Radiator Sandwich / Stacked Radiators / "SLI-Radiators" Test results

Lately, I've seen a lot of discussions about stacking two slim, low-cost radiators instead of using one thick, high-end radiator.

Since high-end radiators like those made by Feser and Thermochill often cost more than twice of what slim, low-cost models cost, radiator-sandwiches could be an economically sound alternative.

I had the chance to squeeze in some stacked-radiator testing. The emphasis here is on "squeeze in" since I'm really busy with several other testing projects. So forgive me if this isn't as elaborate as you might hope for.


Test setup:
In my test setup, I'm using some pretty low-end cooling, which has to do with an article that I built this setup for.

- Q6600 @ 3.2 GHz with Nexxxos XP Waterblock
- HD 4850 with EK FC Waterblock
- XSPC X2O 400 Pump
- Magicool Slim 360 radiators
- three Nanoxia fans

I ran this setup first with just one of the Magicool radiators and then with a radiator sandwich structured as Radiator -> Fans -> Radiator.
Here are some impressions of the sandwich:

http://www.abload.de/img/0427cyf.jpg

http://www.abload.de/img/0439emh.jpg

http://www.abload.de/img/041k0vs.jpg


Testing method:

I run Furmark to load the GPU and Prime 95 to load the CPU for 45 minutes to warm up. Then I run it for another 20 minutes during wich I log all temperature data in five second intervals. The data of those 20 minutes is then averaged out to get the results.

Since we're testing radiators here, the most relevant temperature difference is that of water-air.

Here's the results:
http://www.abload.de/img/graph1wasserluftsandwii69l.jpg

I tested once with 750rpm and once with 1400rpm. As we can see, the sandwich gains a good two to three Kelvin over the lone radiator. This also translates to lower CPU- and GPU-temperatures.
As you can see, for the radiator sandwich it doesn't seem to matter what fanspeed is used. Since the T-Balancer, the device I use for temperature-logging has a resolution of only 0.5 degrees, we're certainly not seeing the whole picture here. But I guess we can say that stacked radiators are less sensitive to fan speeds than lone radiators nonetheless.

As for flowrates, the additional radiator made them drop by 25% in my testing.

Now, it remains to be seen how the radiator sandwich performs vs. a high-end radiator. I just so happens that I'm testing that as i write... :D

Cheers,
Shane


EDIT:

Temperature difference CPU-Ambient:

http://www.abload.de/img/sandwichgraphnewj3f8.jpg





Scenarios for further testing:

In the meantime, a lot of interesting suggestions have come up. Here's a list of all of the ideas that should be tested. Please contribute! :-)


- Two radiators separately vs. stacked radiators
- Stacked radiators with a total of six fans
- Stacked radiators with a total of nine fans? (Overkill?)
- Stacked radiators vs. PA120.3 (testing in progress on this one)
- Airflow going from first to second radiator vs. airflow going from second to first radiator (as discussed in posts 29-33)
- Radiators stacked right on top of each other: Radiator -> Radiator -> Fans
- Running stacked radiators in paralell

Nice! The only problem i see is the physical dimensions of that sandwich.

Thanks!

Well, the dimensions are actually practically the same as with a thick radiator. With the sandwich you have 30mm deep radiators x2 plus 25mm fan = 85mm. Compared to that, a PA12.3 is 64mm deep plus 25mm fan = 89mm!

Great info thanks!

-sponge

It would have been great if you could have done the 2 rads separate with their own fan to see how much of a performance hit the sandwich design is. This is the primary argument against such a set up.

Interesting. I would have expected the sandwich to yield proportionately larger benefits at 1400rpm fanspeeds, since the air wouldn't get as saturated with heat while passing through the first radiator.

Very interesting. But those barbs!

Fan speeds...why?????

"Fan speeds... why?" What?
I don't understand what you mean. Please clarify?

The fittings are simply the only 90s that I have. But I can tell you that they have next to no impact on flowrates (less than three percentage points per fitting), so I don't see a real problem with using them.

Nice. I was actually thinking about doing this with the BIP GTS and some heavy blowing fans but figured it would be cheaper/easier to purchase a BIP GTX.

Quote:

---

Originally Posted by MomijiTMO

It would have been great if you could have done the 2 rads separate with their own fan to see how much of a performance hit the sandwich design is. This is the primary argument against such a set up.

---

My thoughts exactly. Not exactly fair testing two rads against one alone. Doing a test with two rads separate and then 2 rads sandwiched is what I am looking for here.

another vote for testing the sandwich vs 2 separate radiators to see if the design can work.

Quote:

---

Originally Posted by Sadasius

My thoughts exactly. Not exactly fair testing two rads against one alone. Doing a test with two rads separate and then 2 rads sandwiched is what I am looking for here.

---

I don't have the capacity or setup to test the two radiators seperately as well as sandwiched. I'm pretty confident that if they are used seperately, the two radiators do much better than when sandwiched.

I just wanted to test from a perspective of limited space and finances. Since two radiators separately need lots more space as well as three additional fans it would again be "unfair" to compare them with the rad-sandwich...

