More Radiator-Sandwich testing

[NaeKuh]

Yes, NaeKuh, the same identical heater was used for every test. It has an actual heat output of 289.9W, monitored per Watt-meter.

Don't worry.

i wanted some accurate messurements from stacking since i see it being our next Big mirgration.

[Vapor]

Is this with an "ideal" fan, or something? Because I know a lot of fans perform less efficiently at higher RPMs, when you do a direct scaling from lower RPMs.

This is a property of all axial fans.

If there's any 'flaw' with it, it's at lower RPMs where measuring static pressure and airflow becomes increasingly difficult and therefore applying the property/rule becomes hairy to verify. Highish RPM fans (above 1500RPM) stick to this nearly 100%, in fact, Sanyo Denki uses these rules for rating their fans exclusively (they test at a given RPM and then all variations of the fan have their ratings derived from the first via this method).

What fans don't scale with RPM? I can't think of a single one

EDIT: anyway, it seems there's two things with the sandwiches that need attention: increased restriction and already-high air saturation levels. In Martin's and Skinnee's rad reviews, low RPM air saturation levels are up above 75% it seems. So at best, the 2nd rad can only really 'use' 75% of the remaining 25% (of air temperature headroom)....and that's if there's no airflow loss. Sandwiches might be more airflow loving than the GTX series (which really are a sandwich of sorts anyway, heh).

And FWIW guys, there's basically no low RPM fan that also has high static pressure, just like there's no low RPM fan that has high CFM.

[Sadasius]

Wow this was an eye opener for me. Then again I favor the 'other' white meat or other half of the equation which is fan power. I hate cheap fans...okay maybe not cheap but low performing fans. I have no use for them. If I cannot hook them up to a controller and be able to go from sleeping to having a 747 landing in my room then I have no use for them and they end up in the garbage. Static pressure and CFM are the other half of the equation, not just the rad. I see people spend big bucks on a rad but then turn around and put cheap fans on it. Makes no sense to me. Get a good rad and good fans that can have the power when you want it and you have it made. You can settle thing down (for those with dog hearing) or crank them up (for those who are half deaf like me). There is a reason why San Ace is an expensive fan. Because you are getting what you pay for.

I bet if you did another similar test with performing fans you will see things turn around. Good testing though and thanks for the time for this. I was very surprised the sandwiches did not do better.

[Zehnsucht]

I was very surprised the sandwiches did not do better.

Same here. I hope that ultra kazes 2000 will improve the numbers. That's what I'm ordering anyway

[Sadasius]

Same here. I hope that ultra kazes 2000 will improve the numbers. That's what I'm ordering anyway

If you use a controller just get the 3000 rpm one's. That way you can control it and if you ever wanted to do a sandwich thing you have the capability to do it. I was thinking of doing a sandwich myself actually with 2 480GTX's with fans on both sides and in the middle with some custom air cones to shape the air into the inlet fans to get the most velocity from them. Would probably use delta's though as I think I would need a higher rpm range.

[Jhors2]

If you use a controller just get the 3000 rpm one's. That way you can control it and if you ever wanted to do a sandwich thing you have the capability to do it. I was thinking of doing a sandwich myself actually with 2 480GTX's with fans on both sides and in the middle with some custom air cones to shape the air into the inlet fans to get the most velocity from them. Would probably use delta's though as I think I would need a higher rpm range.

and I thought I was crazy, lol.

[Zehnsucht]

If you use a controller just get the 3000 rpm one's. That way you can control it and if you ever wanted to do a sandwich thing you have the capability to do it. I was thinking of doing a sandwich myself actually with 2 480GTX's with fans on both sides and in the middle with some custom air cones to shape the air into the inlet fans to get the most velocity from them. Would probably use delta's though as I think I would need a higher rpm range.

I'm just about to pull the trigger on buying a lot of parts from PTS, including the stack. I'm hooking up the fans on a controller, yes, the NZXT SENTRY LX. Decisions decisions...

