anyone else not very optimistic about Vega?
HBM2 will help AMD a little but Nvidia is so far ahead of AMD in terms of power consumption and has released product around 6 months ahead of AMD's schedule
the one potential saving grace is for more games with Doom Vulkan-like performance, but only Doom and to a lesser extent Ashes of Singularity has shown very positive results under GCN so far
I really don't want to see a repeat of the Fury line being released and Nvidia overshadowing it with the 1080 Ti release, but Pascal power consumption doesn't make me very positive about AMD being able to get Vega to be more in line with Nvidia's high end performance & power #s.
I would be pleasantly surprised if AMD manages Fury/Fury X vs. 980Ti generation. That would be amazing for them. I fear much worse. Titan XP isn't even the full version and NV already has 3840 CC P6000 due to sell really soon. Such a card would double Titan X's performance. I don't see AMD doubling the performance of a Fury X as they barely improved perf/watt despite all the P10 hype.
To me this is shaping up to be AMD's HD2000/3000 series. AMD needs to increase performance 80-85% from 480 just to match the 1080! Titan XP is almost 2.4X faster than RX 480 at 4K. AMD would need to increase all the resources of RX 480 by at least 2.5X since hardware doesn't scale 100% linearly.
AMD's best shot is for DX12/Vulkan games to start coming out. In DX12/Vulkan, Fury X competes well with a 1070. As long as most AAA games are DX11, Vega will look bad on the review charts. I think NV played into AMD's hand because they priced everything at artificially inflated prices. Even if AMD can't take the performance crown, they should be able to undercut 1070/1080 and beat them on price/performance. This should be the minimum expected given how much later Vega will come out.
If AMD manages to actually have a 4000 SP Vega 10 and a 5500-6000 Vega 11, and the cards are fabbed at TSMC, then we are talking. If their flagship Vega is made at GloFo and has only 4096 SPs, NV should win this generation without much effort.
As I have been doing for almost a decade, I am going to take the path to least resistance/headache for my upgrade path: keep building mining rigs and use the profits to upgrade the main gaming rig when I need the extra performance. I would feel a lot better getting free Vega CF even if it's 25-30% slower than the $1700-1800 USD 1080Ti SLI. In 2018, there will be even faster cards.
I am not happy with the prices of any GPU this generation, but I still want to reflect why for 95% of PC gamers buying Vega or any other flagship makes no sense. $179-209 RX 470, $199-279 RX 480, $249-299 GTX 1060 are all beating November 2013 $699 780Ti:
https://www.computerbase.de/2016-08/radeon-rx-470-test/3/
As I have been saying for years, unless the PC gamer can perfectly time the upgrade path, it's better to buy $300-400 cards and upgrade every 2 years than buy a $650-800 card and keep it for 4. Since AMD doesn't have the resources of NV, it's becoming more expensive to focus on $550-650 flagship die that very few customers purchase throughout the generation. AMD needs a card that's faster and cheaper than the 1070 and then they'll gain market share. Even if Vega beats 1080 and costs $649, it's not going to sell well. AMD doesn't have the same customers as NV. NV released Titan XP and 1080 owners paid $1200 for ~30% more performance. When AMD released Fury X that was about that much faster than the 290X and cost $649, most AMD buyers skipped it entirely. AMD should go back to HD 7000 days were they beat NV on performance and price/performance in every single tier. Vega will do little to gain back AMD much market share because many high end buyers will only keep buying NV or already bought Pascal.