29 December 2021

Case airflow and design...

A false colour image of a turbulent jet; C. Fukushima and J. Westerweel

There's a topic that I've never really broached before on this blog and it's mostly because I haven't had the time though, more precisely, it's because I haven't had the time to do a deep dive into the background of the topic so that I can talk semi-intelligently on it.

However saying that, while I'm not an expert in fluid dynamics and flow design, I do have a lot of practical experience in industrial and research/academic airflow design between building and designing systems that can achieve 10-6 bar and refill to 10-100 bar, to enclosed and open airflow glove boxes. Working with these systems has given me a decent foundation on the principles, pitfalls and limitations of certain designs. 

While I've actually had a post in mind on this subject for quite a time, I wasn't really motivated to really put any effort into it as I've seen movement in the subject from outlets like GamersNexus and LinusMediaGroup in this field and, undoubtedly, they can do a better job than I can in real terms because they have the equipment and funding to be able to do so. Comparatively, my keyboard warrior abilities of putting text (and sometimes pictures!) on a screen really do pale by comparison.

So what did motivate me to write this post? 

Linus' latest case airflow video did and the reason it did is that I believe that they are missing out on a lot of issues in airflow design, causing them to draw some questionable conclusions - something that GamersNexus will not be able to cover with their fan testing equipment either.

So, while I'm not an expert, I do have some expertise in this area so let me give my perspective on case cooling and see what people think about it.

The important subjects...

Let's start off with a quick overview: What are the important things to think about?
  • Flow characteristics
    • Turbulence
    • Laminar flow
    • Volume of flow
  • Fan characteristics
    • Pressure across the fan cross section
    • Blade design
    • Diameter of the fan
    • Speed of the impeller
  • Chamber design
    • Dead zones
    • Flow paths
  • Pressure
    • Static pressure
    • Dynamic pressure
    • Total system pressure
      • Positive
      • Negative
I'm sure I've missed out a bunch of things but these are some of the basic ideas to be thinking about. I won't touch on everything in the above list but will do my best to summarise what I think are the important aspects for case airflow design. 

So, to quote a literary genius: "let's start at the start and take it away".

Example of a ducted test setup from the AMCA Standard 210-16.

Theory vs. Reality...

A lot of people speak about consumer PC case airflow design quoting truisms from mathematical theory and ducting design. They may have some element of truth to what is being said but for the most part, none of these considerations really match up with the reality. 

For example, I see many people stressing about the static vs dynamic pressure performance of the fans they are buying but for the most part it's a foundationless worry because all of these P/Q curves and statistics are generated from a fictional, idealised environment (usually a straight piece of tubing with a certain amount of fan diameter lengths before and after the position of the fan being tested).

There are test setups designed for non-ducted free inlet and outlet setups (as we find in the case of PC designs) however, I have not been able to verify how any PC fan seller/manufactuer tests their fans or whether the methods are comparable across brands.

Further to this, to the best of my knowledge fans are never tested with respect to operating in an environment that includes other fans operating on the same system volume.

A test setup for free inlet and outlet environments (closer to what the environment encountered inside of a PC case) - also from the AMCA Standard 210-16.

These conditions do not exist within a PC case and so the value of the specifications listed for PC fans is relatively limited for the average consumer. In fact, multiple independent testings have shown this. Unfortunately, for this reason, even these independent testing environments only apply to those specific environments and you can see in each scenario, the environment is different but most importantly, each test is being performed on a single fan - not the case for the majority of gaming PC users.

Static vs. Dynamic...

The benefit of a static pressure fan is specifically observed when the exhaust side of the fan is constrained in volume*, not the inlet. i.e. exhausting into a heatsink fin array or into an enclosure where the exit is of a smaller diameter (or overall surface area) to that of the fan. In these configurations, a static pressure fan can achieve the same rate of flow of air as a dynamic pressure fan will at a lower speed of operation, increasing efficiency of operation - presuming that increased static pressure is your goal.

