“I got 8 fans on my PC case!” “That’s nice.” “What?” “I said, That’s nice.” “Black Ice?” “Never mind.”
I’m not going to launch into some Theoretical Physics rant here, but rather some basic tenets of General Physical Science. That’s a class they used to teach at my old High School. Based upon what I see in some of the Cooling forum posts, they no longer offer that class at some schools.
With the advent of warmer weather, I have been challenged to remove my case circulation from contributing to my CPU temperature. My goal is to have my case (aka system) temperature the same as room temperature. It’s not that hard to achieve. You simply have to have enough through-case flow to prevent residual heat.
I use an Enlight 7237 mid-tower case. I do not have enough drives to warrant a full-tower, though in retrospect, my next case will probably be one, for the sake of adding a closed system water cooling solution and perhaps dual PSUs. My 7237 comes with two 80mm add-on fan openings. One is in the lower front and the other, in the upper rear. I originally populated these openings with modest 80x25mm 36cfm fans. The did a fine job when my PC room was 16C, but now, with it being more like 22C, I had to regain that 6C somehow.
Pause a moment and ponder this fact of life. Axial fans only deliver their rated capacity (cfm) in free air space. As soon as they encounter any resistance to airflow (aka static pressure) their performance collapses pretty rapidly. You can meter this by placing your hand behind one and feel the air slipping back out the intake. The thicker the fan, IE 80x32mm versus 80x25mm, the better they handle static pressure, up to a point. Add to that, the fact that it’s bad air to start with. Axial fans propel air in an ever increasing cone with a spiral component to it.
Where does your CPU heatsink need air the most?
If you look at a thermogram of a heatsink while PC is running, most of the heat is concentrated in the center. With the exception of the Delta FFB812SHE, which employs static vanes to de-spiral the airflow, axial fans put out an air pattern with a hole in the middle. So, I was not living under any false pretense that those 36cfm fans were actually delivering their rating.
Nevertheless, I had to get the through-case airflow, regardless of the fans I wound up using, greater than it was. Now, in my 7237, there really was no other place to add additional fans, without cutting holes in the sides, top, or bottom. I really did not want to do this unless I absolutely had to. I like the front to rear, S pattern circulation model and wanted to maintain it.
There were a few, higher rated 80mm fans available, but not that much higher rated unless I was willing to get some of the 69cfm Delta screamers, like the one on my Swiftech MC-462A. That’s another issue I will address shortly.
The 92mm form factor.
While perusing the various web sites that sell fans, I came upon the Sanyo-Denki 92x32mm 55cfm fan. Being thicker than most 92mm fans, I decided to give it a shot. I have to say that as fans go, it is one nice package. It was more tolerant of modest static pressure than anything I had played with to date and without being obnoxiously loud.
Out came the nibbler, hand drill and moto-tool. Every OC’er should have a nibbler and moto-tool. The two fans, one sucking (rear), one blowing (front), lowered my case and CPU temp a few degrees, but surprisingly, still not to my satisfaction. Again, there were more powerful 92mm fans available, not enough to warrant buying more fans to play with. Already, the twin 92s and the 80mm Delta screamer were louder than I ever wanted my computer room to be. Also, with the input and output fans being equal capacity, that left the case at neutral pressure.
When you factor in the exhaust fan from the PSU, I was in the negative pressure zone. That’s not necessarily a good place to be, since you wind up sucking in dust through all the cracks, crevices and removable disk drives. Note that I am fastidious about not allowing any air to get into the case from passive openings. I tape up any holes, slits, cracks and crevices that have any volume to them. This is to encourage the air to flow from one single in-point to one single out-point.
That may be nit-picking, but that is my personal preference. If you get too much air flowing from different directions, it collides and causes strange cooling anomalies. So, what to do? I wanted more cooling air, but I sure did not want more, or louder fans.
The 120mm form factor.
Once I get to nibbling on sheet metal, I have a hard time quitting while I am ahead. Same problem with hedge trimming and tree pruning. So, I did some measuring and decided to change the 92mm front intake fan to a 120x38mm Panaflo that delivers (in free space) 86cfm. In my 7237, that was really pushing the available space in the front. So, now, I had the 120mm Panaflo 86cfm blowing air in and the 92x32mm Sanyo-Denki 55cfm, in parallel with the PSU fan sucking air out.
Though the 120mm Panaflo did a good job, at a very reasonable noise level and yielding a positive pressure to the case. I could still see a few degrees improvement in the CPU temperature by removing the case cover.
One important factor you must consider is the fact that the 80mm Delta 69cfm fan is very powerful and generates quite a strong air wash out the bottom of the MC-462A heatsink, so much that it fights the through-case “S” airflow pattern. Even with the high rate of flow going through my case, some of the warm exhaust air from the Delta was getting back into its intake when the cover was on.
