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

[GDRRiley]

2 minutes ago, sacharja said:

Does anyone know what case was used there?

 

Am using a *fanless* PC and looking for an open / mesh case that supports natural convection, has enough space for my big passive cooling blocks and some 5,25" bays to update usb front connectors in the future. Seems that the case in this video would suit.

for fanless, I'd just go with a test bench/wall mount case

[masamoto]

Now I am just wondering if this is really some Official Corsair response. Or just someone who things they know what they are doing and try to prove someone else wrong.
If this is their way of proving a point and not really following the scientific method. They lost mayor points with me.

[berkut]

I have to say I don't think this test was up to LMG standards Corsairs criticism was about reusing air from one radiator to another your test planly did not do this. in the server with 3 radiators there was no place for the second or third radiator to get any fresh air. the test you did was in a open case the second radeditor had easy access to fresh air from the back of the case.

[Egad]

There seems to be two different things.  The first is the Hackintosh doesn't have "stacked" radiators in that the air doesn't go directly from Radiator A to Radiator B, it goes Radiator A to the internal volume of the case to Radiator B.  I agree with LTT there, the air will still pull heat of the rear radiator on its way out.  In general having one big radiator used on exit, so that your airflow is best, that is:

 

Ambient enters case -> picks up a minor amount of heat off the VRMs and such -> crosses large radiator and pickups CPU and GPU heat -> exits case

 

But as Linus says you gotta work with what fits in the case and the delta between that model and the one they used is probably nominal.  The cases just aren't that airtight and he's right about that/

 

I think Corsair is much closer to be being right about the server build and the direct stacking though, especially once you put the server in the rack and surround with a bunch of other servers kicking out heat.  Nothing LTT did in the video really addresses that minus the one comment about Jake tried removing a radiator, but no clarity on to if this was just done on a table or if the server actually went back into the rack.  To properly address that point LTT should actually 3D the proposed baffle, put the server in the rack and benchmark it during the workday (so while the other servers are also under normal workload and raising ambient temps the room).

[Nystemy]

3 minutes ago, masamoto said:

Now I am just wondering if this is really some Official Corsair response. Or just someone who things they know what they are doing and try to prove someone else wrong.
If this is their way of proving a point and not really following the scientific method. They lost mayor points with me.

My guess that some points were lost in communication.
And looking at the pictures provided by corsair, it is rather evident that they say that there is situations where one can get more optimal thermal performance if one puts the radiators in parallel.

Ie, the radiator is a fairly restrictive air path. So we can have an inlet on the case that is covering a much smaller surface area, and use baffles to guide the air around the radiators while providing peak differential temperature over both radiators. Now this isn't even brought up in the video other then a subtle line in Corsair's picture and one line of text that isn't even readable if one watches the video at a low enough resolution.

So to a degree, we are shooting the messenger while not listening to the actual message. 

[cowsgomooo]

Long time viewer but I made an account just to say this - this video is just FUD. Writing an 8 minute video with the conclusion "it depends" with little to no mention of any of the properties that affect cooling such as thermal equilibrium, air velocity, air pressure, and that doesn't address the the initial claim about stacked IE doubled thickness radiators doesn't inform the audience of anything new. The graph from this 2009 forum post explains things more clearly: http://www.xtremesystems.org/forums/showthread.php?220874-More-Radiator-Sandwich-testing#close.

 

Logic will get you to the same conclusion: cooling relies on temperature difference and air volume to cool. If the first radiator in a setup efficiently cools, then the temperature of the air leaving the first radiator will be almost ideal, meaning the differential temperature used to cool in the second radiator will be greatly reduced. Thus the second radiator will have little impact as Tdelta of radiator 1 >> Tdelta of radiator 2. But by introducing open airflow into the case, then temperature of the intake of rad 2 becomes a function of Tout of rad 1 and Tambient. This is why the experiment showed the 2 radiator loop worked better. That is why in general it will work. But addressing the claim of radiators specifically, without quantitative results showing the effect of increasing Tdelta on the second radiator, we cannot make meaningful conclusions about the cost effectiveness about 2+ radiators.