[Sadasius]

I'm just about to pull the trigger on buying a lot of parts from PTS, including the stack. I'm hooking up the fans on a controller, yes, the NZXT SENTRY LX. Decisions decisions...

You know there is a funny side too. When you go to a LAN party crank up all your fans to full speed. They will sit farther away from you giving you more room!

[zlojack]

Wow...well now I'll have to test mine and see what results I get when I get them in. First just one rad, then adding the stack.

[Zehnsucht]

Just ordered a 220 stack plus kaze 2000's. I'm looking forward to read test results from other members

[gillbot]

you need more airflow.

[Martinm210]

Superb testing as always!

The other variable that may or may not make a differences is using an oversized radiator.

I think the magicool is similar to the XSPC RS120 in that they are more narrow and have less frontal surface area than some radiators like the MCR series. The MCR series have a good 10mm more width which equates to a pretty big frontal surface area advantage. This will play a part in reducing the air pressure drop for the radiator.

Also while I never got the equipment together that was sensitive enough, the actual fan used will make a difference. Each fan has a PQ curve similar, but different than that of a pump. It's different because those curves have very pronounced peaks and bumps in them. It's kind of tricky, but it's very possible the top performing setups just happen to fall inline with the fan PQ curve where it's optimized and other setup may fall in the less desireable regions.

I'm not really sure how you could really determine this short of measuring the PQ curve of the fans at those RPM vs the pressure drop of the rad (which would take some nice equipment (flow chamber) to do, other than perhaps running the test with more than one fan.

Anyhow, fans and air flow pressure drop is alot trickier than I had thought. When I did some prototype testing for XSPC on the RX series, I was pretty suprised what happened with changing the fin densities. I always assumed more density would provide the strongest results with high speed fans, but it wasn't the case. Trial and error was really the only way to find that sweet spot for that particular fan and it's likely that optimum for one fan may not be optimum for another. I have to blame the irregular fan PQ curve for that occurrence and I've seen that in testing.

Great work though, very interesting information.

[Bond Number]

Excellent test, along with stronger fans my only other critique would be to account for flow rate. In the single radiator case you probably have a higher water flow rate than in the dual radiator setup. The only reason why I bring this up is because it might allow people to extrapolate data for other pump setups.

You could adjust the voltage on your pump so that the flowrate was the same for each config or take the flow rate dependence into account by plotting dimensionless delta T’s versus your configuration (include flow rate). The dimensionless delta T is basically your measured delta T divided by a delta T which is only a function of flow rate. This delta T could go as Heat load/ (flow rate * heat capacity). (Of course I am assuming that your heat capacity is not a function of T, which is a good approximation for water @STP with small temperature changes and that your heat load is constant in all experiments). So if you use distilled water the flow rate delta T would be approximately:

Load (W)
_________________________________________________
Flow rate (L/s)*56(mol/L)*4.18 (J/(mol*K))

[Martinm210]

It's pretty tricky to try and account for flow rate. You would need a fixed flow rate chiller with very precise ability to record temperatures, probably down to .001C to even begin to measure the effect of flow rate on a radiator. The performance difference gets skewed very quickly by the pump's heat dump effect. I wanted to do that, but gave up and just started testing everything at 1.5GPM as a fixed number. Not that that's right though, just one method, actually I like the "Fixed pump method" better...

It's actually pretty helpful to instead as done here and test with a fixed pump. This then naturally takes into account the pressure drop differences and pump heat dump differences that a user would experience when switching between one setup or another, so it's actually more of a "real world" method than the fixed flow rate idea. What I've found is the difference is just too small on radiators to even bother with flow rate. As long as you have 1GPM+, it's well within the very flat part of the curve.

BUT...it is important on block testing and why I'm an advocate of fixed pump testing. Using fixed flow rate comparison testing on a chiller will inappropriately skew the number more favorably to the block that is more restrictive, etc.

