Today, the current bottleneck is not due to memory speed/bandwidth on the CPU<->Memory side.
The bottleneck is in the frequency of the CPU and in the software that makes use of the core parallelism. Today, the (general purpose) software is not efficient enough to make use of the so many cores of our chips. Since not all the cores need data at the same time, the memory BUS doesn't get overloaded so we don't have any bottleneck. Some tasks like video encoding (handbrake) make good use of all the cores, but those are exceptions.
There is another task where high memory bandwidth are used. Graphics (GPU or using the APU). In those cases, due to the way graphic programs work, they need a lot of data. In other words, they need a lot of memory bandwidth. And you can easy overload the data bus increasing the resolution;.
In those cases, you would benefit of HBM2 and would be interesting to see solutions integrating HBM2 in the motherboard.
independently of the current need, my personal opinion is that it would be good to have the option of using the CPU to support the GPU with some graphical tasks and other way around with directx12 or vulcan. Maybe, some cores can be dedicated for that in the future, so it could be interesting if INTEL and AMD see that niche and sell us some motherboards that integrates HBM2 for their CPU chips.