 

LTT has produced many useful videos, but this "scientific" experiment is certainly not one of them.

 

 

Edited May 13, 2020 by cowsgomooo
more explain

[jasonvp]

Random slag towards Corsair:  the first mistake made in this video was listening to anything Corsair has to say about water cooling.  Unless they come out and say, "Buy someone else's stuff!" ... best to just not listen to them.

[ninbura]

Having a hard time understanding the true purpose of this video.

Corsair reached out saying that directly stacking radiators results in "negligible to no improvement" and then LTT proceeded to test a completely normal radiator setup in which there is plenty of space / free flowing air in-between radiators. At the end of video, Linus mentions they ended up actually testing the scenario in question off camera by removing one of the stacked radiators in the Minecraft server to find that temperatures increased by 5c, which I'd think would considered a "negligible" difference when taking in account that said radiator was 60 millimeters thick with a 320 millimeter foot print.

There is an obvious bottleneck there when stacking those radiators directly which resulted in very little improvement, would be cool to see a video actually testing the scenario in question.

[Grabhanem]

For everyone complaining about them not testing the stack configuration Corsair was talking about, didn't they mention Jake taking one of the radiators off the server "brick" and seeing a rise in temperatures?

[SolarNova]

1 hour 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.

 

 

 

 

 

Ok so admittedly i cringed somewhat watching this vid.

 

A couple things stood out to me.

 

1stly, you overwhelmed the single rad, so ofc an additional one is going to help, even in a undesirable airflow setup.

2ndly, the fan setup in that wasnt 'normal' , few people would turn the unfiltered rear fan to an intake, I guarantee that helped that top rad a lot.

 

As for the minecraft server build and ur 'test by removing one rad". The important difference wasnt the 5c . .the important difference was what the temp would have been if u allowed that rad to breath fresh air. Sure it lowered temps by 5c by having it there, but its efficiency was greatly hampered by being fed warmed air from the other 2 rads.

 

The point i think corsair was trying to make is that setting up airflow that uses warmed air from rad(s) can make a significant difference to the efficiency of the additional rad(s). And that where possible you should always allow rads to use cool air.

 

This is precisely why my loop setup in a Enthoo Primo has 3 240mm rad all as exhaust and all the other fan positions are intake. I've run a loop in the past using the same 3 rads in a different case where one or more was intake and the other(s) exhaust, the difference is night and day.

 

In smaller case like the one you used where you only have 2 x 240mm spaces + rear 120. The most rads you should put in for efficient use is a 240 at the top and a 120 at the back, leaving the front , and perhaps another fan at the bottom, as intake.

 

In the end i cant say this vid is giving out a great message. I know you say test things ur self and take the vid with a grain of salt and all, but u know full well thats not the message thats going to come across.

 

EDIT: instead of just criticism, here is what i would have done.

Get a case which can support 2 rads, ideally something like a 360 and a 240 so that its harder to overwhelm the single rad setup.

Test with a single 360 as top exhaust with front intake and rear exhaust.

Then test with the front intake going through the added 240mm rad.

Then test again with that front 240 rad just sitting somewhere outside the case so that both get fresh air.

Display the differences.

 

What 'should' show up is the 360 being the worst, the added 240 as intake improving cooling somewhat, but the setup where both rads get fresh air being the best.

 

Then for kicks, grab ur minecraft server and do the same, take out the 3rd rad to see the difference it makes in the setup u have (which u did (5c apparently)). Then allow the rad to sit out feeding on fresh air, it should help by more than 5c.

[DJRWolf]

It could also be important what rad is first off of your heat sources. For example if Rad-A is on the top of your case and the air is going out and Rad-B is at the front where the air is being brought in in the following loop directions.

 

Top First: Rez > CPU Block > GPU Block > Rad-A > Rad-B > Rez

Front First: Rez > CPU Block > GPU Block > Rad-B > Rad-A > Rez

 

With those two setups I would think the Top First setup where the water first goes to the rad mounted in the top of the case and the air from it is going strait out would perform better because it would be the hottest and then the water that has already been cooled by that rad then goes to the front rad to be cooled further before going to the reservoir.

