Table of Contents
Does fan stacking really live up to all the attention it gets? It seems like every once in a while, heatsink vendors look for something to hype sales and come up with stacked fans. Intuitively, placing one fan on top of another makes sense – “doubles the airflow” is what seems to be the common rationale.
Fan Stacking
Sorry folks – the truth is a lot harsher. Fan stacking can work, but only under certain conditions; if these conditions are not present, fan stacking can actually result in worse performance. Even if the necessary conditions are met, don’t think you can get twice the airflow – if you’re lucky, MAYBE an additional 20%.
As I understand the theory behind fan stacking, the essential problem is that the air leaving the first fan’s blades comes off at something like a 45-degree angle. Unless the second fan can take the air in undisturbed at that angle, all you do is create a lot of turbulence and noise, with little or no performance boost.
The ONLY way you get a performance boost of 20% or so is if the second fan is rotating in the OPPOSITE DIRECTION to the first (thanks to Andy Lemont for the primer).
Now to show how, in fact, these principles play out, I set up a test as follows:
Using the same heatsink, I used a Papst 33 cfm fan (which rotates counter-clockwise), A Delta 38 (clockwise), and a pair of YS Tech 26 cfm fans (also clockwise). Not many fans rotate counter-clockwise; the Papst was the only one I had.
I then ran the fans in a variety of positions, one fan and stacked, to show what performance gains are possible. I did this on an ABIT KT7 running Prime 95 with a Duron 800 @ 1000 MHz, 1.9 volts. I used the Omega HH23 to measure temps and calculated C/Ws for each condition.
Test Results
CPU Temp | Ambient Temp | Delta | C/W | MBM Temps | |
Papst 33 + YS Tech 26 | 33.7 C | 22.2 C | 11.5 C | 0.22 | 39 C |
Single Delta 38 cfm | 35.1 C | 22.9 C | 12.2 C | 0.23 | 41 C |
Single Papst 33 cfm | 33.7 C | 21.3 C | 12.4 C | 0.24 | 40 C |
Stacked YS Techs | 39.7 C | 23.3 C | 16.4 C | 0.31 | 45 C |
Single YS Tech | 39.3 C | 22.4 C | 16.9 C | 0.32 | 45 C |
C/W = Delta / CPU Watts
Interpreting C/W: For every watt the CPU radiates, the heatsink will cool the core by the (C/W x watts) plus ambient temp. For example, at an ambient temp of 25 C, a C/W of 0.25 with a CPU radiating 50 watts means that the CPU temp will be 50 x 0.25 = 12.5 C over ambient temp, or 37.5 C. The lower the C/W, the better.
The best cooling is with a single clockwise rotating YS Tech on top of the counter-clockwise rotating Papst. Bringing up the rear are the YS Techs, both spinning in the same direction. Note the minimal performance boost (Delta column) of the stacked YS Techs compared to using just one (0.5 C); compare that to the single Papst compared to the stacked Papst gain (0.9 C). In truth, pretty minimal gains, but less so for doing it “the wrong way”.
I included the Delta 38 as a “check” to the stacked Papst gains. Stacking is 0.7 C better; hardly worth the effort, not to mention increased NOISE. Certainly not a gain you could justify on price/performance.
Conclusion
Buying a heatsink that features stacked fans is great for the vendor and a poor deal for the buyer. This is a clear example of where more is NOT better.
Sometimes you write something and it strikes a nerve; fan stacking is something a lot of you have tried with varying results. Read below for their feedback and responses
Jeff emailed:
One trick that you did not try, which might be of some interest to me and others, is adding a chamber between the fans. I think you might be able to use a 2″ long piece of PVC pipe for this, it should help “straighten the air out”. This should increase the performance of the two YS-techs as well as the Papst+Delta combo.
I thought this deserved a shot, so I used a fan body (no fan, just the rim); very tall assembly
with the following results: I dropped CPU temps by 0.2 C with the spacer in place. It helped a little, but still (IMHO) not worth the effort.
However, others seem to have better results, so read on and maybe try out some of these ideas:
From Daniel:
I just read the article about the stacked fan scheme used as a marketing gimmick. Stacking fans is not so efficient but running two fans onto the same heatsink can be a good thing. Just angle each fan at about 45 degree angle pointing in on each other and then block the air from coming out the sides.
From Kevin:
One way around the turbulence problem would be to place a fan either side of the heatsink, with a “lid” to create a tunnel effect, this should reduce the turbulence as the airflow is straightened out by the heatsink fins. I agree that this will not give a major performance boost but may just serve to increase noise.
From “Old Man” (Hungary):
I can’t agree with the double-fan article. I tried a lot of positions
between the fans. You can try it out: take one fan and aim it at your face so you can feel the wind. Place the second one after the first, within 2-3 cm (depends on the fans_. You will get much better performance. The noise is much lower than one FOP32, the performance is better with 4 degrees celsius.
