AMD recently released the preliminary details of its upcoming Vega GPU architecture, outlining its design and improvements.
AMD plans to launch its latest high-end desktop graphics card range this year, and these cards will be based on the Vega GPU architecture.
Current Radeon RX-series graphics cards are aimed at the mid-range performance market and do not compete with Nvidia’s high-end Pascal cards: the GeForce GTX 1080 or Titan X.
The company has not revealed details of specific Vega cards, but it has outlined some big changes which could shake up the desktop GPU market.
AMD’s Vega graphics cards will be a departure from traditional GPU design, using new technology to provide solutions not yet offered by Nvidia’s desktop cards.
It is important to note that the Vega architecture is aimed at high-end gamers and hardware enthusiasts and will have additional uses in professional and machine-intelligence fields.
While Nvidia’s high-end Pascal cards use enhanced GDDR5X VRAM, AMD has opted to use a new type of memory for its Vega cards: HBM2.
HBM was used on several Fiji cards, although they were constrained to a maximum of 4GB due to the limitations of the first generation of the technology.
With HBM2, graphics cards can sport up to 32GB of VRAM, while having a smaller PCB footprint than GDDR5.
Another advantage of HBM2 is an increased bandwidth-per-pin, double that of the first iteration of HBM.
This means that Vega graphics cards could have higher memory bandwidth in addition to larger amounts of VRAM.
GCN vs NCU
AMD’s Polaris graphics cards used traditional GCN compute units, which Vega will forgo in favour of AMD’s new NCU microarchitecture.
This means the graphics cards will no longer be limited by traditional issues associated with GCN and will be built to handle any operation – from gaming to professional workloads.
It is unknown how these changes will affect performance, but Vega’s deviation from traditional GPU design means it could have the potential to outperform Nvidia’s Pascal lineup.
AMD said Vega graphics cards will launch under various Radeon brands, depending on their operation.
The company will release Vega GPUs geared towards gaming, professional, and deep learning under the brands Radeon RX, Radeon Pro, and Radeon Instinct respectively.
“It is incredible to see GPUs being used to solve gigabyte-scale data problems in gaming to exabyte-scale data problems in machine intelligence,” said AMD senior vice president Raja Koduri.
“We designed the Vega architecture to build on this ability, with the flexibility to address the extraordinary breadth of problems GPUs will be solving not only today, but also five years from now.”
Graphics cards based on the Vega architecture are expected to ship in the first half of 2017.