 

Another thing to consider would be adding a single rad to the back along with the dual on top instead of a dual at the front  and a dual on top so that nether rad is putting hot air into the case. Could be that is better performance due to the air across what is effectively a triple would be cooler then the quad area of the two double's due to the heat from the one in the front being pushed into the case.

[SolarNova]

7 minutes ago, DJRWolf said:

It could also be important what rad is first off of your heat sources. For example if Rad-A is on the top of your case and the air is going out and Rad-B is at the front where the air is being brought in in the following loop directions.

 

Top First: Rez > CPU Block > GPU Block > Rad-A > Rad-B > Rez

Front First: Rez > CPU Block > GPU Block > Rad-B > Rad-A > Rez

 

With those two setups I would think the Top First setup where the water first goes to the rad mounted in the top of the case and the air from it is going strait out would perform better because it would be the hottest and then the water that has already been cooled by that rad then goes to the front rad to be cooled further before going to the reservoir.

 

Another thing to consider would be adding a single rad to the back along with the dual on top instead of a dual at the front  and a dual on top so that nether rad is putting hot air into the case. Could be that is better performance due to the air across what is effectively a triple would be cooler then the quad area of the two double's due to the heat from the one in the front being pushed into the case.

Watercooling 101, loop order is irrelevant. Waterloops are equalized.

[Nystemy]

Corsair sates:
"Use a baffle to guide airflow from the front of the case more efficiently, so that both radiators can be used optimally."
And the way the video answers the message is: "wha? should I not stack radiators? So one is better than two...."

Honestly, Corsair just came with advice on how to do it more properly.
A radiator is a fairly large restrictive surface, one can use a fairly slim amount of front panel space and still get adequate airflow out of it.
Now, with a proper baffle setup, one can rotate the three radiators, put baffles between them, and feed each one its own air from the front panel. Giving the system optimal efficiency.

Here is a rough literally thrown together 3D model and paint overlay: (the air gaps on top left and bottom right will need baffles too, just didn't feel like modeling them for a quick sketch. Not to mention the fans...)


Here one can also connect the water path in parallel through these three radiators to lower the flow resistance for the water, increasing flow speed and the cooling performance of our water blocks. All without needing to worry about loop order through our rads. (though, a sufficiently good pump might already though be limited by max flow speed, so this might not matter. But for weak pumps, it can be worth taking in mind.)

As pointed out, radiators have very restrictive air paths, so we do not need a lot of space in front of them for them to still get plenty of air flow.

Corsair were though having a different baffle idea that didn't involve rotating the rads. (in this case 10 degrees.)
But Corsair's idea were only involving two radiators, while @LinusTech used three.

Also, another advantage of using radiators in this fashion is that we also lower the air resistance, so the air flowing through the case cooling other components will also be more efficient in its job to keep VRMs and such in check.

[Enderman]

1 hour ago, Doommius said:

So, i just watched the video and i have 2 comments, i would say water flow has a lot to say here as well, if you the loop goes:

CPU -> GPU -> front rad(intake) -> top rad(exhaust) -> pump -> 

vs

CPU -> GPU -> top rad(exhaust) -> front rad(intake) -> pump ->

 

as the last will will be able to cool batter as it has a higher delta T for both rads.

 

Just my two cents. would like to see something like this as well for the Minecraft server build with flow back to front rad and front to back rad.

Flow order basically doesn't matter.

The delta water temperature before and after a radiator is a fraction of a degree, assuming you're using a decent pump like a D5 and not a tiny AIO pump.

 

1 hour ago, GDRRiley said:

Corsairs model seems to be over estimating how efficient the rads are. I don't think any rad would take say 40C water with 20C air and turn it into 39C air and lower the water temp. There also maybe a point that a single bigger (360 or 420 or 480) is better than 2 240 but that would mean a lot more testing which I'd love to see.