EDIT: If no one else steps up and does the testing, I might be able to fit it in at some later point. ;-)

I definitely appreciate your efforts toward this info, so--Thank You! I am contemplating the same basic concept for my dual car rad setup for this summer, and trying to determine if it would be beneficial, the best setup and plumbing or flow course.

:up:

Quote:

---

Originally Posted by Sadasius

My thoughts exactly. Not exactly fair testing two rads against one alone. Doing a test with two rads separate and then 2 rads sandwiched is what I am looking for here.

---

I agree completely, but personally the testing that you've done already offers some benefit to me. However, I think we already knew that 2 rads would beat one, so the value is limited.....:up:



My car rads are not in a pc (duh!), but the principles should be the same. Personally, I would appreciate any info that results from the configuration below:

http://i50.photobucket.com/albums/f3...h-Modified.jpg

I am working on other things atm, but I have been giving this some thought and the air flow/water flow above--I believe--would produce the best results...for a sandwitched setup. I may be wrong though....:up:

If you would care to check out that setup.....You could save me some guesswork....

Either way: Good Effort.....:up:

Thanks for doing this :) Looking forward to more from you :up:

It was nice of you trying this setup out. Please can you make out time to try Naja002´s idea, i know i once made a proposal about this configuration. The inlet/outlet rad and air direction will play an important role in this setup, in as much as another set of fan will reduce the temp too.

@HESmelaugh
:up:

I think you should really think about (better: test ;)) Naja002s proposal ...

Quote:

---

Originally Posted by Sadasius

My thoughts exactly. Not exactly fair testing two rads against one alone. Doing a test with two rads separate and then 2 rads sandwiched is what I am looking for here.

---

x100.. i mean, sure 2 work better than one, but without knowing how well (or dismal) the sandwich works vs the 2 rads separated is the only relevant test here. It's kinda like saying, 2 D5's has more pressure than one.

Quote:

---

Originally Posted by HESmelaugh

I don't have the capacity or setup to test the two radiators seperately as well as sandwiched. I'm pretty confident that if they are used seperately, the two radiators do much better than when sandwiched.

---

Why not? Never know until you try.

first, let me say that I really like all the feedback and input I'm getting here! XS sure appreciates it's testers. :up:

The reason I could only do such a brief test is that I'm currently working on quite an extensive series of testing that will illustrate how best to upgrade a LC setup. After that, I have another CPU-waterblock roundup scheduled and that will eat up a lot of my free time. After that, even more stuff is scheduled, though some of it less pressing.
I only mention this to explain why I can't deliver the testing you asked for right away: I have only one test-rig and my days are, alas, limited to 24 hours each...

Having said that, I will get a new bench-table type rig in a few weeks. It will allow lots of flexibility with radiators and make this kind of testing easier. So keep bringing the ideas for what should be tested and I will get to it ASAP.

what about connecting radiators in parallel like car intercoolers. has anyone tried this?

How does it compare to a single thick radiator like a TC 120.3 with three fans is another suggestion.

why not just paralell them too while your at it.

And then throw them in a 400 mile wind tunnel to test efficiency.

ROFL...

This is why sometimes i keep quiet with my test.

Great test tho! :up:

You just proved physics to be correct. :P

But i would really try a paralell design over a serial and compare that.

Anyone have a spare MCR320 they would be willing to loan me for a comparison test? I already have one, just need #2 for a couple weeks as the other piece of bread in this sandwich. :D

edit:
Or a GTX480, but I figured a MCR320 would be easier.

Quote:

---

Originally Posted by mcoffey

yeah,

And regardless, I'd need to see what the CPU temps were before and after adjusted to ambiant differnece before the test is of any real use to me.

And thanks for taking the time to test,

andyc

---

That I can offer:

Temperature difference CPU-Ambient:

http://www.abload.de/img/graph2cpuluftsandwich2rnib.jpg


I've also finished the first test-run with a PA120.3. It looks like the sandwich performs a little better, but I'm not sure if I really got all the air out, so don't take my word for it just yet.

My approach to testing this came from reading about two other, similar tests that showed next to no effect for the sandwich vs. just one radiator. It seemed like the air got so warm after the first radiator that it just didn't do anything for the second one anymore.
That's why my focus was on verifying or disproving the hypothesis: Radiator sandwich is no better than single radiator. I must admit that I just didn't test enough scenarios, but as I said: I'll do more runs when I find the time.

Oh, and someone send Skinee a radiator! Two sets of data are always better than one (especially when they completely contradict each other?). :)

Quote:

---

Originally Posted by HESmelaugh

"Fan speeds... why?" What?
I don't understand what you mean. Please clarify?

---

Clarification (sorry): To what do you attribute the non-effect of the fan speeds on the "sandwitch" setup?

As for the barbs, I race sail boats and we spend about two weeks getting the boat ready for spring racing. There's a guy in the fleet that spends all winter working on his boat. Some of the things he does are rather trivial and would only result in a few seconds savings over a 3-4 hour race ...but he does hundreds of them things! and his boat is actually a couple of minutes faster. (So cunning and trickery have to be our tools!)

