When you set your fan speeds in the BIOS after building your PC for the first time - if you're working with a modern motherboard/bios - you'll likely have access to a tool that resembles a graph with dots on it, where the Y-axis is fan speed in % of max speed and the X-axis is temperature values in degrees celcius. This is, if you are using PWM fans (fans that change speed based on temperature information sent from the motherboard), the most straightforward way to set up your fan profiles. A profile determined by dots you place on the graph will tell the fan under what conditions (i.e. at what temperatures) to spin at a certain percent of max speed.

Here's my observation - fan speed changes are more distracting than (relatively) loud fans. And my suggestion: Don't make a straight or consistently ascending line on the graph, for example 40 degrees = 30% speed, 50 degrees = 40% speed, 60 degrees = 70% speed, and 70 degrees = 100% speed. Don't do this because the result will be that your fans constantly speed up and slow down dynamically as the temperature changes under normal usage at safe temperatures. This is, for me, much more distracting than having the fans run consistently at one speed a bit faster/louder. Here's what I've done instead:

In my bios fan control settings there are four dots on the graph and a graph for each fan. I've set all of the fan graphs up as follows: (point 1) 0 degrees celcius = 50% speed; (point 2) 60 degrees celcius = 50% speed. These two values ensure that the fan runs at 50% the majority of the time - namely when the CPU temperature is anywhere between 0 and 60 degrees celcius. This is both quiet enough for me and keeps the temperatures around 35 degrees at idle. The fan speed doesn't change at all up until the CPU reaches 60 degrees, which is the max temperature I've observed any CPU core to reach under normal non-gaming or rendering workloads with the fans at 50%. So to summarise: now, most of the time the fans are running at 50% speed and the CPU temps are below 60 degrees; Next, the last two dots are set with CPU intensive scenarios in mind. The third point is 65 degrees = 75% speed (could be 80% speed if you're worried about thermal runaway, or 70% speed if you want it to be a bit quieter at this temperature, there's room for adjustment here), and the fourth 70 degrees = 100% speed. The reason for this 10 degree span is twofold: For one, this is the temperature range which is easily handled by my cooling solution with fans at close to full speed when the CPU is at 100% utilisation. This is also (70c) the max temperature I would prefer my CPU to run at for longer duration, although it is well below throttling temp and even 10 degrees below the widely assumed safe operating temp for my chip. Just an aside, putting both latter values (graph dots) at 70 degrees = 100% speed should have had the same effect, as the fan speed would increase linearly by 100% over 10 degrees in either scenario. I just like to use all the dots.

These values are an example. EDIT: as other users have pointed out, the fact that these values work for me doesn't mean they'll be perfect for your cooling setup. They could be a bit too relaxed if your chip is quite hot in general or too aggressive from a noise standpoint if the CPU tends to briefly fluctuate above 60 degrees under medium workloads. END EDIT. The key to quiet fans is to prevent them from switching speed at lower, "safe" temps, and instead to have them aggressively ramp up when the temperature goes over whatever value you're comfortable having the CPU run at over a longer period of time. Note, this isn't my approach with my graphics card, there I'm a bit more focused on avoiding thermal runaway and tend to have the card fans ramp up to 100% long before any thermal throttling would occur, as well as have them run faster at lower temps to prevent the near-throttle temps from ever occurring. In general, a graphics card will reduce its processing speed (core and/or memory clocks) before it hits a safe thermal limit - for my rx 480 the performance throttling temperature is 83 degrees, but the "safe" max temp is over 90. CPU's won't thermal throttle their processing capability (clock speed) based on temperature until they reach unsafe temperatures, so the motivation for fan speed curves is different for these two different scenarios.

EDIT: update 11.02.2020, Ryzen 2 CPUs WILL throttle their boost clocks long before reaching unsafe temp levels. For example, my CPU is at 65° under an all core 100% load, my boost clock on all cores is about 3990-4000mhz; my CPU is at 70° in the same scenario, the CPU boosts to 3950mhz. This may also be the case with intel CPU "turbo boost" level clocks. So it's up to you to ascertain if this is the case and then choose a performance and noise level you're comfortable with if you want your CPU to consistently boost as high as it can with your respective thermal solution (CPU cooler).

What a CPU still won't do is drop below stock clock speeds until it hits unsafe temps (thermal throttling), and in this way it is similar to gfx cards whose algorithms consider core/memory load as well as temperature to decide when to boost above stock ("boost clock", "game clock" or whatever your GPU AIB chooses to call the above stock overclock).

Just thought I should update this because I wrote it long before experimenting with my r5 3600 and the case cooling. Spending an inordinate amount of money on good airflow+static pressure fans nets me about a 50mhz all core boost clock with an aggressive fan curve by keeping CPU temps between 60-65 at max utilisation. Wasn't worth it.

Hope this helps some people out there, happy building :)

Late edit: Link to second post with additional information on fan curves.