pretty meaningless.

[Nystemy]

2 minutes ago, moriz1 said:

in an ideal setup, Corsair's proposed air baffle setup WILL likely cause better cooling, since each radiator surface gets their own fresh air intake, and each exhaust are isolated from each other. the second radiator does have an airflow restriction due to the smaller opening, and the first radiator would have to generate additional pressure because its fans have to pump air through a smaller exhaust.

Everything went so nicely up until this point...

It is indeed true that pushing air from one large unrestricted opening down to a smaller unrestricted opening does need additional work to either increase air pressure or speed through the smaller opening. But that isn't the case there.

A radiator isn't an unrestricted opening, its a large fairly restrictive mesh. Ie, a radiator can have the same air resistance as an opening that is far smaller than the radiator's total size.

Ie, the opening to the second radiator, and the exhaust from the first aren't majorly increasing air resistance in practice. (It will still increase, but this is very very marginal to the air resistance imposed by the radiators themselves.)

[moriz1]

5 minutes ago, Nystemy said:

Everything went so nicely up until this point...

It is indeed true that pushing air from one large unrestricted opening down to a smaller unrestricted opening does need additional work to either increase air pressure or speed through the smaller opening. But that isn't the case there.

A radiator isn't an unrestricted opening, its a large fairly restrictive mesh. Ie, a radiator can have the same air resistance as an opening that is far smaller than the radiator's total size.

Ie, the opening to the second radiator, and the exhaust from the first aren't majorly increasing air resistance in practice. (It will still increase, but this is very very marginal to the air resistance imposed by the radiators themselves.)

are you implying that i'm wrong?

 

i said that the smaller opening and exhaust will cause airflow restrictions, but didn't say anything about whether they'd be relevant... because (AFAIK) nobody has gotten around to testing it. what you wrote doesn't confirm nor deny anything that i wrote.

 

as for the restrictive nature of radiators vs smaller intakes/exhausts, we actually don't know at this point which end up being more restrictive, hence the need for testing.

[Nystemy]

7 minutes ago, moriz1 said:

are you implying that i'm wrong?

 

i said that the smaller opening and exhaust will cause airflow restrictions, but didn't say anything about whether they'd be relevant... because (AFAIK) nobody has gotten around to testing it. what you wrote doesn't confirm nor deny anything that i wrote.

 

as for the restrictive nature of radiators vs smaller intakes/exhausts, we actually don't know at this point which end up being more restrictive, hence the need for testing.

In engineering, one should generally only look at effects that are relevant.
Also, testing of how large an open duct needs to be to not restrict air flow to a radiator has been done multiple times. (primarily by car manufacturers)

If the opening is about 1/4 the area of the radiator, it won't have any meaningful impact on air flow. As in less than a handful of % more restrictive than a radiator sitting in a wall moving air from one chamber to another. So the server itself most likely already have more air resistance behind the radiators then what our additional duct will add.
And looking at the server in question, we can reasonably get more than a quarter as an opening, so this shouldn't even remotely have a noticeable impact.

Especially compared to the benefits of giving each rad its own air flow.
Even if it were noticeably more restrictive, we could still reasonably expect better performance.
Just take the numbers from my earlier post above where I calculated the expected thermal performance of three different implementations.

If we say that Corsair's and my solution have 20% less performance due to increased air resistance, then the LTT solution still gets beaten by a fairly large margin regardless. (And we wouldn't even remotely expect such a stark increase in air resistance.)

[moriz1]

26 minutes ago, Nystemy said:

If the opening is about 1/4 the area of the radiator, it won't have any meaningful impact on air flow. As in less than a handful of % more restrictive than a radiator sitting in a wall moving air from one chamber to another. So the server itself most likely already have more air resistance behind the radiators then what our additional duct will add.

 

do you have any reference material that supports this? i tried a quick google and wasn't able to find anything.

 

27 minutes ago, Nystemy said:

And looking at the server in question, we can reasonably get more than a quarter as an opening, so this shouldn't even remotely have a noticeable impact.

 

i don't think that's the case. in one of the pictures from a previous post (which looked like a still from one of linus's videos), the gap available for such a baffle would be much smaller than 1/4th the length of the radiator, more like 1/10th. at that point, it might be worth testing just how restrictive the opening really is.

 

the baffle layout also uses two radiators, vs the triple-stack that Linus is currently running. would two radiators with better airflow beat three radiators in a stack? again, we won't know until we test.

