By the time D3D12 becomes mainstream, Intel would already release the 9xxx series, with 6 cores replacing current 4 cores x770 for high mainstream CPU. AMD also has to deliver a ZEN 2 CPU to be competitive with that too.
By the time Intel releases i7 97xx (cannonlake?) we'd still be getting ~5% IPC boosts & that's me being optimistic. Coffeelake (i7 87xx) will be just a tick & probably yield lesser IPC gains than Broadwell did wrt Haswell, not to forget Intel will still be wasting ~50% of the die on IGP.
So basically one would be wasting ~50% of his precious chip on a
useless IGP, that's what I've heard gamers say, & still be getting less than 10% IPC over & above Skylake. Also DX12 will be adopted much faster, because of win10 & how quickly we've moved on from PS4/XBO to a Polaris based GPU powering the next gen of consoles.
P.S. Just to add to that, why do you think Intel never released a 57xxK for desktops? They simply couldn't get Broadwell to clock high enough for it to be a compelling upgrade option over the 4790K, that they made it look like a dud actually helped Skylake is another matter entirely.
Now with Kabylake being named as 7xxxK they're doing this
mindshare game, just by renaming the chip they're making it look better than what it is i.e. just a respin of Skylake. But here's the catch, with 97xx they'll have to clock it at or above 4GHz & deliver ~15% IPC gains to make it a more compelling purchase than Kabylake, or get it to clock ~4.5GHz & deliver less than 10% IPC.
There's also this little step they'll have to take with 10nm shrink, we all know how 14nm turned out, & how that pans out remains to be seen. I'm pretty certain however that even if they fail to match Kabylake's base clock people will obviously point towards
more cores, you know the opposite of what they're saying today
If I'd be Intel I would certainly be sweating right now because Zen being this good actually throws a spanner into their tried & tested formula of milking the consumers with no competition at bay.