Vertical vs. Horizontal Case Cooling

Introduction

Exceptional airflow is something that is essential for a cool-running computer. Traditionally, computer cases are oriented with the air intake in the front of the case and the exhaust in the rear. However, some manufactures have lately been orienting their cases 90 degrees from the norm with the air intake at the bottom of the case and the exhaust at the top. The theory behind this design is that since hot air rises, having the airflow in a up/down orientation allows the case to work with, rather than against, the forces of convection.

Procedure

Our main question is: does a vertical airflow in a chassis improve cooling? To answer this question, we will separately test two chassis in both vertical and horizontal orientations. Although we will be testing multiple chassis, it needs to be clear that we will not be directly comparing the chassis against each other. What we want to find out is if having an up/down airflow alone really makes a difference, not whether one case or the other has better overall cooling. Vertical cooling cases have a vastly different internal layout than traditional cases, so we will be using both a traditional cooling case as well as a vertical cooling case for all of our testing. The two cases we will be using are the Antec P183 V3 (traditional cooling) and the Silverstone FT02B-W (vertical cooling).

The Antec P183 V3 has a more restricted airflow and caters more towards quiet operation.	The Silverstone FT02B-W has very low restriction on airflow, which is great where noise is not as much of a concern.

All of the components we will be using for testing are high-wattage parts with very high thermal outputs. This means that some of our testing may give us temperatures that are much higher than what we would normally deem safe, but temperature variances should become more pronounced by using these very hot components. It needs to be noted that realistically we would never have these components in the Antec P183 V3 with the fans on low. We would normally either turn the fans up or mod a side fan into the side panel so that the case would provide adequately for all the components. For this testing, we actually want the case to be running very hot with low airflow to ensure that any forces of convection are not overrun by high-flow fans. We are trying to represent both ends of the cooling spectrum and these two cases are almost the embodiment of high airflow, and quiet cases.

Setup

Both cases will be setup with the stock case fans running at their lowest settings. All testing will be done with both a traditional air-cooling CPU heatsink (Gelid Tranquillo Rev2) and a closed-loop CPU liquid cooler (Coolit ECO II). Normally we would configure the Coolit ECO II as an intake in the Antec P183 V3, but in an effort to "work with convection" when case is turned on its front, we will be leaving it in the default exhaust configuration. This will raise the CPU temperatures on this cooler, but is more appropriate for this comparison.

Temperatures will be examined at both idle and load, with a combination of Prime95 and Furmark used to put both the CPU and GPUs under full load for 10 minutes. Temperature recording will be performed with Coretemp and GPU-Z, as well as with a thermal imaging camera. A controlled environment with an ambient temperature of 24 Celsius will be used for all testing. The margin of error for our temperature readings is 1 degree Celsius. There is a margin of error for the thermal imaging as well, but as we are looking at images and not hard numbers it is very difficult to define. We will be relying on our years of experience working with thermal images to determine which variances are worth pointing out and which can be attributed to normal testing fluctuations.

Testing Hardware:

Results

Before we examine the results, we should point out that all of the thermal images have been rotated so that the motherboard is in the traditional upright position to assist in visual comparisons. With that said, lets examine the results:

Additional note: some people have been inquiring about the blue dot in the middle of all the thermal images. This is the CMOS battery which apparently never gets hot even when the system is under very high load.

Vertical Orientation				Horizontal Orientation
CPU Idle: 28c				CPU Idle: 29c
CPU Load: 62c				CPU Load: 63c
GPU Idle: 46c				GPU Idle: 46c
GPU Load: 96c				GPU Load: 96c
Vertical Idle Thermal (rotated for visual comparisons)				Horizontal Idle Thermal
Vertical Load Thermal  (rotated for visual comparisons)				Horizontal Load Thermal

For our first test, the results were somewhat underwhelming. Simply put, there is minimal difference between the system in a vertical or horizontal airflow orientation. The one degree difference in CPU temperatures is within our margin of error, and there are only a few minor differences in the thermal images. The thermal image for the vertical cooling shows only slightly better cooling to the MOSFET heatsinks and a single IC chip above the CPU under load. The thermal images are very close, but the one red IC chip on the horizontal load image is just enough for us to call the vertical configuration the winner.

