Started an interesting discussion with Bluefront here about how big a role natural convection plays in cooling PCs and how much of a difference to cooling going against it makes.

I've always been of the opinion that it's nice to go with natural convection if possible, but, as I believe it plays a minor role in PC cooling, I don't think going against it should make much of a difference. That said, I haven't done any controlled experiments to look into it, so I'm open to ideas. The simplest experiment I can think of, if you have a standard ATX bottom to top airflow, is to simply measure temperatures with the case the right way up and then upside down. I'll give it a go this weekend. Anybody else is welcome to try it and to add these and other experiences.

Opinions?

I have no hard numbers to cite nor recall any detailed experiments, but it's been clear to me that convection is only really useful in a passively cooled system.... where it's critical to use every trick to move the heat.

With even a tiny bit of forced airflow, you can go against natural convection with impunity. With the amount of heat in most systems, the rising force of convection doesn't stand a chance against forced airflow from even a <800rpm fan. Obviously I'm not talking about any "serious" gaming rigs.

In general, I'm only sold on passively cooled systems where...

1) the thermal envelope is very small -- ie, there's little heat to begin with
2) very dusty conditions prevail, where sealing the system may help defend against early breakdown of components (including fans)

The cooling advantages of a couple of quiet slow spinning fans are immense, compared to their negatives.

This is a complicated subject, from many view-points. But start off with an SPCR view-point......we want to make our setups operate as quietly as possible, but cool enough to avoid over-heating problems. This mandates a quiet airflow technique. The use of just enough forced airflow, and no more, is the number one priority to creating the quietest possible computer. (water cooling is a different question). Of course choosing quiet components makes the project easier. But it's not impossible to use hotter components in a quiet computer.....just harder. This brings us to convection......heat always tries to rise, even if it is being forced (blown) by a fan. And to build the quietest possible computer, we need to make use of every cooling technique.....including convection.

IMHO.....the design of the case and the integration of ATX form factor components is the start of the problem. We are saddled with components of a particular size and shape, but we are free to re-arrange those components in any number of ways. The case itself is the first hurdle. Look at most std tower cases....power supply on top, along with large optical drives, both of which block natural convection's way of removing heated air....out the top. Having a solid flat top on a case does have advantages....you can't spill stuff into your computer, and you can set things up there, and you can shove the computer under a desk with no clearance on the top. But from a "remove the heat quietly" view-point, a standard case setup, is far from ideal.

There are examples of computer cases that attempt to solve some of these problems. A P-180 puts the hot PSU on the bottom, which brings new problems, such as cable issues, and the convection thing itself.....putting hotter components in a lower position, is not ideal. The standard ATX PSU location blocks convection airflow from the rest of the computer. My own solution to this problem.......move the PSU partially out of the case. Buy a fanless efficient PSU that requires little airflow. This opens up the top of a tower case setup, enabling better convection airflow. Like this.....



There are countless ways to make use of natural convection.....all of which make a quieter computer a real possibility. This is a subject that needs to be explored further. There is no one answer here. But natural convection exists, and is quiet. We all need to make better use of it.

i think the easiest method would be to use a thermal/infrared imager to timelapse the flow of heat in the case?

As I recall from physical chemistry of macromolecules- the amount of energy conveyed by convection is function of height, (cross-sectional) area, viscosity, and difference in free energy (temperature difference between the top and bottom). Oversimplifying a bit more, the narrower the tube in relation to volume, the better it will work. Things like turbulence (obstructions in the case, like cables or fans) would increase the viscosity.

In order to promote a sharp gradient, you could:
* Put all the heat generating items on the bottom and put all the cool stuff on the top
* Place the hot and cool sections as far away from each other as possible
* Minimize anything that would impede laminar flow (increase viscosity)


My favorite examples of where it worked well are:
The Mac Chimney - the voltage regulator for the CRT would get insanely hot and fail spectacularly.

Jonathan Ive's original G3 iMac. The heat rising through the vents from the CRT pulled air into the bottom of the case. The EM shielding did a significant amount of conduction while the convective airflow traversed along the surface of the EM shield to cool it.

One is somewhat more effective, the other is more attractive.

I have a Lian-Li PC-A05 build in my home office that I'm going to work on this summer that I hope will rely solely on convection. If I have time to take pictures, I'll post numbers when I get them.