What if the intake on the HSF was getting fresh room air instead of case air combined with some of its own backwash? I reversed the direction of the 92mm and 120mm fans so that air was coming in the back opening, adjacent to the HSF and exhausting through the lower front opening. This reduced my CPU temperature about 2C, but now I was back to negative case pressure.
I reduced the 120mm fan voltage and consequently its airflow capacity until I got a slightly positive case pressure. I was wishing I had tried this with the dual 92mm fans first, as I now had a gaping hole in the front of the case. I was within 2C of what the CPU temperature was with the cover off, but it happened to fall at one of my OC’ing failure thresholds.
All CPUs have certain temperatures where once you cross them, you have to drop back a speed increment. With my setup, I could run my 1G/266 Tbird stable at 10×150 until the CPU temperature hits 30C. Then I can run 10×145 until it hits 37C. Then 10×140 until it hits 44C. All these at a core voltage of 1.85V. Yes, I did try less voltage, but no soap.
Believe me, I have tried just about every imaginable trick in the book by now. Also, it is important to note that all this endeavor has been to satisfy my own definition of running stable. By a stable OC, I mean one that runs Prime95 Torture, error free, until the CPU temp stops rising and also runs 3D Mark 2001, error free, immediately afterwards, while everything is still warmed up.
I have a rheostat on the Delta fan because it does not have to run at full speed except when I’m running an intense 3D game, Engineering Modeling Programs or a CPU stress test. Essentially, when the CPU is running at full load. So, Now, I’m out something like $60.00 for different fans, still not hitting the temperature I wanted and my PC is too loud to be around for protracted periods of time. God, I love this hobby.
Then, it hit me, though not for the first time: I have worked with Centrifugal blowers, for many years, in high power (>1.0kW) Radio Transmitters and Amplifiers. I like these kinds of blowers. Not just because they put out large amounts of airflow, but it’s Good Air. By that, I mean that the air comes out in, for lack of a better term, a solid tube. I like to think of this as cohesive air. Dense and uniform across the entire outlet opening.
The idea of using a centrifugal blower re-dawned upon me while perusing through a WW Grainger catalog, looking for a replacement blower for an amplifier. There, on one of the pages was a powerful (170cfm) blower, but unlike every one I have ever seen, it ran off of 12VDC instead of 120/240VAC. Add to that, the fact that is only cost $42.00 ($37.00 with company discount) and the hook was set.
A quick phone call to the local Grainger and I had one ready to pick up on the way home from work. “A fool and his money…” you say. Okay, I got this baby home, put a 4-pin connector on it and plugged it in.
It sounded like a gale force wind blowing through the tree tops.
Admittedly, I could not hear the motor, but then I could not hear much of anything. Papers on my desk across the room were launched in a mad dance; I struggled to point it in a non-destructive direction while trying to get it unplugged. Okay, some taming was definitely in order here, but I was awe struck by the raw power of it.
One of the beauties of DC motors is the fact that you can vary their speed by varying the voltage feeding them.
Still, I was challenged as to how to implement this brute into my system. Clearly, it had more than enough power for my needs, but I also had to couple it to my PC in a sensible fashion. This was getting interesting. Now, it plainly had enough good air flow to push in one end of the case and out the other, now what could it do to get rid of that noisy Delta fan?
Not wishing to make a permanent “blowhole” in my 7237 cover, I went into my workshop and with a few quick table saw maneuvers, I emerged with a piece of particle board that was exactly the dimensions of my cover. Some quick calculations, a bout with the saber saw and I had a cover with a three inch square opening directly over the MC-462A heatsink.
Before testing this odd looking arrangement, I removed both the 92mm and 120mm fans. I further closed up the 92mm fan hole with some duct tape, but let the 120mm fan hole open. Then I mounted the blower and it looked like this:
I plugged the blower back in (12V) and fired up the PC. The sound was not as overwhelming as earlier, but the sound of the rush of air coming out of the 120mm fan hole was like sticking you head out a car window going 65 mph. Okay for a short test. Not knowing entirely what to expect, my eyes were glued to the Fluke digital thermometer attached to the thermocouple in the heatsink.
