What you write about is pretty interesting, but Iris Pro and Haswell-E are at the highest end of the market (where volume is really low).
In contrast, even most people in this enthusiast forum only use a Core i5 (of some type) that they may have been holding onto for quite some time because they haven't found the need any extra CPU performance.
You must concede, though, that the i7-4970k and the i7-5820k are basically the "gamer's choice" chips from the Intel stable right now. By 2015/16, the 5820k will probably have established itself as the chip-to-get for anyone who is serious about high-end gaming (except for the occasional oddball who wants the 5960x, but really now . . . ). Those are the chips people look to when they describe the state of modern PC gaming. If you want to cause an uproar, launch something that beats the day's halo chip and then step back and watch the madness. If you can!
Intel most certainly can beat their own product, and my guess is that, somewhere along the line, DX12 titles will start to show up big on Broadwell Iris Pro or some later Skylake product.
As much as GPGPU will play a bigger role in the future, neither AMD nor Intel sell a CPU today to offer a benefit years down the road
I think Kaveri could have done it had the software stack been in place. Mantle really hasn't made the 7850k shine over, say, the 9590 yet. But if you look at the raw computation capability of the 7850k, you must recognize that its plucky little iGPU can
totally smoke a 9590 in 32-bit floating point, 32 bit integer, and 64-bit integer. 64-bit fp is a bit iffy on Kaveri. On my 7700k, my Steamroller cores were putting in significantly higher 64-bit fp throughput @ 4.7 ghz when compared to the 384 shaders clocked at 1028 mhz. I imagine the situation would be a bit better on the 7850k, assuming memory bandwidth didn't mess with that particular bench too much. Hmm, maybe I should test that one of these days . . .
Regardless, show me a high-end DX12 title that can heavily tax single-card and SLI/Xfire configurations just with 3d and load up CPUs with tons of physics calculations and whatever else, and I'll show you something where that iGPU will really help. A lot. What it might not do is correct certain stutter situations perhaps related to the way graphics drivers work.
It will be rather curious if they launch an Excavator die by the time they should have their next wonderchip, designed by the hands of an EE God, rolling out on the market.
Might be a WSA thing, but here's the other thing we need to remember: integrating a particular x86 structure into an APU requires a non-trivial amount of work. AMD is short on R&D resources right now. Launching both a server-level Zen part and a cut-down Zen APU at the same time might be more than they can handle. Remember that doing so would require moving their GCN cores to the 14nm Samsung/GF process as well. Will they even have their flagship dGPUs on 14nm by that time?
It will be cheaper and easier for AMD to modify Carrizo to make Bristol Ridge than it will be for them to make an APU out of Zen. Zen APUs come later.