AMD is set to showcase its highly anticipated GPU architecture, Vega, on 5 January 2017, in conjunction with the upcoming CES 2017 event. Interestingly, the company has also made a website which display a countdown timer to the preview event. In addition to that, there’s also a teaser video that contains no clear indication of what Vega is capable of.
Unfortunately, it’s quite difficult to pinpoint what exactly will be said about AMD’s Vega GPU architecture during the event. One has to remember, however, that this is merely a ‘preview’ event, so we wouldn’t expect the company to be announcing new Radeon graphics card on that day – but let’s hope they do.
Fortunately, folks at 3DCenter forums have managed to discover a hidden ‘word cloud’ inside AMD’s special Vega website. The word cloud apparently details the improvements that Vega will bring when it is released, but it isn’t clear whether the upgrades are based on a comparison between Polaris or Fiji.
So, according to the word cloud, AMD’s upcoming Vega GPUs will come with two times more bandwidth per pin, high bandwidth cache, significantly increased power efficiency, a next generation compute and pixel engine, an increase in peak throughput per clock, and plenty more. One rather interesting improvement that was mentioned was “8x capacity per stack” which strongly indicates that Vega-based graphics cards (probably not all) will be using the HBM2 memory controller.
Alas, we can only be sure of all these details on the 5th of January during AMD’s Vega preview event. Speaking of which, the event will start at 10pm local time.