Well, I see barbs as being one thing among many too. So, 1% here, 2% there, and pretty soon you're talking about 10°C!

Quote:

---

Originally Posted by Naja002

I would appreciate any info that results from the configuration below:

---

Since the deltaT across the water inlet and outlet will likely be less than 0.5°C, I doubt that there would be much difference at all. My only thought though is that the greatest cooling happens when the deltaT between the water and the air is the greatest. ?????

I generally agree on your "lots of little things make a difference" theory. As soon as someone sends me better barbs, I'll use 'em! :-)

I attribute the non-effect of the fan speeds simply to error in measurement. The device I use for logging the temperatures has a resolution of only 0.5 degrees, so the difference might simply be smaller than that. Apart from that, even with long test runs, there's always a margin of error. So I would say in reality, the results are very close but not, as it seems from the data, identical.

Interesting results.. Thanks for the test.
I wonder how it would compare to a PA or even 1 more set of fans on the top or 2 more sets of fans for the top and bottom.

Quote:

---

Originally Posted by HESmelaugh

My approach to testing this came from reading about two other, similar tests that showed next to no effect for the sandwich vs. just one radiator. It seemed like the air got so warm after the first radiator that it just didn't do anything for the second one anymore.
That's why my focus was on verifying or disproving the hypothesis: Radiator sandwich is no better than single radiator. I must admit that I just didn't test enough scenarios, but as I said: I'll do more runs when I find the time.

---

In the pic below--if the air flow is reversed...I would expect the results (no effect) to be as you stated above. Air/water flow is important in this type of setup and, again, I could be wrong....but it seems like common sense to me. :up:

Quote:

---

Originally Posted by Naja002

http://i50.photobucket.com/albums/f3...h-Modified.jpg

---

I love this !!

Thinking of getting another Black Ice 360GT to do it...

Thank you
Sergio

I think I see what you mean, Naja002. With the airflow in the wrong direction, you'd get relatively warm air blowing on relatively cool water, right?

The two options should definetely be compared to each other.

I'll start a list of scenarios that should be tested in the first post (tomorrow, though. On this part of the globe, it's getting really late).

Right. If the ambient air goes over the coolest water first its going to get slightly warmer (1.5C or less) which will be removing heat from that water....it will then go over the much warmer water--removing the bulk of the heat. Its about temp difference. The air going over the coolest water will remove some more heat (energy), but not enough to increase the air temp v. the water temp in the hottest rad. So, the air should still have a significant temp difference compared to the water in the hottest rad....and remove the bulk of the heat as it passes through.....

The reverse would just be reheating cooled water.....

Quote:

---

Originally Posted by Naja002

The reverse would just be reheating cooled water.....

---

no. water temp is always higher than ambient, you would never get "reheating" of the water. the delta air-water would just be smaller

- Update -

Added the following list of test-scenarios to first post:



Scenarios for further testing:

In the meantime, a lot of interesting suggestions have come up. Here's a list of all of the ideas that should be tested. Please contribute! :-)

- Two radiators separately vs. stacked radiators
- Stacked radiators with a total of six fans
- Stacked radiators with a total of nine fans? (Overkill?)
- Stacked radiators vs. PA120.3 (testing in progress on this one)
- Airflow going from first to second radiator vs. airflow going from second to first radiator (as discussed in posts 29-33)

Quote:

---

Originally Posted by SNiiPE_DoGG

no. water temp is always higher than ambient, you would never get "reheating" of the water. the delta air-water would just be smaller

---

Reheating of water woun´t happen but the aim of the second rad is defeated.
Front mounted rads perform better than rear or top mounted rads that has to draw warm air from inside the case.
The direction of air flow/which rad gets the water first in stacked rads will play an important role when the configuration is being used. I have tried single rad with front and later back mounts and i noticed that i lost about 4~5°c difference in temp with front mounted rad but other hardwares get heated up from the heat being dumped in the case. Thats why i went with a external mounted rad. The bottom line is that when the water flow from rad1 to rad2 and the air flows thur rad1 to rad2; the water exiting the rad2 will not loss enough heat due to warm air coming from the first rad1

Quote:

---

Originally Posted by Shocker003

Reheating of water woun´t happen but the aim of the second rad is defeated.
Front mounted rads perform better than rear or top mounted rads that has to draw warm air from inside the case.
The direction of air flow/which rad gets the water first in stacked rads will play an important role when the configuration is being used. I have tried single rad with front and later back mounts and i noticed that i lost about 4~5°c difference in temp with front mounted rad but other hardwares get heated up from the heat being dumped in the case. Thats why i went with a external mounted rad. The bottom line is that when the water flow from rad1 to rad2 and the air flows thur rad1 to rad2; the water exiting the rad2 will not loss enough heat due to warm air coming from the first rad1

---

The significance of this greatly depends on the rate of airflow.

would the flowrate of the coolant have any impact on this?

can you add in scenario:

rad 1 - rad 2 - 38mm fan

thanks for this tests, i'm looking for your next review since i'm building a similar watercooling setup.

