Moving Coffe Lake desktop chips to 2017 is probably a response, but this is not a new architechture.
I'm not so sure of the "no new architecture" part, consider this:
2015 → Skylake was released on time despite 14nm issues, µarch was ready anyway
2016 → Cannonlake... but 10nm is delayed, no tick that year yet µarch should be already fully designed
2017 → Icelake... 10nm still delayed, coming last quarter at best, same applies here µarch must be ready
So... 14nm++ coffelake is pretty much a stopgap before 10nm yield is good enough for larger dies, we know it's gonna be 6 cores and not much larger than Kaby (150mm^2 with IGP too), but we also know that there's no point in releasing Skylake 6 cores again when that's already happening in august (Skylake-e), thus my only explanation is that Coffe lake might incorporate some µarch improvements.
Maybe they'll port Icelake cores onto 14nm and call that Coffelake, maybe only Cannonlake cores, I don't know for sure but that's what I would do if I was at Intel.
It should also explain the 15% increase cited here:
Now Kabylake didn't achieve exactly 15% single thread improvement on all chips but it's close: stock 7700K runs 4.4GHz all core turbo while 6700K does all core at 4.0GHz → 10%
Extra clock/OC headroom alone grants better single thread here, but I don't expect 14nm++ to do as much... the slide clearly doesn't talk multi core performance so for Coffelkae to get >15% increase they must be combining say ~5% extra clocks and ~10% ish IPC gains.
6 cores, higher IPC than Zen (10-20%) and higher clocks should make for a compelling product if they price it at around 300$... oh you get the useless IGP too!
Realistically speaking: 8 cores is great and all but I'd rather stick with 6 faster ones in the future, the same way 4 faster cores were and are still suggested over 6 now.
6/4 = 50% more cores, 8/6 it's just 33%: multithread shouldn't be much different in this case but you get significantly higher single core speeds for common tasks.