Building a Better ATX Case

I recently decided that my computer was powerful enough for pretty much anything I actually do with it. As a PC nerd I still wanted to tweak and fiddle, so I decided to focus on quietness. I like to think that I have particularly sensitive hearing, so that was a nice big complex project to embark on. Since then I’ve read quite a bit of cooling and quietness theory, particularly on the excellent website There is plenty of debate on the right way to do things, and money = progress, but by doing nothing but switching out a few fans and a cooler(and a tiny bit of undervolting) I’ve managed to make my PC quieter than ambient at full load while simultaneously running ~20 degrees cooler!

This process got me thinking about airflow, which is the second most important element of maintaining a cool PC, the first being contact between the CPU and heatsink. The nearly unbiquitous design for motherboards and computer cases is called ATX and it celebrates its 20th anniversary this year. I’ve read a few articles in which people have complained about it being a dated design over the last few years, but none of them mentioned airflow. Back in 1995, graphics cards didn’t have fans. Often CPUS didn’t either. The Pentium 1, released in 1993, had a Thermal Displacement Power of 8W (some desktop processors have hit a TDP of over 100 in the last few years). Further, HDDs were so loud that they would drown out case fans to some extent, meaning that those fans could be run faster without being noticed. Perhaps most telling of all, Intel originally intended for airflow in an ATX case to go in the opposite direction it does now[1]!

Nowadays we have graphics cards with massive cooling requirements, and CPUs that demand meaningful air cooling. My Radeon 7970 has a TDP of 250W…before overclocking. While CPUs have gotten generally cooler recently, GPUs have been on an upward trend. Fortunately our heatsink and fan design have improved along with those requirements, and we can get that heat away from the chip quickly and quietly. The problem is that the air then needs to go somewhere. That’s where ATX falls flat corsair-550d-install-3on its face.

This is a picture of the case I have, a Corsair Obsidian 550D. Of course this isn’t mine (I wish that looked so neat), but it’s a very typical setup for newer cases. Air is sucked by two fans in the bottom right and flows through to the top left where another fan blows it out. On the way it needs to collect as much heat as possible and then get out immediately. The problem is that middle bit. In ATX, the CPU sits above the expansion slots which hold (in this example) a pair of very hot graphics cards which can only expel a fraction of the air they process out to the left. This means that any air sucked in is being heated up before it gets to the CPU. It is impossible to reduce the temperature of a chip below the temperature of the air flowing over it. This means that the bottom graphics card is making the card above it hotter, which in turn is making the CPU hotter still. This problem is compounded if, as I do, you have a cooler which blows all of its air upwards rather than flowing as much as possible directly outside. It would be better, though not perfect, if the airflow order were reversed because the CPU has a lower operating temperature than the GPUs, meaning that the air would be cooler moving through the case, even though it would ultimately exit the system at the same temperature. Unfortunately the exhaust fan needs to be higher than the intake fan (because hot air rises), so it is not possible to simply reverse the fans on a standard case. Remember that the original specification had the fans placed in the same places, but with reversed airflow. Then it would have been PSU->CPU->GPU->exhaust, which is a much better order.

I could go on, but I think you get the point. ATX cases have somewhat poor airflow. I’ve started pondering improved case designs that would allow better airflow, giving us more efficient cooling and therefore quieter systems. To that end, I’ve come up with a radical design for a case which is compatible with ATX motherboards. It would work even better if the motherboardwere changed as well, but by keeping it I am able to ensure that this could theoretically be made and sold. Though probably not cheaply.

C:\Users\Adrian\Desktop\Case Design Layout.pngMeet Negative Nelly!

The case layout to the right is a representation of NNs layout as viewed from the front. It is broken into 3 compartments, each of which is partially isolated from the others. The green line on the far right is a standard ATX motherboard. The CPU is housed in compartment A, the GPU in compartment B, and everything else in compartment C.

Note that there are two sets of lines defining the case. The inner lines are actually a tray that sits on roller bearings and can be pulled out to build and modify the PC. This is necessary because the external “shell” must be airtight in order to benefit from the cooling effect of negative pressure. See where the name comes from?

Positive pressure is when the intake fans are blowing more air into the case than is being exhausted, negative pressure is the inverse. Since very few cases are perfectly balanced, all PCs have either positive or negative pressure. There is much debate about which is better, and the argument for negative pressure is as follows:

Noise is caused by more fans running at higher RPMs.

When a fan pulls air, it creates a vacuum behind it which will fill with more air almost instantly.

if the case is exhausting more air than it is pulling in, the vacuum effect will pull air through more quickly without requiring more fans or higher RPMs.

Therefore: Negative pressure = quieter!

