In April 2025, LinusTechTips visited NASA's Langley Research Center at the invitation of the Flow Physics and Control Branch to seek an answer to the question: how far does a case fan have to be from a flat surface?
The article - authored by Dr.Louis Edelman - delves deeper into the test design and components, expanding on the approaches and methods used.
I’m kind of surprised there isn’t a CFD setup for PC builders yet. We could model all the parts and run some simulations, as the last hurrah for our old systems. Manufactures could show off their cred by making increasingly detailed models of their parts. It would all be very fun.
I do CFD in my day job, though not for electronics cooling. I don't think this is as easy as you imagine. It's relatively easy to make pretty pictures, but just because the picture is pretty doesn't mean that it's physical accurate or mathematically correct. Lack of resolution could be an issue, but there are plenty of more subtle problems as well. Jet impingement is known to cause problems with turbulence models, though some models claim to solve the issue. Plus, turbulence modeling isn't always predictive, and might require a certain amount of calibration any time a model is used in a new scenario. Add on top of that the fact that the computational cost of these simulations often is extremely high, even with turbulence models. Maybe people building PCs have plenty of unused CPUs and GPUs, though.
Unfortunately, I don't think CFD and turbulence modeling are things that you can just start doing well without learning a lot before starting.
You are probably right, my only exposure to CFD was through listening in at conferences, haha. It seems neat, though. They always had the coolest pictures.
I wonder, could there be any play in the fact that PC cases tend to be a little bit less general than just, like, any 3D model? There are only so many cases. Plus most of the parts are rectangular, and most of the surfaces are aligned the same set of axes.
Another tip: Blowing air onto a component, instead of sucking it away, is much more effective for cooling. Basically if you put your hand where the cooled component is you should be able to feel the airflow.
In practice this meant flipping the "exhaust" fans in one of my systems. It resulted in temperatures that were consistently a few degrees lower. Note in this instance there wasn't any ducting involved.
Yes, positive pressure is also essential for keeping dust off components, which becomes insulating if left unchecked. Make sure intake fans have filters and you're golden.
In April 2025, LinusTechTips visited NASA's Langley Research Center at the invitation of the Flow Physics and Control Branch to seek an answer to the question: how far does a case fan have to be from a flat surface?
The article - authored by Dr.Louis Edelman - delves deeper into the test design and components, expanding on the approaches and methods used.
I’m kind of surprised there isn’t a CFD setup for PC builders yet. We could model all the parts and run some simulations, as the last hurrah for our old systems. Manufactures could show off their cred by making increasingly detailed models of their parts. It would all be very fun.
I do CFD in my day job, though not for electronics cooling. I don't think this is as easy as you imagine. It's relatively easy to make pretty pictures, but just because the picture is pretty doesn't mean that it's physical accurate or mathematically correct. Lack of resolution could be an issue, but there are plenty of more subtle problems as well. Jet impingement is known to cause problems with turbulence models, though some models claim to solve the issue. Plus, turbulence modeling isn't always predictive, and might require a certain amount of calibration any time a model is used in a new scenario. Add on top of that the fact that the computational cost of these simulations often is extremely high, even with turbulence models. Maybe people building PCs have plenty of unused CPUs and GPUs, though.
Unfortunately, I don't think CFD and turbulence modeling are things that you can just start doing well without learning a lot before starting.
You are probably right, my only exposure to CFD was through listening in at conferences, haha. It seems neat, though. They always had the coolest pictures.
I wonder, could there be any play in the fact that PC cases tend to be a little bit less general than just, like, any 3D model? There are only so many cases. Plus most of the parts are rectangular, and most of the surfaces are aligned the same set of axes.
Cabling might be a problem.
Another tip: Blowing air onto a component, instead of sucking it away, is much more effective for cooling. Basically if you put your hand where the cooled component is you should be able to feel the airflow.
In practice this meant flipping the "exhaust" fans in one of my systems. It resulted in temperatures that were consistently a few degrees lower. Note in this instance there wasn't any ducting involved.
Yes, positive pressure is also essential for keeping dust off components, which becomes insulating if left unchecked. Make sure intake fans have filters and you're golden.
Was this one of the trips on the good-as-new plane? The video was an interesting watch.