Ninja Airflow Testing....Tracing Turbulence Noise.

Bluefront · Post by **Bluefront** » Sun Oct 14, 2007 5:52 am

I hope this isn't getting too boring.....but there are many different possible configurations to this test. I'm only showing a few.

Now you all know a 120mm fan should cool better than a 92mm fan....right? The problem with a Ninja is with the OEM 120mm fan mounting, much of the airflow is wasted. The only way to make use of this wasted airflow is by the use of a duct. Here's one way of doing this....there are other ways.

The idea is to gradually increase the Ninja dimensions to match a 120mm fan dimensions. If you do the dimension change too abruptly (like with a 92/120 adapter), you will lose much of the benefit. Here's the numbers at the same ambient as the last test.

120mm low-speed Yate Loon......45 CFM

@12V.....1dle.....50%.....75%.....100%
.............28C.....32.........35........37

@5V.......28C.....34.........36........38

This works out to a CPU temp improvement of 3C @100% over this same fan mounted directly on the Ninja, at 12v or 5v. I consider this significant, considering how much trouble I had gaining even a one degree improvement with the other duct. And the numbers are a little better than the 92mm Yate Loon with much tweaking to the other duct. Perhaps with some tweaking, this duct could do better. The 92mm YL and this low-speed 120 YL are very close in CFM.....42/45.

Conclusions....if you want to gain any benefits of the 120mm fan, over a 92mm fan on a Ninja, you'll have to devise a duct of some sort, to capture the entire airflow of a 120mm fan.

Scythe......are you listening?

Felger Carbon · Post by **Felger Carbon** » Sun Oct 14, 2007 10:37 am

With apologies: this is not about a duct, or a 120mm fan. But Bluefront discusses that 120mm fans are too big and 92mm fans are too small, so...

viewtopic.php?t=34467

ddrueding1 · Post by **ddrueding1** » Sun Oct 14, 2007 11:49 am

Nice duct Bluefront. I like the idea of venting out the back side of the case; this leaves the top/back for case ventilation fans, and is the shortest route a CPU exhaust could take in most cases.

I've been thinking of ducting my Ninja to the 200mm fan on the top of my Antec Nine Hundred. Even on low it does 82CFM!

My fear with ducting has always been the surrounding components. What is the airflow like on the way into the duct? Is any of that airflow going across the motherboard? Could a duct on the intake side encourage airflow past the mofsets, northbridge, or RAM?

Bluefront · Post by **Bluefront** » Sun Oct 14, 2007 11:56 am

Heh....Actually I think a 92mm fan is plenty to cool a Ninja. But if you're trying to use the airflow through the Ninja to cool other pieces in the computer, a larger fan would probably be better......a fan with more airflow.

There's one problem using a 100mm fan.....there's only the one brand to choose from (that I know about anyway). Plus the Ninja comes with a fairly quiet 120mm fan. Plus if you do buy a 100mm fan, you've got an odd size that might not fit anywhere else.

If I had a 100mm fan to try here, I would have......I don't. I've got dozens of different 120mm fans.....I used one of the quieter models I had in storage.

Bluefront · Post by **Bluefront** » Sun Oct 14, 2007 12:23 pm

Cooling the other computer pieces, in addition to cooling a Ninja......with a single fan......is a somewhat different subject. The Ninja design does not lend itself well, for cooling the mother-board. But it is possible.

This Ninja duct uses a rear 120mm fan blowing inward through the long black duct.. The airflow is internally divided......part goes upward through the bottom vent holes of the Fanless Zen PSU, part goes through the Ninja, part loops around and goes upward through the bottom of the Thermalright NB heatsink, and part goes toward the MB components between the Ninja and the I/O plate. Very little airflow makes it all the way to the Ram chips....but some does. The exhaust goes upward, and out the top of the case.

The airflow with this duct is not ideal for cooling the CPU.....but it does cool most everything inside the computer. The video card has it's own fan in this setup, as does the HD. This setup is making good use of the results of these tests.......though it was built a year ago.

cmthomson · Post by **cmthomson** » Sun Oct 14, 2007 5:04 pm

Now that's one elaborate duct.

Of course you and I will never agree on which direction to run the fan...

Bluefront · Post by **Bluefront** » Mon Oct 15, 2007 2:30 am

I usually build cases using extensive intake filtration, as you probably know. So I normally have fans blowing inward with the filters right behind the fan. But I think this last particular duct setup, would work almost as well with the one rear fan blowing outward.