Quote:

---

Originally Posted by septim

can you add in scenario:

rad 1 - rad 2 - 38mm fan

---

Done. I don't know if I can do it with 38mm fans though. Firstly, because I don't have three of a kind right now (that would be a minor problem though) and secondly, because it's then less comparable with all the other scenarios.


Just let me throw this idea out there: Let's assume that Skinnee (or someone else) also gets a chance to test stacked radiators. How about I test with 25mm fans and someone else with 38mm fans (for all the scenarios)?

How did you screw the rads together?

Hey ive got a extra TC 120.2 maybe I can sandwich that with 1 of my TC 120.3's Wonder if that would make much difference on my Gtx280's

Quote:

---

Originally Posted by masxerofkartis

How did you screw the rads together?

---

Like this:

http://www.abload.de/img/044u8fj.jpg

Each fan is attached to the lower radiator at two diagonal points and to the upper radiator at the other two diagonal points.

Quote:

---

Originally Posted by SNiiPE_DoGG

no. water temp is always higher than ambient, you would never get "reheating" of the water. the delta air-water would just be smaller

---

Agreed. Just trying to keep it simple for those going through the mental exercise...

If the flowrate is to your concern and you want to avoid elbow fittings, how about turning one of the radiators 180 degrees, so that the fittings sits opposite of one another.
Thats how I have my 2 rads in my TJ07.
Something like this:

Quote:

---

Originally Posted by Naja002

Agreed. Just trying to keep it simple for those going through the mental exercise...

---

Quote:

---

Originally Posted by silverphoenix

The significance of this greatly depends on the rate of airflow.

---

Well i rest my case for now cos am a wc noob.:p:
Waiting for test results ;)

- UPDATE -

Here are the CPU minus Ambient temperatures including a comparison to the PA120.3:

http://www.abload.de/img/sandwichgraphnewj3f8.jpg

This has also been added to the first post.

Cheers,
Shane

Superb work! Thanks so much for taking the time and putting in all the effort to try this, it's something that has been asked about before.

I really didn't expect it to do as well as it did. Generally I've found the double thickness is really only good for about a 10-15% gain if that over a standard thickness. Looks like this sandwiching may be work 30% or more...that's very good!

I think this sandwiching is getting additional gains because of fan turbulence internally. Just like the shroud testing I found there was an optimal distance from the radiator, and I can only attribute that is due to some sort of fan scrubbing turbulence thing when the fan is close to the fins.

Anyhow, this is very interesting, it appears to do better than the double thickness rads and costs less....wow!

Any thoughts on sound quality differences?

Thanks!! Amazing work!!

Off to strapping a few rads together for my regular system...:D

I was surprised myself that the sandwich beat the PA. Just to make sure, I did two test-runs each.
And I have to admit that I really don't have enough data to be able to understand how or why the sandwich is so effective.

On a completely different note: How great would two BlackIce rads with LED-Fans in between look? :-)

Quote:

---

Originally Posted by HESmelaugh

I was surprised myself that the sandwich beat the PA. Just to make sure, I did two test-runs each.
And I have to admit that I really don't have enough data to be able to understand how or why the sandwich is so effective.

On a completely different note: How great would two BlackIce rads with LED-Fans in between look? :-)

---

So all you used was one set of fans betwen the radiators?
25mm too.. hmm
Did you try it with more fans, middle & top?

Quote:

---

Originally Posted by HESmelaugh

On a completely different note: How great would two BlackIce rads with LED-Fans in between look? :-)

---

ahem..

2 fessers 480's with 8 sanaces in the middle and 3 or 4 DDC's :rofl:
http://i125.photobucket.com/albums/p...a/IMG_0095.jpg

i'll tell you when oracle is finished.. :rofl:
This will be semi bling build...

@Eddie:
Yes, I used one set of fans between the rads. No other setups tested yet.

@Naekuh:
Are you going to sandwich 'em? :D
Anyhoo, the I think of the whole radiator stacking as something more for those among us who don't have a money-tree in the back yard (or a back yard, for that matter)... :-)
Just kidding.

Too bad the radiators dont have holes for fittings on top so you could toss the 90's out.

Interesting testing :)

Quote:

---

Originally Posted by HESmelaugh

@Eddie:
@Naekuh:
Are you going to sandwich 'em? :D

---

yup. double stacked san ace core. :up:

I KNEW IT:D

Cant wait for the Swiftech version.

wow, HESmelaugh killer PA120.3? ;)
nice test. i have mcr320 and mcr220, i need to try to do the same

The other test you could try is running the radiators in parallel instead of series. This is essentially how the double thickness rads run and why they generally have lower pressure drop.

Running them in series should perform better with the higher flow rate, but parallel might be an option if restriction was a concern.

Running two separate rads should be worth roughly half the previous water/air delta or a 50% gain, but 30% gain for only 3 fans is a pretty healthy gain.

What actually would be interesting is to see if using shrouds has any effect, i.e. Rad - Shroud - Fan - Shroud - Rad. I wonder if the shrouds would focus the air flow better in this scenario.

Wow, those are fairly impressive results. It'll be good to see how the Rad > Fan > Rad (or Rad > fan > Rad > fan) setup goes compared to two separate radiators, but this certainly looks hopeful for a lot of people who can't fit two 120x3 radiators in a case...

