AMD will release its new GPU architecture in the first quarter of 2017, and it will be called Vega.
The company officially announced at CES 2017 that Vega was designed to deal with complex computing processes, and gaming won’t be the only one to exploit its features. According to the company, AMD’s Vega will also be remarkably good running tasks related to graphic design and machine intelligence.
The company didn’t share some important details about its new product at CES 2017, such as the exact release date or its price. However, AMD did reveal some of the most notorious features the new GPU will have.
Vega will count with a new memory architecture, advanced pixel engine and significant improvements in the compute engine and geometry pipeline among other characteristics.
Vega will tackle complex computing and machine learning
AMD said Vega was devised to improve the handling of data-intensive workloads, something that the previous Graphic Processing Units cannot accomplish in an efficient manner. That means AMD will provide new possibilities for graphic design, PC gaming, and machine intelligence.
AMD executive Raja Koduri said they developed the Vega architecture was built on the ability to solve data problems at an exabyte-scale, with the flexibility to address the extraordinary series of challenges GPUs will be dealing with in the near future.
Current GPUs can’t solve some major problems related to the processing of complex geometry and data sets too large to be treated in an efficient way. The new Vega architecture will open a new path of possibilities to improve complex tasks machines are not able to perform today.
AMD’s Vega is more than a gaming graphic card
The new GPU will be solving data problems in gaming at a gigabyte-scale and in at exabyte-scale in machine intelligence. The new GPU will also introduce new features, as it will have the most advanced GPU memory architecture.
Vega will come in the form of new high bandwidth cache, doubling this the bandwidth per pin. It is also optimized for streaming large data sets, and it can work with various memory types. It counts with up to 512 TB of virtual address space.
The new GPU also counts with a new generation of pipelines. That means the programmer can get incredible efficiency when processing complex geometry. It also comes with improved load balancing for consistent performance.
Another characteristic that makes Vega unique is the new Radeon pixel engine which works with a Draw Stream Binning Rasterizer. That will improve performance and power efficiency.
Source: AMD Vega