It's impossible to show the complete duct arrangement, because this duct is composed of a number of smaller pieces, mostly held together by velcro. Since there are separate internal airflow paths, heat from the PSU would never go through the Ninja, with the rear fan either sucking or blowing. The benefits of ducting a Ninja to also cool other parts of the computer.....are largely unknown territory. Not many examples of this technique are around...

Anybody seen another combo duct for a Ninja?

Oh....the North Bridge cooling part of this duct is hard to see, and it gives another dimension to possible airflow routes with a Ninja. With this duct you have air blowing level through the back-side of a vertical Ninja. But the bottom of the duct also has a gradual curve that directs air upward from the bottom of the Ninja at the same time. This gives the over-all airflow an upward bias, causing the airflow to travel the "long way" through the Ninja.....from corner to corner. It works extremely well, considering the airflow from that single rear fan is shared by four separate paths.

Here's another view of the NB part of the duct. At the time of the photo I was trying out a duct arrangement for the video card that directed airflow upward also (the case exhaust vents are on the top of the case w/no exhaust fans).

And here's the final setup for this combo-Ninja-duct.

Three fans total....the rear Ninja-duct fan, an internal spot fan for Ram cooling, and the lower intake fan setup(inside another duct) that cools the hard drive, and blows air at the bottom of this fanless video card. No exhaust fans.....very quiet.

Bluefront · Post by **Bluefront** » Tue Oct 16, 2007 12:13 pm

I modded the wood duct in the previous photos, which I'm now calling a Wing Duct. This is different from your normal duct. It opens up the two sides of the first Ninja wood duct, adding two air channels that point at the mother-board. In effect, this turns a Ninja into a partial "blow-down" heatsink.

I've been measuring some other temperatures during this testing. This new duct lowered the usual ram temps 3-4C. And even better.....the upper left corner of the MB, measured 63C in places with the first duct. With this new duct, I'm now measuring <50C. Not bad. Of course this new setup has reduced the airflow through the Ninja, but does not affect the CPU temperatures more than 2C.....far as I can tell. My best temperatures with this duct had the fan on the bottom blowing upward into the Ninja, and at the MB. When I tried it with the fan on top, sucking out of the Ninja and off the MB, the temps were slightly higher. With the fan on the bottom blowing upward, the NB temps were also improved.

I have a computer already setup and running that I'll try out this new duct. I have high hopes for this new device.....which might benefit from some tweaking. Anybody want to buy the patent rights to the "Ninja Wing Duct".

ddrueding1 · Post by **ddrueding1** » Tue Oct 16, 2007 1:19 pm

Sweet. In the battle of getting an entire system effectively cooled using a single fan, you are certainly making headway. I'm still debating whether it is more effective to build custom ducts to route the air, or use custom heatpipes to move the heat. Likely some compromise will work best.

Bluefront · Post by **Bluefront** » Wed Oct 17, 2007 2:31 am

Well look.....for the average guy it is quite possible to build custom air ducting for his computer case. Hell you've probably got all the necessary pieces just laying around somewhere. The duct in these last photos was built out of scrap wood, some aluminum roof flashing, and a bottle of glue. For tools I needed a drill, a jig-saw, and some sand-paper. Estimated cost of materials.....<$1.

Could I have duplicated the effects with heat-pipes and custom heat-sinks....maybe. But I don't have such things, and certainly can't make them. Not to mention.....this duct can be moved from case to case, and will fit many different mother-boards. Try to duplicate that with heatpipes....

I like building things. Making ducts is easy.....had this one not worked out, I would have started over with no money lost. This last duct is a prototype...and no doubt could be improved upon by many SPCR members. Use it as a starting point, and improve it. For instance....it could be entirely made out of aluminum sheet metal, to match the looks of the Ninja.

Bluefront · Post by **Bluefront** » Thu Oct 18, 2007 8:34 am

More changes to the test procedure....some minor tweaking to the "wing-duct" and changing the CPU to the hottest one I have that the MB will run.....a P4-3.4 Northwood, which runs about 100W. This is just about as hot a currently available CPU as you can find. If this duct can handle 100w, it can handle anything.....it does.

Here's the setup I ran, with a full-sized fanless 7500 AIW which partially inhibits airflow to the fan.....not much however. The fan is a Mad Dog sleeve bearing, 92mm, rated at 40CFM @12v. The fan controller goes from 5.6V-11.2V. Ambient today is 26C.

I won't bore you with the page of results, but here is what it boils down to....