Thanks a lot for the tests, and you're giving me ideas *credit card commits suicide*

Eller

I'm planning on stacking 2 120.1 rads.

Quote:

---

Originally Posted by Sadasius

My thoughts exactly. Not exactly fair testing two rads against one alone.

---

I'm confused, how is it unfair? Obviously in a test to see which single rad cools better you wouldn't do that but this one is just to show you can get the same cooling as PA120.3 for cheaper right?

Ok, I added testing the rads in paralell to the list of things that need testing.

I'm also thinking of using a different testing method. I would like to have a heat source inside the water (in a reservoir). It should be possible to get a more constant heat output and more precise results running something else than a cpu and gpu right?

Quote:

---

Originally Posted by HESmelaugh

I was surprised myself that the sandwich beat the PA. Just to make sure, I did two test-runs each.
And I have to admit that I really don't have enough data to be able to understand how or why the sandwich is so effective.

---

The reason is pretty obvious.
The second rad adds more surface area for cooling and despite getting hotter air at the intake with lower CFM, it's still able to dissipate more heat than a single rad.

Btw, I had this idea in my mind too some time ago, but the wiser choice for this sandwich setup will be the XSPC rad, the R120-T (V)
http://www.xs-pc.co.uk/r120tvbig.jpg

with the horizontal ports which has less impact on flow and easier for setup, can have you choose between running two different loops, one loop in parallel or one in serial choices are fun :D

Quote:

---

Originally Posted by HESmelaugh

Ok, I added testing the rads in paralell to the list of things that need testing.

No need to test if you're not interested, it's a tremendous amount of work for maybe no difference at all...just a thought. I'd probably be more interested in comparing the same with more fans, something like a PA with 6 fans and the sandwich with 6 fans or something like that. Whatever you're willing and have time/interest to do.

I'm also thinking of using a different testing method. I would like to have a heat source inside the water (in a reservoir). It should be possible to get a more constant heat output and more precise results running something else than a cpu and gpu right?

---

You can use the water heaters like I was doing if you want, but you'll have to take the heater apart and solder the thermostat solid. There is a little metal spring that pulls the contact away when the temperature is high. I just solder that contact together and slide it all back together again. The biggest pain is getting the thing apart.

for a triple rad I would use at least a 300W heater, two for a total of 600W is even better if you're trying to split hairs.

The hardest part of the radiator testing for me was waiting for equilibrium and ensuring a constant ambient. It doesn't matter what the ambient is, but it needs to be holding constant. If it's rising or lowering, you're delta will be off a bit. Unfortunately it usually takes a couple of hours of testing before you've warmed up the testing area to a point where you might have a fixed ambient.

Also lot of air in sensors helps, you probably can't have too many and they all need to be in the same exact place for each test because you will get radiant heat from the radiator. Probably best if your air in sensors can be several inches away from the radiator if possible.

Finally the fan RPM is another pain. As the fan warms up the RPM will increase. You have to let the fan warm up for about 10 minutes at the desired RPM or constantly monitor and adjust.

Rad testing is one of the more difficult ones I've found, it's a PITA. Oh yeah one more, don't change the air flow in your room. A small think like a fan running in the room will change your results I noticed this when I got my air conditioner and used the fan to circulate air in the room better. I had to retest anything I wanted to use for a comparison.

Yet another reason you are highly regarded Martin.

Thanks for the info, Martin.
I was thinking about getting two heaters with 300W each. I still need to make a reservoir where I can submerge them, though. I already use multiple air in sensors and have a rig where their distance to the radiator is always fixed.

I probably won't be able to get results as precise as the ones we got used to from your testing, but an improvement over the setup I have now should be easily achievable.

Changing fan rpm is something I have also noticed but don't have a good way of controlling yet.

I just had an idea. Maybe there could be an easy alternative to double-threaded radiators that would allow us to connect any two radiators together.
I tried to illustrate with a quick sketch:

http://www.abload.de/img/radconnectorcopy1xcn.jpg

There is something similar available for connecting two GPU waterblocks. Example: Link
All that needs is another G1/4" thread in the middle and you've got a T-line for connecting two radiators in paralell. The radiator-threads could then simply be facing each other.
What do you think? Would this work?

Yeah that would work great as long as both radiators matched. Running the two in parallel would probably change the temps very slightly as you wouldn't have one rad with slightly warmer water then the other.

What about increasing the spacing between the radiators taking an old fan and cutting out the fan and just use the frame. and then you could have a set up like this.
___________________
[_______Rad 1_______] Outlet Rad (Cooler water)
[_______Fan 1_______] *Possibly switch fan 1 and spacer depending on results.
[______Spacer_______]
[_______Rad 2_______] Inlet Rad (Warmer Water)
[_______Fan 2_______]

with the air flowing downward through Rad 1

I wonder how that would perform on a dual or even triple 120 rad setup.

with the koolance sli connectors and the cube like -| |- only for an 38mm fan though. this is the neatest way to stack rads in parallel...

