AMD let it be seen that AMD Vega will have an update, possibly during the first half of 2018, to compete with the NVIDIA Volta, something that is difficult when Vega 2 will have a 14nm + lithography.
We have been talking about AMD Vega for weeks and it is that AMD graphics cards are highly anticipated, but we have not yet seen a gaming version, as we have seen the Vega Frontier Edition, intended for the professional sector. It should be noted that Vega does not come to compete directly with Volta, but has been developed to compete with Pascal. Volta's competition will arrive next year and it will be Vega 2, graphics cards developed in 14nm +, an update of the lithography on which the AMD Vega that are about to arrive are based.
This revision of AMD Vega would arrive next year, even before what we might expect and it is that apparently AMD has already finished developing the current Vega and they will only dedicate themselves to production and polishing drivers, for next year you will launch Vega 2 with 14nm +. It seems to be something similar to what happened with Polaris, first releasing a full range and then a revision, with a slightly modified development and higher frequencies and little else. What we know is that AMD is already working hard on the new graphics cards, which if we look at the release date of the RX 500 Series, they should arrive six months after the first Vega, possibly during Q2 2018.
It should be noted that AMD will take a long time to provide data on AMD Vega 2, since they have not yet launched the RX Vega and they cannot anticipate anything. These new GPUs should have an improvement in terms of performance per watt, in addition to an increase in frequency, although in Polaris the improvement has been minimal and the increase in consumption has been quite high. Something that we find interesting to know is what AMD will do with the memories. The HBM2 right now are the best option, because the GDDR5X in power are below, but we know that SK Hynix in its portfolio already has the GDDR6 and NVIDIA will surely opt for these memories for the more powerful Voltas.
Perhaps the million dollar question is whether the next generation Vega GPU can really compete with the NVIDIA Volta. We have seen that Polaris has not been able to with Pascal even by chance, since the RX 480 performed below the GTX 1060 or on par with luck and the RX 580 performed on a par with the GTX 1060 and in some cases, slightly above . NVIDIA had the high-end market to itself, with the GTX 1070, GTX 1080 and Titan Pascal, but that was not enough and it has released the GTX 1080 Ti and Titan Xp, not counting the update to the GTX 1060 and the GTX 1080, which have seen GDDR5 and GDDR5X memory updated, respectively.

"AMD had the high-end market to itself"
Apparently it is in the DNA of this website to defenestrate AMD, they still do not have a proof for it, only guesses are too much and they poison.
When I defend Intel, I don't see you coming en masse to complain, but when I say AMD, you come in like a headless chicken.
I thought it was a mistake, but I think it was with another intention now that you say
It was a failure, it was not with another intention, I promise.
This time it was a mistake. Nor do we crucify them for such a mistake.
It was late and tired and I was school. Sorry, I've fixed it.
How unprofessional you are, if I were your boss you would be fired, it is not the first
If I imagined it was a mistake, nothing serious, I only said it to be corrected.
If you want to do something with Vega 2 ... it will have to be something better than the simple change of node as with Polaris, in addition to containing consumption, TDP and increasing frequencies ... I think that a change is also necessary, even if it is minor in the architecture.