ED NOTE: I wonder if the CPU cooling observed is more due to airflow cooling the motherboard.
From Alan:
Quick point: Any company that sells coolers with stacked fans doesn’t enclose the space between them, for example the old Step Thermodynamics coolers [ED NOTE: Not quite so – some do]. They weren’t like one duct, with a fan at the bottom and fan at top. They used one fan, then another separated from it. Try that. I placed a fan on top of my Gorb and dropped the temp; then I got an FOP38 and SERIOUSLY dropped it.
From Randy:
I stacked 2 26 cfm YS Techs; the one closest to the CPU jumps from 4600rpm to 5200 rpm, and the top fan jumps from 4600 to 4800 rpm. Now if the fan on the bottom is running at 5200rpm = 30cfm or something like that, then does that mean I have 30 cfm of air moving to the heatsink?? I may be off by a 1-2 cfm, but I get a few C cooler also. Why would the bottom fan be spinning at a higher cfm if it wasn’t spitting out more air through its blades?
From Larry:
I just read your evaluation of the dual fan approach. I also read your comments on the new heatsink with the fan with the vanes. I suspect
that using two fans with one turned off may be better than having both running. A fixed fan would do much the same as the vanes. I can’t see the
orientation of the vanes from your picture (ie are they in the same or opposite direction?) [ED NOTE: Opposite]. I suspect you want it in the opposite direction to straighten out the airflow. Also a slightly thinner (or one with less curvature to the blades) dead fan is probably better. But I suspect that that a dead fan may be the champion.
From Jerry:
Funny, I just tried this same thing last night with very different results.
I stacked a Delta 4700 RPM fan on top of a SANYO 5100RPM fan and got an effective reading of 6400RPMs on the bottom SANYO from the mobo header (KT7, using VIA Monitor and MBProbe). It also lowered my highest fully loaded temps by 4C from 51C on a Athlon 1200 @ FSB+10 to 47C (MBProbe).
I thought the noise excessive but not terrible. I used electrical tape to stack them and agree that it would not be good long term. I have the fan for a FOP38 coming to upgrade my FOP32 and was curious to see what this rig would do while waiting on the mail.
From Thomas:
I tried fan-stacking on a Swiftech MC-462 with two YS Tech 80mm 3500rpm
fans. The rotation speed of the lower fan increases 300rpm (curious
???). The temperature of the Athlon 1200 goes down 3 degrees Celsius.
Perhaps the result of the stacking has something to do with the sort
of fan used??
From Pablo:
I have played with some fan stacking in the past; this was done with two 80 mm fans that ran in opposite directions. The results were very very impressive.
I have not done the research as to how much CFMs are put out by the single fans, but placing them together (stacked) was much better than two fans (stacked) that ran in the same direction.
I mounted them on a cheap (small) heatsink with some ducting just to play with. If there were a Black Delta that ran in the opposite direction than the regular ones, I would stack it in a second on top of mine. I believe that would be killer CFMs from 60 mm fans.
From Karl, aka space, from Norway:
Just wondered if the attached picture I made would be of any interest. In it I try to describe what happens when stacking fans vertically or in parallel.
When stacking two fans, the theoretical max airflow will be the same, but theoretical max pressure will double. From what people are saying about how small a performance increase they get, one could guess the airflow with just one fan is closer to max airflow than to max pressure* (Green x on “single fan” in picture).
So stacking a fan doesn’t help much. Adding a fan in parallel would more likely help, but the airflow will far from double. This is because to double the airflow through an obstacle, one must also double the pressure (so actually 4 fans are needed).
What I’m telling here doesn’t take into account that stacked fans produce more turbulence, as you wrote in the review.
*: From looking at some real charts of fans that typically are used, this is also the “area” that the fans are intended to work in.
From JohnT:
To understand the effects/advantage of stacking fans, one must look at the fan curves for the fan(s) involved and know where on that curve the fan is operating.
Most manufactures of fans publish a curve of airflow vs backpressure for their fans. This shows how many cubic feet of air the fan will move against a given amount of backpressure.
Any heat sink will require a certain level of pressure to be developed by the fan to move a given volume of air through the heatsink – this is known as the load curve of that HS when it is plotted for several different flow rates. For normal operation, this load curve will be parabolic in
shape: To flow twice as much air through the HS will require the fan to develop four times as much pressure (and require the motor of that fan to develop 8 times as much power).So, whether a second stacked fan will make a difference is a direct function of where it is operating on its curve. For example, if the backpressure is low, the fan/HS combination will flow almost the rated no-load rating of the fan.
If however, the fan is heavily loaded, i.e. flowing only about half its rated cfm due to the high backpressure, then adding a second fan will indeed almost double the airflow.
However, note this flow rate will be less than or equal to the rated flow rate of one fan – not double that rating. If the RPMs of the fan were constant, then the net maximum flow with two fans (or evermore) in series will be equal to or less than the rating of one fan alone.