Winner: Vertical cooling, but just barely

Vertical Orientation				Horizontal Orientation
CPU Idle: 28c				CPU Idle: 28c
CPU Load: 58c				CPU Load: 59c
GPU Idle: 46c				GPU Idle: 45c
GPU Load: 96c				GPU Load: 96c
Vertical Idle Thermal (rotated for visual comparisons)				Horizontal Idle Thermal
Vertical Load Thermal (rotated for visual comparisons)				Horizontal Idle Thermal

With the Tranquillo CPU cooler, the results between the two orientations are identical. All temperature readings are within our margin of error, and the only differences between the thermal images are so slight that we are attributing them to camera angle or normal testing variances. So for this round, neither orientation has an advantage over the other.

Winner: Tie

Vertical Orientation				Horizontal Orientation
CPU Idle: 43c				CPU Idle: 43c
CPU Load: 101c				CPU Load: 94c
GPU Idle: 52c				GPU Idle: 52c
GPU Load: 109c				GPU Load: 109c
Vertical Idle Thermal (rotated for visual comparisons)				Vertical Idle Thermal
Vertical Load Thermal (rotated for visual comparisons)				Vertical Load Thermal

The airflow in the Antec P183 V3 is much more restricted than the Silverstone FT02B-W so we expected the the differences between vertical and horizontal cooling to be more pronounced in the Antec P183 V3 testing. We ended up being correct, but not in the direction we anticipated. At idle, the temperature readings are the same between both orientations. The thermal images do however show slightly better cooling with the case in the horizontal orientation, with the chipset and MOSFET heatsinks as well as CPU radiator all running slightly cooler. While it is still somewhat minimal, it is definitely a larger variance than we saw in the Silverstone FT02B-W testing. At load, we finally got some clear results with the CPU running a full 7c cooler in the horizontal orientation. Given the vastly different CPU temperatures, we expected the thermal image variance to be larger than the idle images but they were surprisingly similar. The CPU radiator is indeed much hotter, but the chipset and MOSFET heatsinks are basically identical in the two images. The area to the left of the chipset is however much hotter in the vertical orientation than in the horizontal orientation.

Winner: Horizontal cooling

Vertical Orientation				Horizontal Orientation
CPU Idle: 31c				CPU Idle: 31c
CPU Load: 74c				CPU Load: 75c
GPU Idle: 53c				GPU Idle: 52c
GPU Load: 109c				GPU Load: 109c
Vertical Idle Thermal (rotated for visual comparisons)				Vertical Idle Thermal
Vertical Load Thermal (rotated for visual comparisons)				Vertical Load Thermal

While we were able to see temperature differences in the Antec P183 V3 with the Coolit ECO II comparisons, with the Gelid Tranquillo cooler there is absolutely no difference in cooling between the two orientations. There are a few variations in the thermal images, but most of them are actually caused by the slight height change when the case is turned on its side exposing slightly more of one component or another and not by any actual difference in temperature.

Winner: Tie

Other Considerations

Cooling performance aside, there are a few physical challenges to vertical cooling.

1. Having the air intake at the bottom turns the chassis into a vacuum for dust. If the system is not raised up off the floor, it will have a much larger dust problem than a traditional case as the intake fans are right at ground level where dust tends to congregate. Removable filters over the fans definitely helps, but in our experience it is rare that many people (tech literate or not) remembers to clean out the dust filters on their computer as often as they should.
    
2. All the ports for the motherboard and any PCI cards will end up at the top of a vertical cooling case. This may or may not be a big problem for the end user depending on the positioning of their case. The exposed ports on the top of the case will also make your system much more prone to damage if a liquid is spilled onto the case than a traditional case with a completely or near-completely sealed top.
    
3. Some components may not fit properly because most manufactures design for traditional oriented cases. For example, a long 5.25" device may not fit in the Silverstone FT02B-W as it may come into contact with a PCI card that is installed in one of the lower PCI slots.

Conclusion

In terms of performance, we were surprised by the results. Convection should (in theory) aid in cooling with a system running in a vertical orientation.  However, our measurements do not support that theory. The Antec P183 V3 showed our only large variance in cooling, but with the horizontal orientation running cooler. So what is causing these results?