Anyhow, my 2C I wouldn't bother with trying to measure flow rate effect. It would be nice data to measure and see, but pretty darn hard to measure without some very expensive equipement and for practical purposes so small it's not going to add up enough to make the effort very worthwhile.

[Bond Number]

Excellent test, along with stronger fans my only other critique would me to account for flow rate. In the single radiator case you probably have a higher water flow rate than in the dual radiator setup. The only reason why I bring this up is because it might allow people to extrapolate data for other pump setups.

You could adjust the voltage on your pump so that the flowrate was the same for each config or take the flow rate dependence into account by plotting dimensionless delta T’s versus your configuration (include flow rate). The dimensionless delta T is basically your measured delta T divided by a delta T which is only a function of flow rate. This delta T could go as Heat load/ (flow rate * heat capacity). (Of course I am assuming that your heat capacity is not a function of T, which is a good approximation for water @STP with small temperature changes and that your heat load is constant in all experiments). So if you use distilled water the flow rate delta T would be approximately:
Load (W)
__________________________________________________ _______
Flow rate (L/s)*56(mol/L)*4.18 (J/(mol*K))

[Martinm210]

It's pretty tricky to try and account for flow rate. You would need a fixed flow rate chiller with very precise ability to record temperatures, probably down to .001C to even begin to measure the effect of flow rate on a radiator. The performance difference gets skewed very quickly by the pump's heat dump effect. I wanted to do that, but gave up and just started testing everything at 1.5GPM as a fixed number. Not that that's right though, just one method, actually I like the "Fixed pump method" better...

It's actually pretty helpful to instead as done here and test with a fixed pump. This then naturally takes into account the pressure drop differences and pump heat dump differences that a user would experience when switching between one setup or another, so it's actually more of a "real world" method than the fixed flow rate idea. What I've found is the difference is just too small on radiators to even bother with flow rate. As long as you have 1GPM+, it's well within the very flat part of the curve.

BUT...it is important on block testing and why I'm an advocate of fixed pump testing. Using fixed flow rate comparison testing on a chiller will inappropriately skew the number more favorably to the block that is more restrictive, etc.

Anyhow, my 2C I wouldn't bother with trying to measure flow rate effect. It would be nice data to measure and see, but pretty darn hard to measure without some very expensive equipement and for practical purposes so small it's not going to add up enough to make the effort very worthwhile.

[nikhsub1]

Thanks for the testing, I know how time consuming and tedious it can be. You really spent a lot of time on this and it is a great service to us all! I won't say I knew but it turned out just as I thought it would. You said that don't assume rad stacking is bad, here is where I disagree. It is bad. I have always said it is bad. Sure you may find an odd combination where it works OK, but fans are much like centrifugal pumps... if you choke them they don't work well at all, just as your tests have shown. Maybe, maybe if you have some uber air free flowing rads (think 5-7 FPI or something) and play with fan direction... As a general rule though, never stack rads it's just not smart. Until someone can prove (and who better than you?) an exacting combination that TRULY works well, just say no. Again, thanks for the testing!

[Hondacity]

yeah we appreciate the testing...

now i know what to do with my extra$

[Taylor01]

Wow! And I was ready to buy the stacked mcr320's! If I can't put the with low RPM fan and you figure a 320 stack is $75 and a reg. 320 is $50 = $125 Why wouldn't I just buy a Feser for $130? I don't know what swiftech's tring to pull here?

[Jhors2]

As Gabe said people requested this very thing, its especially advantageous for people who already have one of their radiators. It works well with either high RPM fans, or fans with a pretty good static pressure rating. Its mainly for space saving, obviously having the radiators apart are going to yield best results.

[Hannibal Lecter]

I thought that the stacked MCR320s run in parallel, contrary to the test so I believe those results cannot be directly compared.

