Going through the FPGAs board manufacturers came across the HMC(Hybrid Memory Cube) http://static6.arrow.com/aropdfconversion/a034c723e572f45efe84162a58a14f71cedf33fa/hmc_webinar_july_2017.pdf.pdf promising technology basically coupled with UlTraScale FPGA or Intels Aria 10 FPGA you have the bandwidth and capacity to crunch any memory hard algo more efficiently than GPUs and even todays ASICs. Its supper expensive compared to ddr3-4 but in long term investment will return its money
This PDF you linked is a bit... "too much technical talk"... but from what I get they're describing this: https://en.wikipedia.org/wiki/3D_XPoint which is around some time now (the PDF you shared is from a year ago).
I know Intel is using this technology but it's veeeery expensive and I can't see it being used for mining anytime soon. I didn't know Micron has something similar too though, thanks for that!
In any case, this is not an FPGA-specific feature. It's a technology; and since Micron is in collaboration with Nvidia, if this technology becomes mainstream then expect to see this kind of memory implemented it in RTX 2180ti or something.