AMD Unveils Its Vega GPU Architecture for 'Exabyte-Scale' Computing
By Shirley Siluk / CIO Today. Updated January 06, 2017.
Set to be released sometime in the first half of this year, AMD's newest GPU architecture -- known as Vega -- is designed to tackle complex computing challenges in not only gaming but in graphics design and machine intelligence, the company said yesterday.
AMD's preview of Vega was short on some details such as when exactly it will hit the market and how it will be priced. However, the company described some of the new GPU's highlights, including a new memory architecture, an advanced pixel engine and improvements in the compute engine and geometry pipeline.
Vega's architecture is aimed at improving the handling of data-intensive workloads that previous GPU designs haven't been able to manage effectively, according to AMD. In development over the past five years, Vega "enables new possibilities in PC gaming, professional design and machine intelligence," the company said.
Flexibility for Future Demands
"It is incredible to see GPUs being used to solve gigabyte-scale data problems in gaming to exabyte-scale data problems in machine intelligence," AMD executive Raja Koduri said in a statement. "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."
Koduri, who is senior vice president and chief architect for AMD's Radeon Technologies Group, added that Vega's new high-bandwidth cache in particular is "a pivotal disruption that has the potential to impact the whole GPU market."
GPUs are increasingly seen as having an edge over CPUs when it comes to data-intensive compute demands, especially those that require a lot of parallel processing. AMD said Vega will deliver even further improvements through a new memory subsystem design and high-bandwidth cache controller that allows it to "address very large data sets spread across a mix of memory types."
Applications for Machine Intelligence, VR, More
AMD is touting the coming release of three Vega-based technologies: the Radeon RX designed for gaming; the Radeon Pro for animators, filmmakers, graphic designers and engineers; and the Radeon Instinct (pictured above) for machine intelligence in server computing.
First unveiled in December, the Instinct accelerators, for example, are designed to help companies better use and understand the huge volumes of data being generated by today's applications and devices, according to AMD. Instinct is aimed at providing "a blueprint for an open software ecosystem for machine intelligence, helping to speed inference insights and algorithm training," the company said.
Meanwhile, the Radeon Pro technology will enable high-resolution visualizations for virtual reality applications, AMD said. Uses could include "immersive and instinctive computing" such as real-time collaborative design reviews and walkthroughs by product teams working remotely in different parts of the world.
According to the investing site Seeking Alpha, AMD's new Vega architecture "should make AMD once again competitive with both Intel and Nvidia and provide the company with significant growth potential."