Quote:

---

Originally Posted by JvT

If the flowrate is to your concern and you want to avoid elbow fittings, how about turning one of the radiators 180 degrees, so that the fittings sits opposite of one another.
Thats how I have my 2 rads in my TJ07.
Something like this:

---

How about just some shrouds.

Holy banana! Where are you going to fit that? ^^

Honestly though, if you've got enough space for that construction, you might as well equip both rads with fans with the airflow pointing away from the center. That way, both rads get fresh air.

Yes, this is true. If I were to stack radiators it would be somewhere like the lower compartment of a P18x series or perhaps TJ07.

Anyways, just popped into the thread to congratulate and thank you for your testing HESmelaugh, keep up the superb-ness. :p:

Quote:

---

Originally Posted by HESmelaugh

I just had an idea. Maybe there could be an easy alternative to double-threaded radiators that would allow us to connect any two radiators together.
I tried to illustrate with a quick sketch:

http://www.abload.de/img/radconnectorcopy1xcn.jpg

There is something similar available for connecting two GPU waterblocks. Example: Link
All that needs is another G1/4" thread in the middle and you've got a T-line for connecting two radiators in paralell. The radiator-threads could then simply be facing each other.
What do you think? Would this work?

---

Wouldn't that just be a T-line?:confused:

Quote:

---

Originally Posted by HESmelaugh

Thanks for the info, Martin.
I was thinking about getting two heaters with 300W each. I still need to make a reservoir where I can submerge them, though. I already use multiple air in sensors and have a rig where their distance to the radiator is always fixed.

I probably won't be able to get results as precise as the ones we got used to from your testing, but an improvement over the setup I have now should be easily achievable.

Changing fan rpm is something I have also noticed but don't have a good way of controlling yet.

---

Perfect!

There's no right or wrong way IMHO, just things to watch for and try to keep the same across comparative tests.

For Fan RPM, I've done it a couple of ways:
-Variable PSU and monitored RPM on the crystalfontz, could also be monitored on any MB fan header. You just need to splice off the yellow and black wires for the RPM to the fan header and use either the variable PSU or fan controller to tweak the voltage until the RPM meets the target. Plus or minus 20RPM or so is about as good as it gets.
-Crystalfontz percentage controlled.

Fan RPM is hard though, slight variations in household voltage will still make it go up an down. Something like a line voltage conditioner might be the ticket for fixing that little issue. In the end, I just decided to log the RPM.

Anyhow, just a few challenges I had.

I think it would also be interesting to try and measure heat dissipated the way the coolingmasters guys do with water in vs. water out and flow rate. That might be a means for getting around all the issues with ambient changing and such, but you'd need to really beef up the precision of the sensors to measure that small of a delta. I had thought about trying to rig about 5 sensors per side and average all 5 sensors to improve resolution. I think you probably need relative accurace down to .01 or so.

Anyhow, the stack rads is pretty neat, keep up the awesome work!:up:

Quote:

---

Originally Posted by Boogerlad

Wouldn't that just be a T-line?:confused:

---

Not exactly. All the T-lines I know have static threads, which makes it impossible to connect a radiator on either side. What we would need is a T-line with rotary threads. :cool:
In case you know anyone who makes such T-lines, please let me know!


Martin: I'll try warming the fans up for a few minutes and then logging the rpm. Might turn out to be the easiest solution. Though I'm afraid I will have to get out a soldering iron (or my purse) to modify/get new testing equipment at some point...


InCore: Thanks!

there's been in my mind bout ... 2 MCR320 vs 1 PA120.3

2 MCR320 vs 1 PA120.3
(34cm + 34cm width) vs (62cm width)

That would be very interesting. Do you have the radiators, or are you thinking of getting a setup like that?

howsa about a single shoud in-between rads that is a simple box / full cover size & have a push /pull fans on the outside of the stacked rads ??

I love the sandwich action buddy. I was considering this instead of a single p.a. 120.3 rad up top. I was also thinking of doing this with single rads in small spaces but wasnt sure how the barb setup would turn out. Nice Job! :D

instead of 90 degree's i'm gonna mod up some copper pipe that will take most of the hard turns outta the tubing runs with the 90's
STAY tuned i'll post pix when i get the setup done .
I like this idea & am gonna run with it ..
........i got a dozen MCR320 0n the shelf ............

with hi speed fans &/OR push/pull hi speed fans would not the placement of the hot side vs the cold side out recieving the coldest air ...
have a minimal effect whether it be hot OR cold ?? meaning series looped ....???
.......................... any thoughts ANYONE ??

cause if this works u can have 2 mcr320 & some hispeed Yate Loons for less cost than some of the high priced rads not counting the fans price & most of us could have preformace at a reasonable cost
YEAH the size would be bigger but FTW on size/room............ Preformace way out takes the CAKE than JUST looking pretty ..................TEMPS Rule

Bigslappy, the are two ways of looking at the order of the rads well 3 if you count parallel! I guess you know what they are but I will outline them anyway and then get to what I believe to be best.

So Rad1 has the hottest water (inlet), this is then passed to Rad2 and then back to the loop.

