In engineering terms:
a. The less work components have to do, the less heat they will use.
b. The less components move, the less noise they are likely to make, and the less wear on bearings.
With any semiconductor-based components, transitions of state use the most heat, because they are transitioning through their analog phase, so the higher the number of transitions per unit time, the more heat they produce, requiring more cooling.
So, to put it into practical terms:
a. Over-spec the system, rather than push it to its limits, as all components will be less stressed, with less transitions compared to their heat dissipation ability.
b. Use larger, slower rotating fans, as they will generally be quieter. Around 120-140mm, 400-500 rpm or less. They will also collect less dust per area on their blades, so will be carrying less weight, leading to less stress.
Unfortunately, higher audio processing capability is at the expense of noise.
The balance is always between:
a. Performance, including minimum noise.
b. Cost.
c. Wear.
There is a sweet point, but it will always have compromises. If audio recording amateurs were as fastidious about their machine’s performance as gamers are, and willing to pay up as much, a lot of equipment discussions here would be moot. Quiet, high-performance audio costs a lot of money compared to normal computing, so get with the program and cough up!