The room temperature was 22C. At the Windows ME desktop, case temperature was also 22C (no big surprise) and the core temperature was 22C. I fired up Prime95 Torture and watched the core temperature rise to 32C. Now, at 1450 Mhz, and 1.85V on the core, my Tbird is generating 84W of heat. Definitely not for the cooling challenged, but 32C with air cooling is pretty darn sweet. A full 5C less than any other setup I have tried. As best as I can figure, the circulation pattern must look something like this:
Okay, got the temperature I wanted, now what about the noise? I left Prime95 Torture running and started reducing the voltage feeding the fan. The noise level of the rushing air went down proportionally until, at 7V, there was only a moderate “whooshing” sound and a low level squeaky sound I believe to be the brushes on the rotor in the blower motor. The core temperature rose to only 33C. At 5V, the “whoosh” was the same As I heard with the 120mm Panaflo axial fan, but the brush noise did not change.
The core temperature rose slightly more, stopping at 33.7C and there it sat.
Pinch me, I must be dreaming!
One fan in place of three, with excellent cooling capabilities, and noise equivalent to a 60mm Sanyo-Denki 26cfm fan!
I sincerely believe the effectiveness of this setup can be realized with just about any decent heatsink, not just the expensive Swiftech MC-462A. For you “nay-sayers”, here is one last picture of the exhaust plume, with the blower running at 5V. I had to take this quickly, before the candle blew out.
So, don’t drill all those holes in your case and populate them with small, noisy, ineffective fans.
Pay once for all and get a wicked looking system, sure to raise an eyebrow or two at the next LAN Party. I want to apologize for making this such a long exposition, but I felt justice demanded the full development description. I hope you enjoyed reading it as much as I enjoyed developing it.
Judging from the hordes of email I have gotten from the original article, not everyone that reads the articles on the front page participates in the forums – namely, the Cooling Forum. I tried to update my article with some tips and additional information in a post there. Therefore, for the benefit of those of you who only read the articles, here they are also.
I apologize for not including the make and model of the blower. It is a Dayton 2C646A, sold by Grainger.
I was able to enhance CPU cooling effectiveness by attaching a short duct which extends downward from the blower opening and stops just short of the top of the HSF. At lower air speeds, it assures that the airflow does not start to disperse before hitting the HSF. I picked up a short length of square aluminum tubing, 75x75mm OD. You can see the method of attachment in the accompanying pictures.
When I originally tested the setup, I used fixed values of power resistors to drop the +12V output from my PSU to test different speeds. If you think dropping the voltage for a conventional fan wastes energy, you haven’t seen anything yet.
This blower consumes 5A of current at 12V, and by dropping it to 5V (very quiet), you must dissipate a lot of wattage in the dropping resistor. I wanted to have a means of continuously varying the voltage and chose to use a three terminal variable voltage regulator.
The LM-338K is the big brother to the popular LM-317. It can handle a minimum of 5A continuous current and, more often, something in the area of 8A when properly mounted. The circuitry is the same as the LM-317, with 1 fixed resistor, 1 potentiometer and 2 small capacitors.
This is the generic schematic for implementing a 3-terminal regulator, whether it is the LM-317, LM-350, or LM-338. The potentiometer (R2) value should be in the region of 1-2K ohms, not 5K ohms, for 12V operation.
There is enough space between the front of the case and the plastic escutcheon to mount the regulator, with a small heatsink and a long-shaft potentiometer. Notice the gaping hole where my 120mm fan used to reside. To borrow a few terms from Jethro Bodine, “The guzzintas can not exceed the guzzoutas.”
That is to say, you have to have enough volume of air exit capacity, to not limit how much air the blower injects into the case and at a reasonable amount of resistance to flow – AKA Static Pressure. Though not nearly as limited by static pressure as an axial fan, blowers are not miracle devices. Even with the pictured hole, myriad of other naturally occurring holes and the PSU exhaust fan, my case is still positively pressurized.
On feeding power to the fan, I chose to port the variable +12V out through a pair of banana jacks that I mounted in a couple of knockouts originally designed for a sound card. I do not know if all mid-towers come with these, but they sure were handy for me. You peltier enthusiasts might want to consider them also for bringing in your peltier power.
Okay, so the colors don’t match exactly. That was all I could find in my junk box.
The final assembly is not unpleasing to the eye. You Enlight 7237 owners may notice the difference between the fan holes in your front escutcheon and mine. I enlarged them for less resistance to airflow. I chose to orient the blower the way it is because the intake is facing downward and less likely to suck in objects that occasionally get launched around my room.
Notice the speed control knob on the front escutcheon. That is why you use a long-shaft potentiometer.
Notice the normal exhaust fan opening is sealed off. The idea here is to get the majority of the air that washes out of the HSF to flow across the motherboard and out the front opening. This setup keeps the Memory, GPU, Northbridge, etc cool – and by cool, I mean at room temperature.
Again, if there are any gaps in the documentation that I overlooked or any additional tips you need, feel free to post them in the Cooling Forum here at Overclockers.com.