1. Cold air over Rad1 first
2. Cold air over Rad2 first
3. Cold air over stacked parallel rads

We will now sidestep to theory for a second.

Imagine 2 tubes one inside the other, one filled with hot water the other with cold water (the following is dependent on flowrates so we will assume they are the same for both hot and cold) so you have water at 10°C (cold water) and water at 90°C (hot water) and you want to cool the hot water down, in co-current flow (option 1 above) your hottest water contacts the coolest water (largest driving force - good) assuming you have an infinite length of pipe the coolest you can get the water is 50°C and as the cold water gets hotter the driving force shrinks so it takes longer and longer to reach equilibrium. If you had counter-current flow (option 2) then along that same length of pipe you could theoretically cool the hot water to 10°C and heat the cold water to 90°C the driving force will remain constant over the entire length of the pipe however it will take longer to cool the the hot liquid to say 80°C as the driving force would be less.

Stepping back to real life now the question we have to ask is do we when running under option 1 heat the air coming out of the first radiator enough that it significantly impacts on cooling on the second this is a difficult one to answer as it will depend on a number of factors but most importantly on the airflow rate in relation to the liquid flow rate, assuming that liquid flow rate is fixed its all down to fan speed. If we have higher fan speed we decrease the heat transferred to each volume of air passing through the radiator this increases the driving force. So this means that if you use high speed fans then the order of radiators is probably of little importance as the air passing over each one will be of similar temperature (option 1 may provide small benefits due to the higher initial driving force), if however you use slower spinning fans then you would expect option 2 to be the better performer as the temperature gradient over both radiators would be more consistent.

Parallel rads option 3 is really of no use when stacking unless you want higher total liquid flow rates generated from the lower pressure drop associated with this flow regime.

So my advice would be option 2 cold air cold rad this should provide the best average performance over a wide range of fan speeds. But probably worth testing to see what actually happens!

As it happens I have 2 PA160's stacked together in a push pull config Fan->Shroud->Rad->Rad->Shroud->Fan and I really should get round to testing them with the air flow the other way (currently using option 2 with 38mm fans ~1200rpm) but I have more pressing matters to attend to atm though that should be finished in a couple of weeks and then I will give it a go.

Sounds very nice, bigslappy! Looking forward to seeing what you can come up with. :yepp:

Seeing that there were two tests on german forums that showed almost no temperature improvement from a single radiator to stacked radiators, I assume that it makes a significant difference, which rad gets ventilated first and which second.
More testing will give us more of a clue.

Right I have been thinking about it and decided that radiator order probably makes very little difference, my theory is right but due to the nature of the materials we use ie cooling water with air it just doesn't really make that much of a difference.

If a rad is dissipating 100W and the water flowrate is 1.5gpm then the water in water out delta is ~0.25°C, 200W = ~0.5°C and so on.

If you had an airflow rate of 50cfm and that was absorbing 100W then the air in air out delta would be ~3.5°C.

If you make the assumption that heat dissipation is linearly linked to air in water out delta (which from Martin's tests seems to be reasonable for this exercise at least) then if the first rad can dissipate 100W and raise the air temp by 3.5°C (say air in 20°C and water in 30°C delta first rad 10°C) then the second Rad the delta is 6.25°C and so the second rad can remove 62.5W (oh the joys of picking your starting figures).

So in theory as long as you can provide sufficient airflow to the rads to overcome the additional pressure drop then you should see a decent increase in performance otherwise you saturate your air and it can no longer remove heat from the second rad. Guess it will all be a balancing game!

This may be helpful in figuring out radiator order. This is the XSPC RS360 which is similar to the magiccool rads I believe, I think they are both a bit narrower than the MCR series (120mm vs 130+).

Anyhow, I recorded both air in and air out:
http://www.martinsliquidlab.com/img/...0-Thermal1.png

Somewhat surprisingly I used fairly consistently only about 50-55% of the air's capacity regardless of fan used. The more dens radiators were getting upwards of 80% or so depending on the actual air flow.

This appears to make sense in term of the additional 30-40% you gained stacking them.

With that I can only assume with the double thickness rads like a TFC, your net gain from sandwiching would be much lower because you've already used up nearly 80% of the air's cooling capacity with the lower fan speeds. Here is the TFC 480:
http://www.martinsliquidlab.com/img/TFC480-Thermal1.png

It shows the percentage of air capacity used was in the 80% range for low speeds down into the high 60% range for higher speed fans. That would mean stacking them probably is only good for about 5-10% gain low speed and maybe 20% or so with high speed fans.

It would be interesting to see though. Who's going to stack two 480 GTX's with nine mega high speed deltas and see how much heat they can dissipate..:D

I like the two stage cool order myself, second rad getting the coldest air in, but I wouldn't be suprised if in the end either order turns out pretty much the same. The biggest difference in stacking of the rad gain will be in the rad itself and fan combo, higher speeds should gain more, but it really depends on the rad. I wouldn't expect the double thickness rads to see much gain in stacking though, not enough air capacity left.

