We have some more information about the AMD’s next-Gen Greenland GPU, which will be fully based on a new microarchitecture which has been developed from about two years. The most interesting thing is that they leave the node of Graphics Core Next (GCN) architecture with a new ISA (Instruction Set Architecture).
AMD’s Greenland GPUs will feature between 8-16GB of super-fast HBM2
This new architecture would be 200 percent more efficient compared to GCN, all accompanied by a manufacturing process that would be between 16nm FinFET and 14nm and give a serious leap in terms of the number of Stream Processors in comparing with Fiji architecture (which gives life to the Radeon Fury) by the use of 15 to 18 billion transistors (Vs 17 billion in NVIDIA Pascal). The new AMD Greenland GPUs powered GPUs would only be available in models of 8 and 16 GB of HBM2 memory, which provide twice bandwidth that throws the current HBM1.