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!