AMD has already released its rough GPU architecture roadmap showing both the upcoming Vega GPU architecture, scheduled to come sometime next year and the new Navi GPU architecture, scheduled for 2018. Today, we have some first information on the Vega GPU architecture.

AMD's Vega 10 is expected to be the flagship GPU in the Greenland GPU lineup, featuring anywhere between 15-18 billion transistors. As it will be competing with Nvidia's high-end GP100 GPU, it could pack up to 4096 Stream Processors based on AMD's new Graphics Core Next 4.0 GPU architecture.

It is obviously clear that Polaris GPU architecture will skip the 2nd generation High Bandwidth Memory (HBM2) as that one will be reserved for Vega GPUs, offering up to 1TB/s of memory bandwidth.

These information also suggest that Polaris GPU architecture will be quite similar to already available Fiji GPU architecture, only featuring improvements coming from 14nm FinFET manufacturing process, while the next big architecture changes should come with Vega GPU architecture.

Nvidia is expected to unveil more information on its next-generation Pascal GPU architecture at GTC 2016 in April.



Source: Wccftech.com.