Dynamic pressure fans are designed to move a large amount of air through a system without build-up of pressure or without expectation of back-pressure build-up due to the system design. Think of the blades of a propeller-driven aeroplane which have a higher angle of attack... or in PC fan design. Of course, a lot of modern aeroplane propellers are able to dynamically alter their angle of attack in order to optimise their performance at a wide variety of airflow speeds, whereas in the PC fan space, we're still stuck in the dark ages of static props (though i can't imagine how expensive such a fan would be!).

At any rate, the point is that these fans are designed in such a way that any build-up of pressure within a system will quickly nullify their efficiency, reducing the air volumes pushed through a system in a given unit of time. The same caveat that applies to static pressure fans also applies here too - if you block the fan intake or exhaust, you ruin the performance of the fan!
*Please note the difference between a constrained volume and a ducted volume - they are not the same! e.g. A static pressure fan will work better if the inlet is ducted but the outlet is free than in a situation where the fan is just blocked. A static pressure fan is not a "fix" for drive cages blocking air access or when there's very little clearance before the fan fixture for air to pass.

Under pressure...

There is one more important feature to note about fans and that's how they affect the airflow - there's a good amount of energy-driven turbulence directly at the exhaust of any fan assembly which, over distance dissipates to become a more laminar (or less turbulent) flow due to surface friction. Usually, in a tube or cylindrical duct, that's approximately 10 diameter lengths from the axial fan in question unless the airflow is realtively low and thus the calculated Reynold's number is also quite low.

Of course that estimate comes back to the whole "theory vs. reality" thing and, in reality, there is no truly laminar flow in a "free" system as we observe in PC cases. However, it does bring me to my first point - moving less turbulent air more slowly can cause more changes in air volumes, which means bringing more fresh air into a case which allows your components to cool more effectively. 

Conversely, pushing too much air into a case more quickly will cause more turbulence, back-flow and other effects that actually reduce the amount of air volumes being passed through the case, resulting in worse thermals on your components as the air within the system will need to absorb more heat.

This brings us neatly onto the topic of...

Please forgive this crude drawing of turbulent airflow through an axial fan with a free inlet and free outlet, as is often found in a PC case...


"A three-dimensional time-dependent motion in which vortex stretching causes velocity fluctuations to spread to all wavelengths between a minimum determined by viscous forces and a maximum determined by the boundary conditions of the flow. It is the usual state of fluid motion except at low Reynolds numbers." - link to secondary source.
Turbulence is bad, yo! ... and it explains the problem with lack of reproducibility of results between different labs and consumer test outlets such as LTT, Anandtech and GN. Seriously, everyone likes to talk about how great turbulent flow is because it's better at cooling because of the greater "surface area"* and there are many examples of people stating that you "want" turbulent flow without really specifying exactly where you want it. 

Specifically, you want turbulent flow at the point where you wish to extract (heat) energy, otherwise, you do not want turbulent flow for one simple reason: anything that disrupts the flow of air is detrimental to the velocity of the air and, by extension, the volume of air being moved through a system.
*More accurately: the molecules in the air flow are interacting with each other in a greater manner, providing more collisions and thus opportunities for transference of energy.
Here, people might begin waxing lyrical about how we should be talking about vortexes when speaking about airflow in turbulent systems but, in my opinion, they're missing the wood for the trees - turbulence is the presence of vortexes, they are one and the same and at this scale (and for the purposes of case airflow design) we don't need to focus on the micro effects. So, I'll just refer to the entire regime as a turbulent system and be done with it.

Cause and effect...

From my understanding, turbulence is caused when airflow is disrupted by a change in velocity due to difference in pressure caused by a difference in momentum between various molecules within the airflow across a cross section in volume. There are three primary causes for difference in pressure - compression of an air volume (e.g. caused by an impeller acting on the volume of the air, or airflow moving into a smaller volume of pipe), friction (caused by part of the air volume interacting with elements that cause drag through physical collisions with the molecules in the airflow - e.g. a surface), and diffusion of an air volume (caused by exit from a smaller volume to a larger one - e.g. an exhaust).