 

i'm disregarding your angled radiator setup because it looks completely impractical due to space constraints. not to mention, it appears to be complete disregard the fans. throw those into the mix, and the added thickness would cause them to be almost parallel to the sides of the case, completely eats up any possibility of airflow to the second and third radiators (not to mention the loss of pressure due to the almost 90 degree turn that the airflow would have to take... twice, on each radiator).

[Nystemy]

1 hour ago, moriz1 said:

do you have any reference material that supports this? i tried a quick google and wasn't able to find anything.

 

Here is a diagram explaining friction losses in a duct, ie air resistance of a duct of a given cross sectional area at a given flow rate:

 

1 hour ago, moriz1 said:

i don't think that's the case. in one of the pictures from a previous post (which looked like a still from one of linus's videos), the gap available for such a baffle would be much smaller than 1/4th the length of the radiator, more like 1/10th. at that point, it might be worth testing just how restrictive the opening really is.

A rack server is between 435-450mm wide, the radiator has 4 80mm fans, so the radiator is about 320mm long + end caps of about 20mm.
Ie, the gap that is left is still about 70-100mm or about 1/5 to 1/3 the area of the radiator itself. So I don't have a clue how you get that to 1/10th.
 

1 hour ago, moriz1 said:

the baffle layout also uses two radiators, vs the triple-stack that Linus is currently running. would two radiators with better airflow beat three radiators in a stack? again, we won't know until we test.

This can be calculated, and that I have done a few posts up. (And the answer is, two radiators are most likely more efficient than 3 stacked ones.)
 

1 hour ago, moriz1 said:

i'm disregarding your angled radiator setup because it looks completely impractical due to space constraints. not to mention, it appears to be complete disregard the fans. throw those into the mix, and the added thickness would cause them to be almost parallel to the sides of the case, completely eats up any possibility of airflow to the second and third radiators (not to mention the loss of pressure due to the almost 90 degree turn that the airflow would have to take... twice, on each radiator).

This better?

Each duct here is only 35mm.
For a combined area of 105mm.
So at an airflow of 1600 liters a minute. (Ie, server grade fans on full tilt and then an ample amount more. Ie, you won't have this amount of airflow ever in any server.)
Then the ducts will only create a pressure difference (ie air resistance) of 2pa/m. And that is very very little considering that these ducts are only about 0.3 meters long.
So the fans won't consider these ducts as even the slightest bit restricting.

The radiators on the other hand can have air resistance figures that literally dwarfs the air resistance of these ducts.
A primary component of air resistance is surface drag, a radiator's main goal is to offer as large amount of surface area for the air to flow over as possible, ie it gives a very high amount of surface drag as a result.

Anyone who has ever used a radiator can though already say that pushing 1600 liters a minute through a 80x320 radiator will require some fairly massive fans to say the least.
That radiators have a high air resistance is fairly obvious to anyone who has used these devices.

Edit:
Also, if one wants to make a comment that only a third of those 1600 liters would go through the radiators, then still, that is an impressive amount of air, and we would reasonably expect far lower air flows than that. And a resultant lower air resistance in our ducts as well.

Edited May 15, 2020 by Nystemy

[moriz1]

2 hours ago, Nystemy said:

A rack server is between 435-450mm wide, the radiator has 4 80mm fans, so the radiator is about 320mm long + end caps of about 20mm.
Ie, the gap that is left is still about 70-100mm or about 1/5 to 1/3 the area of the radiator itself. So I don't have a clue how you get that to 1/10th.

the radiators used in that case is 365mm wide. so it's actually about 15-20% of the radiator surface. add in some losses due to tubing runs and whatnot, and 1/10th isn't too far off.

 

3 hours ago, Nystemy said:

This can be calculated, and that I have done a few posts up. (And the answer is, two radiators are most likely more efficient than 3 stacked ones.)

you didn't calculate a damn thing - you guestimated. whether any of it translates to results in the real world is anybody's guess at this point.

 

as an example:

 

my personal rig is pretty similar to Linus's, but much jankier: i'm using a modified Rosewill 4U server case with the drive cages drilled out, with two 360mm radiators (largely mounted using packing foam, tape, and prayer). i made an air intake baffle that runs on the bottom of the case to a gap between the two radiators, and the fans on each radiator blows away from each other, drawing the intake air to the front and back.

 

exhaust air are separated from the intake through baffles, so they don't mix. half of the exhaust is set out to the front, so there's less chance of warming up the internal components.

 

layout-wise it's pretty close to Corsair's, where both radiators get fresh air intakes.