[Taylor01]

As Gabe said people requested this very thing, its especially advantageous for people who already have one of their radiators. It works well with either high RPM fans, or fans with a pretty good static pressure rating. Its mainly for space saving, obviously having the radiators apart are going to yield best results.

And still a single rad. on its own (not two apart) performed better than two sandwiched together. I'm sorry I feel that the manufacturer is try to pull something. Why don't the post test on there sit for the stacked rads. They have test results for everything else?

[Jhors2]

Well the tests arent entirely conclusive until you get a full test with bigger/thicker fans with higher CFM....I'm curious to see how it scales with some thicker/faster fans.

[Bond Number]

Sorry, I was not clear in my post. I agree that a fixed flow rate test is not the way to go. I was trying to imply that by following the fixed flow rate test with another solution. A dimensionless delta T, where the delta T is non-dimensionalized with respect to the flow rate. This dimensionless delta T is a better measure of the radiator or series of radiators overall heat transfer coefficient. This way you also maintain all the benefits of having a fixed pump test, but give people the option of extrapolating.
I also agree that the variance in flow rate between different radiators is probably small. But when you are comparing different configurations like a single radiator to 2, 3, or 4 radiators in series, flow rate may be of importance. I understand that this test was with only two radiators, but in the future if someone wanted to compare it to series of 3 or 4 it might be nice to compare data.
As for measuring the flow rate one could perform a separate, cheap test on each configuration. Simply fill a large tub with water then pump the water through the configuration. On the other end record the time it takes to fill a known volume. Density changes between the two types of experiments (Delta T and flow rate) would be small. If you really wanted to, you could record temps in both types of experiments and convert flow rates, but probably not necessary

[HESmelaugh]

Wow, some very interesting feedback here since my last visit!
Thanks to everyone! I read all the posts carefully but I can't reply to each one fully.

The point Martin brings up is one I hadn't even begun thinking of. I checked the measurements from my last radiator test and saw that some of the radiators I had here vary in width between 12cm (Watercool) and 13.8cm (Thermochill). That's a considerable amount difference in the surface areas of these rads! So there's another thing that's almost impossible to take into account without actually testing sandwiching with a whole range of different radiators. ^^

Concerning flowrates: I did do some measurements on flowrates. It was very clear that parallel flow is hugely beneficial to flowrates. The reason I did some flow-rate testing was to find out whether it had an effect on the results. Ie. whether the differences might be because of the flowrates rather than air pressure drop.
I consider my testing equipment to be fairly precise, but I couldn't measure a difference in temperatures between about 1.5GPM and well below 1GPM all other factors being equal. So I just assumed that it didn't make a significant difference for radiator testing.

Finally, I want to address the Stackable MCR thing: I really don't want people to jump to conclusions about the MCRs because of these tests. First of all, keep in mind that the MCR have lower FPI count than the radiators I used. Second, keep in mind that I tested only up to 1200rpm, which seems to be considered a medium or even slow fan speed by many here.
It could have been different: I could have come along with some tests between 2000 and 3000rpm (assuming I had the fans for it) and maybe some spectacular results for the sandwiched rads. The point is this: All the test really shows is that air pressure drop matters a lot (a lot more than I had thought). If anything, I want to encourage people to test their own setups and combinations. I don't publish these tests in an attempt to "preach the truth". As you know if you've spent some time with liquid cooling testing, there are simply too many variables for there to be "the truth". This is why the guys in Germany who have everything running in a single loop with ultra low speed fans and relatively weak pumps can say that flowrates don't really matter (and they are right) and the guys on XS with several loops and several pumps each can say pressure and flow are super important (and they are also right).
There are a million ways to make an LC-setup work. And there are another million ways to make it work a bit better. Try it out for yourself.


P.S.: @Bond: I want to try if I can use your formula and see what results I get. Don't know if I have enough data for it, but I'll give it a shot. Thanks for the suggestion!