Comparison of single and dual inlet fans on airflow - an example of poor test setup. [LinusTechTips]


Taking into account that turbulence reduces air flow and air volume changes per unit time, one major aspect that many PC fan setups never take into account (as I stated previously) is the effect of multiple fans on each other. Unfortunately, nobody ever designs their cases like the server industry which takes into account the flow of air around components and they design the inlet and outlet fans to work in harmony in order to optimise the rate of air changes per unit time which improves component cooling.

What you will also notice about the server industry is that they do not include fans on their components, everything is "passive" (well, in reality they rely on that airflow provided by the case fans). The reason for this is that fans operating in a perpendicular plane (and also in the plane but at a different differential pressure across the fan cross-section) to the airflow will disrupt that airflow, increasing turbulence, including back flow, and reducing the amount of fresh air that gets to each component - especially those downstream.

You'll also notice that in server designs, each fan is ducted so that its airflow does not interfere with the one sitting next to it in the housing. This is a classic mistake from PC builders (as observed in Linus' video, screen-capped in the above picture). An aspect of unducted, freely situated fans is that the turbulence generated by the fan itself reduces the efficiency of its own airflow causing backflow from the exhaust to recirculate into the inlet of the fan.

Placing more than one fan adjacent to each other in this configuration will not double performance and you will not get backflow between the edges of the fan assemblies but instead generate WAY more turbulence which in turn reduces the pressure differential across each fan assembly, decreasing airflow and thus decreasing the amount of air volumes being pushed through the fan.

In Linus' video, the blame is placed on the limited intake of the front side vent, however the turbulence generated from the two high speed fans means that the size of the vent is the least of their concerns - ignoring the fact that the vent definitely has more surface area than a single or dual fan surface area.

The issue with this case design isn't that the vent placement is sub-optimal it's that the system between the vent and the exhaust of the fans is not closed and that the multiple fans at the front of the case cause more problems in this particular layout by reducing the efficiency of each other through turbulent interference.

A depiction of the airflow optimisation in a Dell server rack. [Source]

The goal...

In order to have the most optimal cooling solution for a PC case, we need the largest volumes of "clean" and cool air flowing through a system with a decent amount of turbulent flow at the point of heat extraction. This is our holy grail. Unfortunately, consumer PC cases are not really designed to be great at this and neither are PC components designed in this manner.

Which brings us to the final point...

An excellent example of an NZXT BLD kit. Personally, I think these are a great idea, though disclosure - I own an NZXT 510 case as depicted here... in fact my self-built PC is effectively this build, only with more RAM.

Component effects...

It's entirely possible to optimise your design of your case shell for airflow. It's reasonable to configure your inlet and outlet fans to provide a good flow of fresh air whilst reducing static pressure but once you put the actual components of your PC into the case, all that effort goes out of the window.

Let's face it - as I pointed out during the discussion of server designs, component airflow design is (for the most part) antithetical to the idea of case airflow.

CPUs are pretty bad - in a downflow configuration of the low-end fan coolers, they take perfectly good airflow and treat it to a perpendicular turbulent flow. In a flow-through design, they can cause back flow through pure blockage of a faster-moving stream of air or, in a worst case scenario, cause double back flow and turbulence through a two-fan configuration, causing build-up of pressure before the final exhaust in the system behind the CPU tower cooler.

However bad CPUs are, GPUs are by far the worst contenders for disruptions to case airflow. Aside from being the cause of flow segregation, reduction in air pressure across case inlet fans (see below), dead spot creation and perpendicular flow creation (i.e. turbulent flow and flow velocity reduction), consumer GPUs are not designed to work with any case designed airflow paradigms.

Great from a build quality perspective... not fantastic from the point of view of airflow, though better than Linus' design... (I forgot a red arrow from the external exhaust of the GPU back into the case from below it in the PCIe slot covers)

Yes, that's right the beloved GPU is our own worst enemy in airflow design within our beautiful cases. This is something that Linus mentioned in his video but I think that the issue runs much deeper than that they used a GPU that was too large for the case - it's not that the GPU was too large, it was the fact that the GPU doesn't take into account the case design and neither does the case take into account the GPU design.

