Intel aren't "that far" off moving to 10nm and their 10nm is basically the same spec as GF's 7nm - GF has a slight advantage to fin pitch and a couple of other areas but nothing that is going to make that much difference and it looks like Intel's process will still have the frequency advantage. GF doesn't seem to be a great option for performance parts really better suited to low voltage operation. AMD would have to move production to Samsung really I think if they wanted to push the MHz race.
It's worth noting that we really don't know what speeds Intel will get on 10nm, and GloFo's 7nm looks set to be considerably faster than their 14nm. They are stating up front that they expect 1GHz higher clock speeds on similar parts so it's not at all unreasonable to believe a Zen 2 based chip on 7nm can achieve 5Ghz where Zen 1 is achieving 4Ghz on 14nm.
I would also say that the slight advantage isn't that slight. IIRC Intel are saying 54 or 52 fin, 36 metal pitch? While Glofo is saying 48/44 fin and 36 metal pitch. I'm guessing the 48 is using only quad patterning(this stuff is getting crazy now) and 44nm is done via EUV at a later date for important layers only maybe.
Once again we seem to be seeing a situation where Intel couldn't launch their desktop because clock speeds are too low early on 10nm to be viable so they pushed desktop back and went mobile only. On 14nm what we really saw was first gen 14nm couldn't match the clock speeds of the parts from the previous gen, then 14nm+ brought clock speeds that 22nm chips were achieving semi comfortably. Delaying 10nm for desktop heavily implies that they are absolutely struggling to match current gen clock speeds which certainly gives a high chance that even the updated 10nm + won't be higher than current clock speeds, just finally reaching the same point.
If Intel gain little to no clock speed again while GLoFo jump from 4 to 5Ghz the massive majority of their current advantage disappears, Zen 2 would also be more likely to make a bigger IPC gain than Intel's next 'real' architecture update as opposed to just being a new stepping of Skylake. Basically their architecture is fairly old and extremely well optimised with not much low hanging fruit while Zen is likely to have some fairly obvious things to address.
GloFo 7nm isn't aimed at low power at all, it's a majorly IBM focused process who have been making 5.5Ghz chips for ages and their processes always kept that in mind.
I forget the name of the guy who spent the last 2 years saying AMD were going to fail because GloFo's 'low power' 14nm process was both incapable of making large chips and incapable of achieving 4Ghz because it was called low power. Fact is Zen is incredibly efficient comparing it to the 7700k/6900k/7900x. For a supposedly awful process that is only for low power, their chips are competitive and certainly don't look as if they are on a far inferior node.
If we get Zen2 and potentially 12 core chips in the mainstream, up to 24 in x399, up to 48 in server and starting 14-16months from now and these cores are capable of being clocked with turbo of 4.5Ghz and overclock to 5Ghz.... Intel will be outmatched. The two things their process has a noticeable advantage on currently, die area and clock speeds both seem to be disappearing for the next node and rather than a 2 years headstart on the node, the headstart looks like it will be non existent when it comes to desktop. That is a serious turn around in such a short space of time.