.....deleted...

I think many people confuse convection cooling with passive cooling.....they are not the same. Passive cooling refers to cooling components without the use of a fan (in the case of air cooling), instead relying on convection to remove the heat.....but not always. The effects of convection are present even when a fan is used.

Here's a experiment I'd like to see performed.....using a Ninja heatsink on some CPU, with the MB mounted vertical. Remove all the obstructions above and below the Ninja. Mount a fan on top of the Ninja blowing downward (fixed rpm), then reverse the fan, blowing upward. Compare the results. Mount the Ninja below the fins blowing upward, then reverse the fan.

IMHO.....the results with the fan blowing upward will be better with the fan either below or above the Ninja. How much better....depends. But with the fan blowing upward, you are using convection, not fighting it. This is not a perfect experiment because of the effects of the location of the other board components. But it should give some positive results if set up properly.

Also.... convection cooling can be used on other computer components, not just the CPU. In the first photo above, you see a fanless Zen mounted partially outside the case. The Zen is cooled by positive pressure induced airflow over the main Zen heatsink. But it is also cooled by convection airflow, from outside the case.

Here's an example of convection cooling using a moderately hot video card(in the same case). I modded the heatsink so it extends past the end of the card. This allows good convection airflow through the heatsink, upward and out the top of the case. I help the convection airflow by placing a fan below the heatsink blowing upward.



In a different computer here's a card modded with a Zalman GPU cooler. I added an airflow deflector to the setup, which forces the airflow from the Zalman upward, out the top of the case. All I'm doing here is helping convection airflow with a redirection of the fan airflow.

I too would like to see an experiment to quantify the role of convection in a forced cooling setup. In the setup that you mentioned Bluefront, it would be good to have the Ninja ducted both upwards and downwards (and out) so that with the fan blowing downwards you don't just end up with the hot air recirculating in the bottom of the case (if PCI slots are closed). if i ever get my new PC built I will try this (this would be with a Noctua NH-U12 heatsink).

Yeah you'd have to duct the Ninja [close off the sides]. If you didn't, the suck mode would suffer. Normally blowing through a heatsink gives better results than sucking. The best setup would be with a fan on the bottom, blowing upward. Like I said, this is not a perfect experiment for judging the effects of convection.

You know that you can set up an easy convection experiment. Heat up a flat piece of metal. Hold your hand over the thing, then under the thing. Above the heated metal feels hotter.....it is hotter. Because of convection air currents. Now just how much this effect has inside a computer is the question. Is the effect enough to make using it worth the effort? My last five case projects have no exhaust fans....... The first picture in this post shows an exhaust fan under the Zen. It's temp controlled, and only turns on if the CPU hits about 52C......which doesn't happen.

Here's another example of using convection in a positive pressure case.....to cool another fanless Zen. There is no exhaust fan......only the rear case vent hole with a deflector to consentrate the passive exhaust toward the bottom of the Zen.

I have thought about this and have come to the CONCLUSION (the way I see it now) that natural convection does not work unless there is a long smoke stack atop of the case. the lower the heat source is, the more channeled the air is and, definitely, the longer the tube is, the more a chimney effect can occur. atx/btx all of them are totally designed wrong. isolating the psu is also wrong! Why?

here's my thoughts:

energy cools the computer. in the form of fans. right? well, energy also gets turned into waste heat which is how natural convection work. I once designed, on paper a 4-5 foot tall case that had the board rotated at a 90 degree angle allowing for the most important piece of eternally hot hardware, the vid card, to be facing up and down. the psu would be also be 90 degrees turned and blowing upwards, or, not blowing at all if it was passive, but its vents pointing up. It would be above the mobo but not in line with it, off to the side so nothign heats up anything directly just adds to the force of air. everything would be pointing up and all heatsinks would be thick metal, long, and with lots of spacing, with fins pointing horizontally to catch the draft of the case.

Anyways, a modern box case is a terrible design. You need to have a:

/ \
/ \____
| |


shape at top vs. a blow hole on a flat top like most cases that try and take advantage of natural air rising do. I would advocate for no horizontal top the case at all and just a nice long slope towards the chimney. Turbulence destroys a chimney effect. One day I am going to make it 6 feet tall and do my layout, but not until I find a more fitting apartment to do that in.

crap! i didnt even see the mac chimney. I discard all references to macs so much that I didnt even look at that photo. Yes, something like this but not just a hat, a whole case with its internals and mobo layed out so it works with everything pointing up.

