Well, so far, bandwidth limitations have prevented AMD's APUs from being compelling purchase motivators. However, that could change with HBM, and Intel really needs to step up their game here. Seems like they are getting into the "next chip syndrome". First it was 14nm and Broadwell Gen 8 that was supposed to be a game changer. Now it is Skylake and gen 9 with edram. I just dont understand how a company with so much technical expertise can not narrow the gap with AMD. I mean they have, but AMD *could* make a big step forward with HBM, and intel seems incapable or unwilling to do it, except in an expensive limited edram product. (Edit: seems to me with all the time, effort, and process advantage, Intel should have caught up to AMD in mainstream igpu performance and been in the lead with the edram versions.)
Right now, it is not a major problem, because their graphics is "good enough", while neither Intel or AMD is powerful enough (igp) to be a compelling solution. But if Zen brings HBM and a shrink to 14nm, the competition will suddenly ramp up quickly. And I know it is a very artificial test, but the starswarm demo on DX12 showed absolutely terrible results even with the edram GT3e, possibly indicating some other serious bottleneck besides bandwidth.
And yes, I agree, the gaming benchmarks on that NUC test they just put out were quite disappointing. Seems the reviewer was very generous in assessing gaming performance. Right now, as I said, that is not a problem, because AMD does not have a capable gaming platform in that power envelope either, but with DX12, HBM and 14nm (if they are even close to being on time) that could chage.