I did not understood really well your question, so i will guess what you wanted to ask.
I suppose you are wondering why the integrated Vega 10 gpu included into the Ryzen APU's did not used dedicated memory instead of the system shared VRAM.
Since the gpu is included into the cpu, together into a same chip, one would need to provide some connections to the external memory module, like the one you find on a normal gpu.
It is not an issue when one deal with a normal gpu instead of an APU, one have enough space on the chip itself to fit all the connections needed to link 4, 8, or X memory modules.
But since the gpu is integrated into the cpu, one can't freely add extra connections to the APU chip and socket, since a lot of the existing connections are already used for cpu purposes.
Hence it's far more easier to use the system shared memory instead adding external ram chip, since one can use the existing shared memory connection already designed for the cpu also for the integrated gpu work.
Obviously the downside is pretty obvious since the gpu is now tied to the system memory speed and capacity for the gpu work.
To be honest, the gaming results on these Ryzen APU are already really good with fast dual channel memory, i suppose future APU will do even better.