With the unveiling of their Radeon Pro Duo graphics card AMD also unveiled their GPU roadmap that looks as far into the future as early 2018. By that time AMD will have launched three new GPU architectures. This starts with the launch of its 4th generation Graphics CoreNext architecture, codenamed “Polaris”, which is expected out by the middle of this year. Polaris is built on the 14nm FinFET process and is expected to offer 2.5x increase in performance-per-watt compared to its GCN 1.2 architecture on 28nm.
Taking a look at the roadmap we can see that following “Polaris” will be “Vega” in early 2017. It appears that Vega will offer a 50% increase in performance-per-watt over Polaris, but the main highlight in the roadmap is that “Vega” will feature HBM2. So does that mean that AMD will be skipping HBM2 on Polaris? We already know that NVIDIA is going with GDDR5X to start on Pascal, could AMD be doing the same? This means that AMD would launch a performance-segment GPU first as an enthusiast product, and then launching a Vega-based big chip with HBM2 in early 2017.
Following Vega we will see the “Navi” architecture, which looks like it will offer 2.5x performance-per-watt over Polaris. It will take advantage of a new memory standard.