Our conclusion is that the forces of convection in a computer are simply not strong enough to overcome even the turbulence and airflow eddies caused by the fans. The lack of any temperature variances when we tested with the air-cooler supports this theory, as air-coolers contribute more airflow and turbulence to the core of the system than a closed-loop liquid cooler (which only has a fan at the edge of the system). The fact that the Coolit ECO II in the Antec P183 V3 was running hotter in the vertical orientation does suggest that the hotter air in a system tends to congregate in the top of a system, but it appears that this is only the case when the air is relatively stagnant (no fans in the area). We believe that this result did not replicate in the Silverstone FT02B-W simply because the airflow in that case is so unrestricted that there is not this pocket of stagnant air for the radiator to sit in. Similarly, when the Antec P183 V3 is in the horizontal orientation, the top case fan provides enough airflow to vent the hot air out of the system before enough hot air can congregate and cause a rise in temperatures.

Our very own company President Jon Bach was so intrigued by these results that he dusted off his college textbooks to try to come up with a mathematical reason for our results. This is his take on the physics involved:

Let's do a thought experiment to see what we would expect in terms of the bouyancy of hot air inside a chassis, and how that relates to the air pushed by a typical 120mm fan. We'll use an average room temperature of 20C (68F).  In our thermal testing, under load with maximum temperatures, we saw chassis panels around 40C average.  Let's round it up to 50C so we're sure to give convection the biggest advantage possible.  Heck, let's double the average air temperature difference measured, and use 60C.

Density of air at 20C: 1.2041 kg/m^3
Density of air at 60C:  1.067 kg/m^3
Bouyancy of 60C air in 20C air:  0.1371 kg/m^3

A P183 chassis is 0.04 cubic meters in internal volume.  Therefore, if you fill an Antec P183 chassis with 60C air, and surrounding air is 20C, the air inside the P183 would have a bouyancy equivalent to the weight of 5.484 grams.  That's about the weight of two pennies. How does that translate into air pressure across a top 120mm fan?  The column of air coming into play has an air volume of 0.00576 m^3, creating a bouyancy of .79 grams over the fan's area of 144 cm^2.  That translates to 0.0054861 grams/cm^2, or 0.54861 mmH20 (the standard unit of measurement for fan static pressure).

That's about half the pressure put out by our Antec TriCool fans on their lowest setting.  That means that even one of the lowest pressure, quietest fans that we sell, would *double* the pressure necessary to overcome the air pressure caused by convection, in the hottest scenario we can imagine.  And this was after doubling the average air temperature increases we actually measured.  When using a 40C average air temperature, an Antec TriCool fan on low setting has 4x the pressure necessary to overcome convection.  Given the results of this thought experiment, the results of our emperical testing are making a lot more sense.  Convection, while a strong concept in thought, simply does not generate the results necessary to play a discernible role in a typical chassis.

What our testing has shown is that the Silverstone FT02B-W has great cooling, but it is not due to the case being able to work with the forces of convection.  Instead, the benefit is due to the fact that the internal layout adds very little resistance between the intake and exhaust fans. In traditional cases, the front intake has to go past the hard drive mounts (and any drives installed in those mounts) before it reaches the motherboard and other hot components. While the distance is not huge, the extra 5-7 inches and blockage by any drives is enough to cause a drop in the amount of directed airflow the intake fans can provide. In the Silverstone FT02B-W, the intake is at the bottom of the case rather than the front, which allows the front intake to completely bypass the hard drive mounts (turning the 5-7 inches into only 1-2 inches). By having the intake fans at the bottom, it also allows for the use of multiple large fans (in this case 180mm fans). Normally there are power buttons, LEDS, brackets and many other things around the front intake that the fans have to work around. By moving the intake away from these obstacles, you can have a very large rectangular area in which to mount fans almost as large as the case is wide.

In conclusion, vertical oriented cases can cool very well, but don't expect a case with an vertical airflow to automatically have superior cooling to a traditional case. You need to take into consideration how restricted the airflow is and the size/location of the fans. In the Silverstone FT02B-W, moving to a vertical orientation allowed for Silverstone to use larger fans and achieve a much shorter run from the intake to the hot components resulting in a very well-cooling case, but not all case manufactures will spend the time to ensure this same level of cooling. According to our testing, the internal layout of a case is still the most important factor to case cooling and not the orientation of the airflow.