Wow, let's not let our desires "bulldoze" everyone into thinking AMD can come out with a 28nm part that's going to just "crush" the competition, or honestly even do they want too? My thinking it's a tough "rabbit from the hat trick" as it's certain they're stuck on the same process. Then the flip side is the 300 Series might not be in the channel all that long... do they want to show everything they might have on the aging 28nm?
Truth is GCN hasn't shown the same luck in efficiency so far as Maxwell, so while they could show an improvement to be at least in-line at the "ticket booth" of the "ballpark", I don’t see them besting on perf/w. If Tonga was the best they could muster 5mo’s ago, I don't see even if they have HBM implemantation this first round it being any saving grace.
Given AMD's response time here... Let me first say, I think AMD had been working alternate paths', but they finally found them not taking them forward enough or cost effectively so they spent some time shifting plans. Was it about HBM or process related... don't know, I do think they juggled several concepts with limited resources. Trailing some 4mo's AMD does need to better the GM204. Especially given that Nvidia went "all in" on efficiency, and basically only provided "One-upsmanship" to over their (own) predecessors. AMD had better get over that low bar, if they can't take advantage of this, it tells somebody really screwed the pouch on engineering resources the last 3 or more years.
I like a lot of others here say, it would be a very bad to "have to" employ an AIO cooler in the "first round" for the reference 380X. I feel it would it be smart play to after several weeks to then offer a AIO Rad-Radon™ version that's OC to the hilt as an off-shoot product offering.
Given the issues/resources AMD has surely had to endure the last 3 years against a competitor that is firing-on-all-cylinders, just to be "in the hunt" at this point is all I can assume as "good enough" headway for now.