 

performance wise? well, its coolant delta T is somewhere around 22C to ambient (which is 20C), with fans running at 1000RPM (6x Noctua NF-A12x25). in custom water cooling, this is generally considered to be pretty terrible.

 

^ this is why it is important to test your hypotheses, instead of relying purely on fancy drawings and guestimates. we won't know which is actually effective until we test.

 

3 hours ago, Nystemy said:

This better?

no. this is about as bad as i imagined it. not to mention, you've made no provision on how to run tubing to this monstrosity, which would reduce airflow even further.

 

i know this is a neat design and all, but i seriously doubt it'll work in the real world, nevermind its obvious space constraints.

[Nystemy]

10 minutes ago, moriz1 said:

the radiators used in that case is 365mm wide. so it's actually about 15-20% of the radiator surface. add in some losses due to tubing runs and whatnot, and 1/10th isn't too far off.

4 x 80mm + 2 x 20mm = 260mm
Being off by 5mm isn't a major issue here. (secondly, the ends are also tapered.)
Also, my estimate were from 70-100mm.
A 435mm wide case with a 365mm long radiator in it leaves guess what? 70mm...
70 / 365 = 0.19178 Ie 19%.
Except, the ends are tapered, so its in fact closer to 20-21% in practice. And that is if the case is 435 mm wide.

The case can just as well be 450mm wide, that gives us another 15mm, and brings up the percentile to about 24%
(My estimate for 1/3 were though based on the possibility of the end caps being smaller. But this isn't the case. (I didn't feel like searching up the exact model of radiator, since +/-5 mm isn't a major issue.))

So 1/10th, is a fair bit off, when actual measurements gets it to double that in a narrow rack mount case.
But that isn't important, since even then, its still a fair bit of room.

 

12 minutes ago, moriz1 said:

you didn't calculate a damn thing - you guestimated. whether any of it translates to results in the real world is anybody's guess at this point.

I guess you haven't done any engineering in your life.
I don't make "guestimates", I make calculations of what can reasonable be expected in real life, give or take a few percent. You will have larger variations due to fan quality at that point. (and yes, I can add that to the calculations as well.)
Yes, I didn't calculate air resistance of the radiator, mainly since we know that it will be in the low Pa/L/minute. Ie, far higher than what the duct would create.
But I did calculate the air resistance of the duct, since that is the more important part when it comes to answering the question, "does the duct restrict airflow". And the answer were, "not meaningfully compared to the radiator itself."

46 minutes ago, moriz1 said:

you've made no provision on how to run tubing to this monstrosity, which would reduce airflow even further.

A 12-16 tube wouldn't majorly impact air resistance. And one can temporarily go down to a thinner tube for this connection if one is bothered.
Do note, the ducts aren't even close to restrictive. So a tube isn't going to make it all that much worse to be fair.

And in regards to your own cobbled together solution:

22 minutes ago, moriz1 said:

i'm using a modified Rosewill 4U server case with the drive cages drilled out, with two 360mm radiators (largely mounted using packing foam, tape, and prayer).

The description of your own case sounds more like a thrown together mess to be fair. Sure you aren't blocking off the rads?

21 minutes ago, moriz1 said:

half of the exhaust is set out to the front, so there's less chance of warming up the internal components.

Sounds like the perfect explanation of recirculating airflow to be fair.... Don't worry about the internal components, everything on a motherboard can take 40 and even 50 degree C air. HDDs might be a bit iffy about it, but even then, have you checked the internal case temperatures of most desktop computers? 30-45 C isn't unreasonable, and some goes to 55-60....

23 minutes ago, moriz1 said:

layout-wise it's pretty close to Corsair's, where both radiators get fresh air intakes.

Are you sure? Each turn the air needs to make require an increase in air pressure for a given flow rate. (Yes, a duct with a turn has a larger pressure drop than the same length of duct without a turn. (And I am not going to go into the whole science of turning air more efficiently in a duct, that is literally a whole field in itself.)) And an increase in needed air pressure for a given flow rate, is what we call higher air resistance.

 

Sure, you might have a poorly put together thing that on the "surface" might look "similar".
Just like Linus and his team made a test to "disprove" Corsair's statement about the hack pro. (It weren't even close to actually being a proper test.)

In the world of engineering, there is many small details that one can easily miss if one doesn't know what one is dealing with.
And such small details can make a huge impact on the end result.
Call me a boomer if you want. But I at least do engineering professionally.

You might have the opinion that one can't calculate expected performance before hand.
And you might have anecdotal evidence supporting your opinion.
I am likely not going to change that opinion of yours.