According to Business Korea, Nvidia will be getting a supply of Second Generation High Bandwidth Memory (HBM) from both Samsung and SK Hynix in order to build Nvidia’s next generation GPUs. Also these chips are planned to be mass produced by Q1 2016, after reliability tests at the end of this year. Although first generation HBM is already in use in AMD’s R9 Fury and Nano series GPUs, anyway these GPUs are made using SK Hynix chips exclusively.
Probably, HBM memory already showed immense power savings over traditional DRAM memory and can also operate at 4-8 times the speed, making HBM an attractive technology for more than just GPUs. It is also said that, Second generation HBM would be double the speeds of the already impressively fast HBM 1.0 memory and will allow capacities which are 4x larger, meaning that a single chip of HBM 2.0 could have up to 4GB of capacity. This will allow future HBM based GPUs to have more and faster memory than that seen in AMD’s 4GB R9 Fury X GPU.
Future is HBM powered GPU that could have up to 16GB of memory if both AMD and Nvidia were to continue using 4 chips of HBM on their GPUs, which will be more than enough for any of Gaming needs.