Yeah that's true, that's some magical radiator they've got right there 

[AnonymousGuy]

http://www.xtremesystems.org/forums/showthread.php?220874-More-Radiator-Sandwich-testing

 

 

Pretty much everything watercooling has been tested already, years ago.  A few takeaways:

 

1) You don't need as much airflow as you think.  It only helps very little to have tons of air flowing through a radiator.

2) You don't need as much pump speed as you think.  Anything above a trickle is fine because the heat loads aren't absurdly large.

[geo3]

11 minutes ago, Nystemy said:

This is also a massive waste of space compared to just stacking them. 

[Nystemy]

1 minute ago, geo3 said:

This is also a massive waste of space compared to just stacking them. 

But it increases the efficiency of our water cooling and increases air flow as well.
Yes, it eats space like an E-ATX motherboard or worse, but it will give optimal performance for the components used.

If space is at a premium, then this of course isn't an ideal solution.

[Esteban_G]

First, If you have enough air flow, the difference will be none or negligible. BUT if you don’t, there are different scenarios and cases where there could be a difference:
 

Situation 1: Two radiator take air from outside the case (corsair recommendation):

Water flow: CPU/GPU -> Rad1 -> Rad2

- Case 1: water reach room temperature on the First radiator. Then the second radiator is pointless because it cannot cool the water below room temp

- Case 2: water does not reach room temperature on the first radiator. Then a second radiator will be able to cool the water even more. BUT as the difference between the water and the air temperature is smaller than in the first radiator, the efficiency (the amount of heat extracted) will be smaller than in the first radiator.

 

Situation 2: Radiator 1 takes air from outside, the other from inside:

Water flow: CPU/GPU -> Rad1 (air from outside) -> Rad2 (air from inside)

- Case 1: water reach room temperature on the First radiator. Then the second radiator could be a problem because (in the worst case) it could be possible that the air going through the radiators is hotter than the water (because it absorbed heat from the first rad), and in that case, the radiator will increase the water temperature.

- Case 2: water does not reach room temperature on the first radiator. Then a second radiator will be able to cool the water even more. BUT as the difference between the water and the air temperature is even smaller than in the previous situation, so the efficiency (the amount of heat extracted) will be lower. Also, it won’t be able to lower the water temperature to room temperature because the air inside the case will be hotter than that.

 

Situation 3: Radiator 1 takes air from inside, the other from outside:

Water flow: CPU/GPU -> Rad1 (air from inside) -> Rad2 (air from outside)

- Case 1: water reach room temperature on the First radiator. Then the second radiator is pointless because it cannot cool the water below room temp.

- Case 2: water does not reach room temperature on the first radiator. Then a second radiator will be able to cool the water even more because the air passing through will be at room temp. Again, as the difference between the water and the air temperature is smaller, the amount of heat extracted will be lower. In this case it could be possible to cool the water to room temperature. An extra benefit is that the air pushed inside the case will be cooler than in situation 2 because this radiator will extract less heat into it.

 

Conclusion:

1- If you push enough air through the rads/case so it never gets hot, you can put the rads in any way you want. In that case all this analysis is pointless.

2- If you want the maximum efficiency with more than one rad, make them take always air from outside the case.

3- If you need to put the rads “in serial”, then you should connect the hottest water (the tube from CPU/GPU) to the rad that takes air from the case first.

4- If you need to stack rads, as in the Minecraft server, then you should connect the hottest water to the inner most rad and leave the outer most rad for last to maximize temperature difference.

5- The benefit of adding more rads to the loop, decreases with every rad as the difference in temperature is smaller.

[Razor512]

1 hour ago, Doommius said:

So, i just watched the video and i have 2 comments, i would say water flow has a lot to say here as well, if you the loop goes:

CPU -> GPU -> front rad(intake) -> top rad(exhaust) -> pump -> 

vs

CPU -> GPU -> top rad(exhaust) -> front rad(intake) -> pump ->

 

as the last will will be able to cool batter as it has a higher delta T for both rads.

