Good news for NVIDIA, because Samsung has announced that next year will begin the mass production of HBM2 (High Bandwidth Memory) memory chips, which will help the green team to not depend only on SK Hynix, which gives absolute priority to AMD as they are already selling graphics card based on HBM1 memory (the Radeon Fury series). With the advent of memory HBM2 that priority will not change, then AMD will have priority to access the memory HBM2 against NVIDIA.
While HBM1 memory has a physical limitation of up to 4 GB of capacity and a bandwidth of 512 GB / s, memory HBM2 allows the use of up to 16/32 GB of VRAM that can exceed 1024 GB / s.
Samsung should not not ignore that in order to full fill the demand for NVIDIA they have to manufacturer large amount of HBM2 chips as at the moment NVIDIA has almost 80% market share.
This time Samsung is not late, since it is expected that NVIDIA Pascal GPUs will arrive in late 2016 and now they have more than enough time to reach an agreement with Samsung and begin to supply the HBM2 chips for mass development of high-end graphics cards. Samsung does not stop there, as it is expected that in 2017 an even super-fast type of memory will reach the network market. The plan is to see HBM 2.0 mainstream by 2018.