My goal is to silence machines that can play 2007 games with max eye candy turned up running at above 40 frames per second. I wish all i had to do was to do some school work/editing/light stuff then I would be making all passive comps for fun. There should be a forum for people seriously silencing gaming beasts with other forums building ground up silent computers for people who do not game. there really is a big difference and much incorrect communication and advice from one camp to the other. I see people still saying how a 300 watt psu is more than enough for a gaming rig, failing to take into account how gaming rigs can actually get quite near 300 watts and pci-e connectors are inadequate on 300watt psu's. Then there are people who say air is better than water cooling for a gaming rig even, which drives me up a wall, etc etc.

Obviously there are any number of ways to make use of convection, but I'm mostly interested in techniques that can be applied to relatively normal sized computers. Here's a way to help along convection and still partially cover the top of your computer......slant the top. The opening at the rear can be vertical, or partially slanted. Both ways work well, better than a raised flat top with a rear facing opening [like my first photo in this thread].

What about the effect of gravity, making it easier to blow air downwards than upwards? I'm guessing it's negligible.

I'm not the total expert here.....just a duffer who likes to build things. But I think you could say that gravity makes convection work. If you throw a stone in a pool of water...it sinks, because it's heavier than water. If you throw a block of wood in the water, it floats, because it's lighter than water.

Works the same with air. Heated air (hotter than surrounding air) rises because it's lighter than the surrounding air, by volume. A helium balloon rises for the same reason. Somebody correct me if I'm wrong.

I suppose that's why it's called "natural convection". Anyway....it's working, or at least trying to work in all computers. If you work with it, you can mostly eliminate exhaust fans.....making for a quieter computer.

that's right.

a square case cannot take much advantage of natural convection besides to have NO top at all. Just like make it a floating hat so dust doesnt fall in. put 1-2 120 fans on front for positive pressure and voom, the air case will run really cold. the tee-pee idea is a good one though. I just do not think cases are tall enough to really get the flow going right.

Natural airflow can work really well. Just see cooling and solar towers.
But when you look inside them you'll see they are very smooth.
Cooling towers even have a special shape to increase the chimney effect.
PC's are just a heap of junk inside.
Airflow will start twirling around and just gets trapped.
Imo the will greatly reduce the positive effects of natural airflow.

It may sound weird but perhaps heat is the way to cool a system
If there is a heater inside the chimney a natural draft will be created...
Time for bed I think

Someone on this forum created a chimney with a 40W lightbulb in to get a draft but it didn't work very good. Intresting idea tho

IMHO....if you don't mind a tall/ugly chimney sticking out the top of your computer, there's no reason it would not work .....with a little help. Start off with positive pressure in the case, add your tall chimney, install a small heater of some sort at the top. Heated air rises, and it will pull hot air out of the case, as it pulls cooler air into the case. The fans blowing into the case will help the effect.

The problem you face here......implementation of the design. The design of such a chimney needs to be perfected. I really don't think you need a chimney to get natural convection to work in a mostly standard looking computer. I'm doing it right now with several different setups. Could you get an improvement with a chimney.......maybe. Would it be worth the effort? Not in my book......

I'm the one who tried the light bulb chimney experiment, and the effect was simply too pathetically weak to be worthwhile. A single silent undervolted 80mm fan provides more airflow, with less power consumption.

IMHO, at the small scales of a computer case, the only place for the "hot air rises" theory is the prevention of air recirculation when intakes/exhausts are on the same side of a case. The effect within the case is too weak to worry about.

Hot air rises, and pulls cooler air upward as it does so. The effect can be easily demonstrated. Getting this effect to work to work in a computer does take some effort/planning/modification.......if you want the outcome to be worth the effort. If you randomly cut a few holes on the top of the case, blow all the case fans inward, hot air will come out the holes on the top of the case. How well this setup works depends on a lot of things, but mostly it depends on a good airflow channel up the center of the case. Here's what I mean..



