AMD has also taken advantage of Computex 2018 to showcase its first graphics core based on the Radeon Vega architecture at 7 nm, a major breakthrough that will enable a dramatic improvement in the features of this architecture.
The new 7nm Radeon Vega GPU is designed for the high-performance computing (HPC) sector, such as servers and workstations. This new silicon is specially designed for deep learning operations in artificial intelligence and automatic learning. In late April of this year, information emerged that AMD was preparing its Vega 7nm GPUs to select customers for the Q2 2018 with specific availability sometime in the fourth quarter of 2018.
The Radeon Vega architecture at 7 nm is based on fifth generation GCN, just like the previous Vega 10 core at 14 nm. AMD has talked about some of the benefits of moving to the 7nm process, which will be at the heart of the company’s GPU roadmap until at least 2020. The move to 7 nm allows Vega to improve its performance by 35% while doubling energy efficiency and taking up half the space, significantly lowering the manufacturing cost.
AMD has shown a prototype of the Radeon Vega Instinct with capacity for 32 GB of HBM2 memory. AMD has released the specifications, but from what we know so far the card could include a 4096-bit bus using four 8GB HBM2 memory stacks each. Unfortunately, at the moment there are no plans to bring Radeon Vega at 7 nm to the gaming sector.