 

Just my two cents. would like to see something like this as well for the Minecraft server build with flow back to front rad and front to back rad.

If having a rad on both the exhaust and intake, it would be better to have the first radiator after the head source being the exhaust. Radiators work best when there is a large heat delta, thus even in an open air test bench, the first radiator is always hotter (you can literally touch the radiator and feel the difference.

 

I personally have not done AB testing in a case (too much work to swap the order), and I have only ever done 3 multi-radiator builds, all other liquid cooled builds have been a single radiator. One could logically assume that if the first radiator in the loop is hotter than the second then having it in the intake will be worse overall since the rate of cooling is lower as the temperature delta reduces (this is why you get diminishing returns will additional radiators. If the initial heat dump happens in the case, then the next radiator will have a smaller thermal delta to work with, and often as the delta shrinks, you need an almost exponential increase in surface area to maintain cooling rate.

[Razor512]

mistaken double reply , not sure how it happened, please delete.

[geo3]

16 minutes ago, AnonymousGuy said:

http://www.xtremesystems.org/forums/showthread.php?220874-More-Radiator-Sandwich-testing

 

 

Pretty much everything watercooling has been tested already, years ago.  A few takeaways:

 

1) You don't need as much airflow as you think.  It only helps very little to have tons of air flowing through a radiator.

2) You don't need as much pump speed as you think.  Anything above a trickle is fine because the heat loads aren't absurdly large.

I find this chart very informative and interesting. One thing missing I would like to have seen, for the instance of stacked rads they they only show contra flow and forward flow, but I would like to see how those 2 compare to a parallel flow set up. So like using a T junction to send water through both stacked rads at the same time rather than in one than into another. Also the top performer doesn't show water flow because I guess it didn't matter and they were doing out one and into the other, but I would like to have seen parallel water flow here as well.  

[Razor512]

They really should label the Y axis on that graph.

[AnonymousGuy]

7 minutes ago, geo3 said:

I find this chart very informative and interesting. One thing missing I would like to have seen, for the instance of stacked rads they they only show contra flow and forward flow, but I would like to see how those 2 compare to a parallel flow set up. So like using a T junction to send water through both stacked rads at the same time rather than in one than into another. Also the top performer doesn't show water flow because I guess it didn't matter and they were doing out one and into the other, but I would like to have seen parallel water flow here as well.  

Parallel vs. serial probably wouldn't have made a significant difference.  The temperature delta between the water and the air is already not that extreme (heat transfer is directly proportional to the difference in temperature), so you're just not going to see much of a difference in the water temperature inlet vs. outlet.  

 

I can tell you with my loop the difference in water temperature on a 240mm radiator with 4 fans is less than 0.5 degrees C.   My GPU at idle heats up the water 1 degree C inlet vs. outlet.

[Nystemy]

9 minutes ago, geo3 said:

I find this chart very informative and interesting. One thing missing I would like to have seen, for the instance of stacked rads they they only show contra flow and forward flow, but I would like to see how those 2 compare to a parallel flow set up. So like using a T junction to send water through both stacked rads at the same time rather than in one than into another. Also the top performer doesn't show water flow because I guess it didn't matter and they were doing out one and into the other, but I would like to have seen parallel water flow here as well.  

The impact on parallel flow would be interesting to show.
Though, since the graph doesn't tell us flow speed, then it really wouldn't say all that much to be fair.

Since the main reason to put rads in parallel is to reduce the flow resistance, in case one's pump starts being the week link.
Since if we have a loop with low flow speed, then the water in the CPU/GPU cooling blocks will "hang around" being warm, instead of quickly moving out to let new cool water get into place.

Though, with a good enough pump, one can't really even measure the temperature difference between the warmest and coldest sides of one's loop, so flow speed is rarely an issue.

[AnonymousGuy]

6 minutes ago, Nystemy said:

Though, with a good enough pump, one can't really even measure the temperature difference between the warmest and coldest sides of one's loop, so flow speed is rarely an issue.

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.