I'm curious how these even thicker rads planned to come out will perform. I wouldn't expect 100mm worth of rad thickness to be worth more than 10% over 50mm of rad thickness. The biggest gains are always in the frontal area, that's why many have gone to slightly over sizing the width or fan spacing, typically one extra row or around 130mm wide.

If I was going to design a rad, I would make one extra wide, but slim thickness. Maybe 160mm wide or as wide as possible to fit in most locations, and build a shroud into the rad for the extra shroud gain. I would also make the plenum chambers double side ported so they could be easily sandwiched. Material costs would then be relatively low, but performance would be fairly high and the stacking ability would make them modular. Maybe some sort of SLI specific fittings to sandwich them an easy means to attach both rads together with 25 or 38mm fans.

It's threads like this that might get them thinking more along the lines of modular sandwich/SLI type radiator setups instead of going extra thick.

Really great test. I'll be looking forward to the difference in temperatures between the sandwich and 2 separate radiators. My guess is the difference will be less than 1 degree. At which point I'll slap myself for not thinking of doing the sandwich myself instead of bleeding and sweating, trying to get 2 separate BIP's into an unfriendly, mid-tower acrylic case. Although future watercooling upgrades now seem less painful for me. Thanks again.

Very interesting stuff. I am just now testing a Black ICE GT 120 vs. a GT 140. I want to find out how much benefit you get from that extra radiator width. According to Martin's theory, I might see some nice gains!

Another important question!!

Can stacking reduce the fan noise as well???? Yes/No

pls check as well .. thx :)

Ok, I'll try to check that. Though I don't have any instruments for measuring noise levels. I'll have to judge purely subjectively...

I reckon the testing should find it's way towards the sticky here in the Liquid Cooling section.

The Rad i'd like to see come out of the Rad Mfg'ers is a Rad that will use the newer 250mm fan as those fans move alotta air & are quite for the amount of air they move & with the Mountiion mods & the like Cube Case systems this would be a great option as u could use the side walls as the rad systems & cover with a high pass thru air filters killing the dust & hide the rad all in the same unit & quite to boot ! Power consumtion would be less as you would have less motors . Even a Tower case could have the 250mm rad/fan system in the side covers .. my Xclio case has twin 250mm fans on it's side cover & with the filter mod i did it to this day has never seen a bit of dust in it & the air blows directly on to the MoBO ..
with a simple shoud system & a rad/fan setup to fit a twin 250mm fan high flow rad would cool quite nicely

http://www.abload.de/img/08x07u.jpg

Parallel radiator fail?

What do you guys think, will this work? How can I make sure that both rads are getting equal amounts of flow?

And yes, I am reviving this thread because yes, I have finally started testing radiator sandwiches again. :)

I was thinking about doing this for my htpc with two dual80mm. I might give this a try here soon. thanks for all the work you put in here guys.

Quote:

---

Originally Posted by HESmelaugh

Parallel radiator fail?

How can I make sure that both rads are getting equal amounts of flow?

---

I don't think you can unless you have the exact same routing to the radiators.
And even then it will take the path of least resistance.

Put the radiators facing eachother, and then hook them up with that koolance splitter (I'm assuming that's what it is)

And yes, more rad stacking tests :up:

This splitter is too long (unless it's modular) and so is this one.

But now that you mention it, I remember what I posted about how an ideal Rad-connector would look like:
http://www.abload.de/img/radconnectorcopy1xcn.jpg

And then I suddenly realized that I might be able to make just that with Delrin Ts and two of these! Thanks for the suggestion!

I'm testing the setup in the above pic currently, but I'll order the necessary parts and then retest with the simpler connection. :)

Quote:

---

Originally Posted by HESmelaugh

http://www.abload.de/img/08x07u.jpg

Parallel radiator fail?

What do you guys think, will this work? How can I make sure that both rads are getting equal amounts of flow?

And yes, I am reviving this thread because yes, I have finally started testing radiator sandwiches again. :)

---

I wouldn't do that. You want the rad that is closer to fresh air to be last in line, since the other rad will be receiving heat dump from the first. The rads should be in series, not parallel.

--Matt

Thanks, Matt. However, I'm only doing this setup for testing purpouses. I'll test several different options, including serial setups. And of course, I will post the results here. :)

Basically If you want to ensure equal flow through each rad you need to make sure that the path to each is the same so currently as you have it setup the rad on the left will be getting slightly less flow than the one on the right, how much will depend on what the difference in pressure drop is over the connections from the splitter to the rad and also the pressure drop flow relationship of the rad.

Basically for parallel flow a rotary T (as you suggest with the Koolance sliding fittings) is the ideal solution.

If I may make a suggestion with the testing measure temperatures in and out of each rad for the water and air.

My 2x PA160's which are stacked (soon maybe to be unstacked if I can fabricate a suitable new shroud) under load I see maybe a 5°C temp change over the first rad in air temp and then only a 1°C over the second, now this is potentially because I am not loading the rads up to anything like their max (I have a water out air in delta of ~5°C) but still it shows how much difference the hot air makes on the second rad.

Perhaps I should sandwich my 120.3 & 120.2 together :)

Once again nice testing HES, looking forward to the new tests.