I was in this camp until I read an article on the Silverstone website[2] on the topic. As a component manufacturer they have the time and equipment to properly test airflow, so heed their words (unless they are trying to sell you something of course). Additionally, every case I have ever owned came set up for positive pressure, as do laptops, so there seems to be consensus among those who should know. The crux of the problem is that negative pressure does not direct airflow. Just like water and electricity, the air filling the vacuum is going to come from the place which has the easiest access (following the path of least resistance), and in a typical ATX case that place is very close to the exhausting fan. This means that the hot air being expelled is being sucked right back into the case, and not even passing over the components we want to cool!

C:\Users\Adrian\Desktop\Case Design AB.pngThis is the motherboard side of NN[3]. Here you can see the single MASSIVE 400mm case fan exhausting air from the case. Because it’s so big, this fan can pull huge amounts of air at low speeds, meaning very quiet operation. Because the shell housing the fan is airtight, all of the air being sucked in must come from the front of the case where the cool air is. There will be some exchange of air between compartments, but that is unavoidable and not concerning.

The next detail you should note is the line between the case and the CPU, sitting approximately where most mobos have a PCI-e 1x slot. In the actual setup I have here this isn’t actually necessary, but it prevents air from interchanging between compartments A and B. Because many GPU coolers blow their air upwards, this blockage exists to prevent as much GPU air as possible from approaching the CPU. Since it will interfere with some motherboards and GPUs, I would make this panel removable in a retail design (note that it does not need to be airtight).

The reason we don’t need it here is that the GPU cooler I used, to my knowledge, doesn’t actually exist. While impractical in a dual GPU setup (which very few people have or need), what I’ve done is planned a frankly ridiculous tower cooler of the type normally used on CPUs. Weight would be an issue and the contact between the GPU and cooler would have to be perfect in order to fully utilize it. However, it does represent an ideal cooling situation.

The last point of interest is cabling. In the layout picture you can see a thick central line in the removable tray. That houses permanent wiring that connects anything which needs to sit between the three compartments. IO panels inside the case link the sections as required, as well as to a front panel which houses this like headphone and USB ports. Once again, in a retail design this wouldn’t be flexible enough, because it locks the user into certain types of cabling which become obsolete or otherwise insufficient. It would also increase cost substantially. Replacing the panels with holes would allow custom cabling with a minimal impact on airflow, though extension cables may be required for some components. Anything at needs to exit the case (such as video cables) can be sent out a hole in the front as seen in the layout

C:\Users\Adrian\Desktop\Case Design C.pngIn compartment C we have all the remaining gubbins. Because they are connected by cables, we have much more flexibility with respect to their configuration. Starting from the top:

The IO panel is directly connected to the wiring passage in the tray, and has all of the USB ports, headphone cables, and whatever else you might want. A disc drive sits below it, with room to add more if desired.

HDDs are mounted longitudinally, right in the middle where they will get more than enough air. In my experience they need virtually no airflow at all, so this is like Christmas for them! Because they have lots of room to play with, they can be suspended elastically from the middle of the case. This has been demonstrated to reduce drive noise substantially. With a specially designed frame this will be no less convenient than in a typical ATX case. This suspension system is the only reason for the leftmost panel in the tray as seen in the layout, so if it were eliminated that would make installation of components on this side slightly easier.

I am of two minds regarding the power supply. I have put it on little feet so that it can be placed with the intake fan down without compromising air tightness of the shell, but in this drawing I have it upside down. Testing would be required to determine the best method, so I left both in to represent both options.

Some people have commented that they feel this case is too large, since the whole C compartment is unlikely to be filled. While this is also true of the standard ATX case (most people don’t have 5 optical drives and 6 HDDS), it is true that this design contains some wasted space. This could be minimized by reducing the fan to the same height as the motherboard. Since the fan defines the height and width of the case while the depth is actually substanitally lower than a standard case, I believe that the total volume could be made smaller than an ATX while permitting the same number of components and providing much better airflow. That is THE trifecta of improvements in PC design.

The only other point that I should address is the notion that the exhaust fan could be exchanged for an intake fan on the other side, creating a positive pressure case which would move a similar amount of air by pushing it out quickly. There are a few reasons this wouldn’t be as good, but the main one is that airflow from an intake fan is shaped. According to Silverstone[4] (who are admittedly trying to sell us something in this case) air moves outward from the fan, meaning that most of the airflow would be on the edges of the case where it isn’t needed. Further, there is a hole at the hub of the fan which has no airflow at all, again reducing air flow where it is needed most. A negative pressure situation should provide a much smoother distribution of air.

Of course it would still work, and I believe better than a standard ATX case. I just don’t see any reason to do it. So, the PC component industry has given open air cases and computers built into desks. Could Negative Nelly be the next big thing?


  3. I have left out representation of the removable tray for clarity.
  4. … =wh10_0061
Posted in Uncategorized | Leave a comment

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>