@5.6V, from idle to 100%.......36/37C----51C
@8.3V...................................35/36C----49/50C
@11.2V..................................34/35C----47/48C

Running two instances of CPUBurn.....the hottest setup I can achieve

@5.6V.........55/56C
@8.3V.........53C
@11.2C.......52C

Proves to me that this "wing-duct" can handle the hottest CPUs, at very low voltages, using 92mm fans rated at 40CFM or less. Put in a 92mm Nexus, the quietest available, and the numbers rise slightly. A hotter video card will skew the results, as will a heavily dampened case with low airflow. But the Ninja handles the heat, quietly, with low airflow.

Bluefront · Post by **Bluefront** » Fri Oct 19, 2007 2:33 pm

These different tests I've been running have proven very insightful to me. This latest test involves me simulating the effects of a case fan, either sucking or blowing at different angles to the Ninja fan.....trying to see if I can find the best setup.

Wow....here's the deal. If I adjust both fan speeds to a certain amount, and tilt the second fan at the right angle, I instantly hear a very loud air-turbulence noise. It goes away with a slight change to the angle of this second fan. The picture shows about the worst angle for noise. Hard to say what exactly causes this....maybe has something to do with the rotation direction of the Ninja fan. Blowing from the left side at the same angle could not duplicate this noise.

This might be the clue to figuring out strange unexplained airflow noises in certain case setups. Could you predict a noise such as this, without actually building the complete setup....doubt it. But it's something to keep in mind when tracking down air noises....

Bluefront · Post by **Bluefront** » Tue Oct 23, 2007 3:45 am

I started a new thread on an up-coming project using this new wing-duct as the only fan in a SFF case. Check it out for a new way to manage airflow with the minimum number of fans.....just one.

But to be more accurate with this new duct, I tried it out on a different mother-board with a different fan-control system....an Intel D865PESO. And I used a second Ninja I have.....just to see if anything changed. Nope....the temps are the same. The fan control on this board is pretty neat. The CPU header is not controlled, but the other two are. They apparently use the system temperature sensor (Intel temp zone two) to vary the fan speed. So I plugged the duct fan into the front fan header. Works....the system sensor reads very close to the CPU sensor, and lags behind only slightly. This gives the Nexus fan the ability to run at 4.5V (PWM control) most of the time.....even quieter.

As these tests have proved, it doesn't take much airflow to cool a CPU....as long as you use a Ninja.

cmthomson · Post by **cmthomson** » Thu Oct 25, 2007 3:52 pm

As you may recall from my 830D DIY article a couple of years ago, the Ninja with a simple duct can handle 150W of heat load with a slow fan.

So your 60W P4 doesn't even begin to tax the Ninja, as you've discovered. A 100W overclocked C2D similarly can be cooled with a slow fan on a Ninja.

Where then is the bottleneck? It's not in the Ninja. It's in the CPU. The heat path from the CPU's transistors to the air flowing across the fins goes through the CPU die (which is mounted upside-down), an internal TIM (thermal interface material), the IHS (integrated heat spreader, aka CPU case, which is a nickel-plated copper plate), the external TIM, the Ninja base block, the heat pipe walls in the base, the working fluid in the heat pipes, the heat pipe walls in the fins, and finally the fins themselves.

Because the working fluid in the heat pipes is so efficient (540 c/g to evaporate/condense water), the temperature at the two ends of the pipes is nearly identical. I've measured this on my Ninja, and although I don't recall the exact numbers, my recollection is that there is about a 2-3C difference, meaning that the Ninja base is only slightly warmer than the fins.

This points to the die/TIM/IHS/TIM layers as the main points of thermal resistance, and thus the main bottleneck for transferring the heat of the CPU transistors to the case airflow. Considering how much larger the Ninja is than the CPU die (especially the fin area), this should be intuitive.

Another way to come to this conclusion is to compare the thermal resistance of the CPU IC and case to the overall measured thermal resistance of the CPU and HSF calculated from the temperature delta and wattage. Intel doesn't publish their resistance spec, but I know that similar packages from other vendors range from 0.12 to 0.15 for chips with significantly lower wattage per cm2 (implying a higher number for the more power-dense CPUs). The SPCR review measured the system at about 0.20.

Felger Carbon · Post by **Felger Carbon** » Thu Oct 25, 2007 5:07 pm

cmthomson wrote:As you may recall from my 830D DIY article a couple of years ago, the Ninja with a simple duct can handle 150W of heat load with a slow fan.

Assuming the exhaust fan is pushing more CFM than the slow fan on the Ninja, then that slow fan should not see any backpressure at all. This means the new 9-blade Scythe SlipStream fan, with its extraordinarily low noise (when backpressure is absent) is ideally suited for this application.

I just now checked again: I can't hear the 500RPM version at 12V unless I hold the fan within 3 inches of my ear! This in a very quiet room with double-glazed windows.