AMD at CES 2017

 

 

 

 

This represents an entirely new and most advanced GPU Memory architecture in the world. Vega GPUs featuring with a high –bandwidth cache and a controller, along with HBM2 (high bandwidth memory2) technology that allows increases transfer rates up to 2GT/s (Giga-transfers per second), and as high as 8 gigabytes of RAM package. Also this will take half of physical space needed on a board compared to an equivalent amount of GDDR5 memory.

 

 

 

Vega architecture is most effective device for transmitting largest datasets and can work with variety of memories up to 512 terabytes of virtual address space.

Vega also has better load balancing with an intelligent workload distributor and delivers consistent performance

 

AMD’s RX480 fix

 

 

 

The next generation geometry pipeline has higher levels of efficiencies on processing complex geometry and transmits more data per –clock than previous Radeon architectures.   In the same time the advanced pixel engine designed to improve performance and power efficiency, as AMD describes, “early culling of pixels invisible in a final scene.”  

 

AMD have also shown doom running at 4K using the flagship Vega graphics card.

According to a report, running at the game’s Ultra graphics settings the frame rate is shown at around 70fps with a few dips below 40fps here and there.

Vega’s pixel engine is a client of the L2 cache, and makes reduction in overhead for graphics workloads which make use of heavy read after write operations.

This high end GPU architecture comes in providing a serious GTX1080 Ti contender.

 

 

 

AMD have confirmed that Vega graphics cards will be released in first half of 2017 where they have already shown a working Vega 10GPU at the CES show. There are rumors of Vega 20 GPU that AMD is working on it, with Global Foundaries.  

 

According to a report there will be a Vega 11 GPU too, it is possible that AMD wants Polaris 12GPUs be shore up the bottom end of the new 500-series range that we won’t see a Vega 11 chip until later on. The Vega 20 is reported that may be a 7nm GPU releasing in 2018

  

 

 

      HBM2Product Configurations

 

 

  

8Gb based

 

9mKGSD

 

5mKGSD

 

3mLGSD

 

Density/Cube

(GB)

 

8GB

4GB

2GB

 

10

 

1024

 

1024

 

1024

 

Speed/pin (Gbps)

1.0

1.6

2.0

1.0

1.6

2.0

1.0

1.6

2.0

Bandwidth (GB/s)

128

204

256

128

204

256

128

204

256

Usage

 

HPC, Server

 

HPC, Server, Graphics, Network

 

Graphics, Cache

 

Config. / System

 

8/6/4 Cube

 

4/2/1 Cube

 

2/1 Cube