That's a photo of a recent project with the top flipped open. The case is positive pressure.....most of the exhaust comes up through the visible channel between the PSU and the optical drive. The XP-120 has the fins orientated vertical. Everything in the case promotes upward airflow. If I put a chimney on top, rather than a flat top with an opening at the rear, this design would cool better. As it is it cools well enough to be very effective.

That's a nice theory, but how much of a difference in temperatures do you get if you rotate the machine 90 degrees onto its side? (If you go a full 180 degrees to upside-down, then you may get recirculation of warm air from the exhaust wafting upward to the intake.)

Apparently, the Roman knew more about convection than we do...



My Art History professor back in college said that even with the hole in the top it never rains inside because of the force of the hot air coming out the top. Of course, you need quite a few intake holes in the bottom (entrances/exit), but that's what dremels are for...

Well I cannot flip the case over to check what happens......the first photo I posted in this thread, shows the case from the rear(more recent photo). The exhaust in that mode, would get sucked back into the intake.

As you see it with the top wide open, you'd think it would cool better. Not so...the PSU temp goes up (less airflow over it) and the rest of the temps remain virtually the same. This indicates to me the design is tweaked to the max.....

That church picture is a good illustration of convection at work. People in the church generate the heat, the heated air goes straight upward, going out the top hole, and pulling air inward through openings at the bottom level. Now if we could make a computer with a similar-looking interior.....

A church?
Hmmm, I assumed it was a datacenter
That hot air rises is a fact. Just take temp reading near the floor and near teh ceiling of your room.
The main question is what are the best conditions?
If a room/space is completely filled with hot air, like an oven, the draft has gone.
Draft can also go down when it starts to twirl.

What, imo, would be the best solution is a hybrid solution.
Something like bluefront has build and a real big slow fan.
Then I'm not talking about 12cm fans but more as big as the top of your case. Say 30cm. Then let it spin real slow. 300rpm and I think you will have a real silent and cool setup.
The problem is that I've never found fans larger than 12cm.
Papst has some but they run at 24v.
68dBA, 1014CFM, 1410RPM @24v. Range 16-28v
Besides of that I have no clie where to buy them.
http://www.ebmpapst.us/index.asp
http://www.ebmpapst.us/allpdfs/1G360.PDF

I have no doubt that your prof knew what he was talking about, but to equate the thermodynamic characteristics of a building at least 1000 times larger to that of a typical PC is folly. Whether it's fireplaces and chimneys or fires in wigwams, the scale of effective convection-only cooling structures is always many times larger (and greater in total heat as well as in temperature differentials) than a PC.

If in doubt, go ahead and drill many holes near the base of a PC case, block all other holes, and cut a single blowhole in the center of the top of the PC. Make sure there are no fans. Now, run CPUburn & 3DMark concurrently on the system, and place the system in the rain. It will be a tossup whether the system crashes due to overheating first or burns out first due to the moisture of the rain.

Well I do have a big fan blowing in from the side.....and it turns slow, always under 800rpm. The only other fan in the case is under the GPU heatsink, blowing upward. So far the design works perfect. And the case looks fairly normal, with a flat top.



MikeC's design would work if you add a few fans blowing inward.....basically that's what I've been doing with my last several projects. Of course a bigger structure pulls more air through it.....but convection is working in a smaller device like a computer. My satellite receiver runs hot (no fans), but the top is completely covered with vent holes. I just can't sit anything above it.

@MikeC
That's a clever thought. Combining passive cooling with water cooling

@Bluefront
Did you really forget the specs or are you teasing me
Can you please tell me a bit more about that fan?
website, dBA, size etc

no disrespect to your prof, but wiki page suggests otherwise:

http://en.wikipedia.org/wiki/Pantheon%2C_Rome





Felger Carbon has done loads of these. check his posts.



true, but not really the subject of this thread. this is more about whether, all other things being equal, there is a tangible benefit from using forced cooling in a way that leverages natural convection (ie if you have the choice of blowing up or down (vertically) through a heatsink, which is more advantageous (if there is any difference at all)?

I've been to the Pantheon when it was raining. Rain enters the top. Also, it's cooler on the inside than the outside--a natural effect of shade.



That's why I suggested testing by rotating the case 90 degrees onto its side, rather than upside-down.

Without a test like this, then you really have no idea how much of a benefit (if any) the chimney effect is providing.