Intel uses a hybrid approach to this.
With your 3 modules of 8GB of ram:
- First 16 GB of your memory will get dual channel speed
- Last 8 GB of it will get single channel speed.
Windows perfectly capable to handling this, as it existed for quite a while (Since the age of Core2 actually).
Quite ingenious, and if your memory utilization stays below the threshold, you'll see the same performance as with just pure dual channel.
p.s. - The unused memory is actually used as disk cache, but for that the single-channel speeds are WAY more than enough.
The same is valid for Quad-Channel systems - It uses as much as it can in the largest bandwidth possible, then dropping speed for the rest of the memory.
It is quite possible to make a crazy system on X299 with 32+16+8+4 ( 60 ) and still get full quad-channel bandwith on the first 16 GB, then triple-channel on the other 12, then dual-channel on 16 more and finally single channel on final 16 GB
~~~
I do not know how Ryzen handles this, the last AMD that I used was an Athlon 64 and that one was reverting to Single channel in this case.
But I guess the new Ryzens are smarter and doing a similar approach as modern Intels...