As mentioned by some of your readers, the RPMs tend to increase slightly (due to the lower torque load on each fan), and thus the flow of two fans can be slightly higher than the rating for one fan.
I have run some tests and measured the backpressure with a manometer and can supply some pictures if you are interested. I tested stacked fans (as many as three), parallel fans, and even a parallel configuration with two fans stacked (four fans total).
I also tested the small blowers made by Nidec and found these to give the highest performance (because of their greater flow rates at high backpressure). I currently run a pair of these blowers (Gamma 30) back-to-back on a Hedgehog copper hs for excellent performance without the noise of fans like the Delta.
They are rated for a flow rate of 25 cfm each at zero backpressure and will still flow a significant amount at .5 inches of water backpressure.
From Thomas:
Heya, nice article on using double fans. I have found one application where it was actually very useful. This is because the airflow may not improve (maybe even degrade), but the pressure exerted is enhanced.
Basically, what I had was a duct from a case fan to the CPU heatsink. The problem was that I had a lot of positive pressure in the case, and the 80mm duct fan could not push enough air so the heatsink got real hot. I added a second fan on top of the first, and it worked like a charm! if you want to read more about it HERE.
From Kevin:
I read your article about fan stacking. One way around the turbulence problem would be to place a fan either side of the heatsink, with a “lid” to create a tunnel effect. This should reduce the turbulence as the airflow is straightened out by the heatsink’s fins. I agree that this will not give a major performance boost but may just serve to increase noise.
From Monza:
In response 2 your fan stacking article, I say scrap all that; just make a shroud (like an intake manifold airhorn on weber carbs) that envelopes the top of your heatsink. Then open up the other end to a 120mm fan sealed at the other end (can be an existing side fan if applicable). Let the 120 blow in on the heatsink, ditch the delta, and go with it.
I did that and my Vantec heatsink (the new one with the cool clip) has never been cooler, and it’s quieter than the Delta (120 is a Nidec or Sunon 70 cfm piece). It beats the Delta by 5-7C at idle.
From Matt:
In response to your article on fan stacking this is the experience I had with my GORB:
I removed the OEM fan and replaced it with a clockwise rotating 7500 rpm 40mm box fan I managed to find at one of my local surplus electronic supply warehouses. After removing the 40mm fan’s shroud, I used a couple of drops of super glue to attach it firmly to the GORB where the OEM fan originally sat. I then used a standard counterclockwise rotating 60mm box fan and attached it directly atop of the GORB.
It was easily held in place by inserting a screw in each of the mounting holes of the box fan and then tightening them down, forcing them in between the cooling fins of the GORB, so both fans blow air into the GORB. It made for quite a dramatic drop in CPU temps compared to any other fan set up I had experimented with on the GORB.
At idle, the CPU temps averaged about 6 F above ambient air temps; under load, the temps only rise another 6 to 7 F over idle. Please note: The GORB used had been previously lapped from other experiments with different cooling fan arrangements. Also, I removed the side panel that has 3 80mm Panaflows attached to it to more accurately get an idea of the performance lost or gained.
The one thing I noticed with the counter-rotating fan set up is the peculiar sound it creates. Although not that loud, some may find it bothersome.
From Dana:
With fan stacking the rule “What goes in must come out” applies (conservation of energy). Stacked fans of the same or similar rating will do little to nothing.
In our business, we use industrial regenerative blowers. They are axial type fans that first compress the air before blowing it out. This increases the total CFM. I don’t know how to reproduce this effect, but it is what fan stackers will have to investigate if they want to really increase CFM and seriously cool their chips. This well could provide good cooling but could also generate waste heat from compression and definitely noise.
From Roger:
Basic Fluid Mechanics (Mech Eng.) gives the clue to what is happening:
- Stacking fans will not double the airflow, all you are doing is passing
the same amount of air from one fan to the other.- Stacking fans increase the static pressure capability of the fan system. What this means is that you now have double the pressure which means the air has a greater ability to force it’s way past the fins.
- The lower fan speeds up because it is receiving air at a higher pressure and has to do less work to move the air, thereby allowing it to speed up.
- If there is a lot of resistance to the airflow then stacking the fans will result in greater airflows are subsequently lower temps.
- The biggest culprit in the airflow game is unbalanced airflow in & out of the case. If you put in too much air you create a high-pressure zone in your case and the CPU fan has to fight this pressure to get the air across the fins. This is why removing the sides of the case result in lower temps.
Lots of good feedback! Fan stacking can work under the right circumstances but don’t look for double the airflow – theoretically impossible. I did find some benefit, but minimal CPU cooling impact. Also, monitoring temps with socket thermistors may be picking up secondary heat path effects rather than the impact on CPU cooling. And finally, give it a try – nothing like a little experimenting to see what happens. Thanks to everyone who sent in comments about Fan Stacking. Obviously lots of interest, so please send in your experiments to share with others.
No replies yet
Loading new replies...
Member
Join the full discussion at the Overclockers Forums →