Seriously, there's a big to-do about the whole ATX12VO standard but that's a small optimisation to a solution that's already in place. Consumer GPU airflow designs just do not work in any case standard - whether that's ATX, micro-ATX, BTX, ITX, etc. etc. Even cases that mount the GPU at the front side of the case in the plane of the motherboard do not help because the flow of the GPU air is in a plane that is different from the airflow through the case.

So if our own components are working against us, is there any point in trying to optimise case airflow designs?

It's important to find the right balance for your particular system... [Jeppe Hove Jensen]

Working towards better airflow design...

The problem isn't intractable. There are more optimal solutions to designing the airflow in our consumer cases without transitioning to the highly controlled and highly undesirable form factors of the server market.

Here is my list of things that I think the average person can do to improve their airflow design (and this applies to the Linus Tech Tips video as well):

1) Stop putting too many fans in your case.
Yes, you! You and your five fans can go take a hike! They're not helping - they're causing too much turbulence through contraflows and back flows within the case! Look up there at that NZXT case - remove the top front fan, it's only causing interference with the exhuast fan of the water-cooled CPU OR issues with that tower cooler as in Linus' video. If you have a fan above the CPU, get rid of that and block the frickin' holes too! They're terrible because that position is either: 
  • a) drawing in fresh air that never reaches any hot components, AND it reduces total system pressure (a combination of all stages of dynamic and static pressures) which will reduce airflow velocity and the total number of volumes of air changes per unit time... or 
  • b) without a fan in that position, just reduces total system pressure and therefore airflow velocity and the total number of volumes of air changes per unit time...
And on that front;

2) Block unneeded air access to the case and case sections.
Seriously, I love my NZXT H510 but it has no compartment separation. The inlet filter at the front side of the case? Not blocked by anything - the airflow of any front-mounted fans can pull from underneath the main volume or from behind the board or (the worst case) even from the main volume itself. That's terrible for airflow. 

Want to reduce turbulence and increase air volumes changed? Block off unnecessary inlets of airflow that can cause recirculation or can reduce air flow performance. (Yes, that includes blocking off that top-space above the CPU!)

3) Reduce the speed of those case fans.
Case fans don't need to be running very quickly to get air flowing into and through the case. Additionally, the lower the speed of the fans, the lower the turbulence and the lower the noise. If a component needs more or less cooling, let it ramp up or throttle down. Let the case fans grant a consistent movable airflow, not something that will be fighting against the component airflow designs.

4) Install airflow separators or a baffle between the CPU and GPU.
Yes, it's ugly and ruins the aesthetics of your PC build but you don't care about that - you care about cool and quiet. Watch that LinusTechTips video and you'll see the CPU (running at a high speed) steals the airflow from the GPU because it is in-line with the axial fan providing the airflow to the bottom of the case. There's no reason for the CPU and GPU to share the airflow in any given system and this is proof of it*.
*Head down to step 6 and you'll see that the design of this case removes any chance of airflow intermixing between components. It's a simple design but effective.

5) Forget about positive and negative pressure setups.
You'll never achieve any sort of appreciable pressure differential from atmospheric in any commercially available case design (they're too leaky) or with the power of the axial fans you can buy in the 120-140mm range.

Seriously, it's mental masturbation. Stop it! You'll go blind.

Focus instead on air volume changes and reductions in turbulence and you will see that a case "at atmosphere" will perform better than one targetting negative pressure or positive pressure. Whatever those pertain to actually be...

Am I crazy or is this the end...?

So there we have it. My thoughts and guidelines on how to make the airflow in cases better... until GPU manufacturers get their heads in the game and start designing the airflow of a GPU with case fluid dynamics in mind...

Let me know if you love this or hate this. Let me know if I've changed your life for the better. Seriosuly, feedback is appreciated because nobody gets better in a vacuum. Except certain types of space aliens...

And with that warming thought, I wish you adieu!

No comments: