I'm not speaking of highly customized cores neither of MSFT eing done by MSF. I've clearly stated that if they are go through the pain of getting "cat" cores on the 14/16nm they might as well improve them. I also clearly stated the rumored amount of the agreement between MSFT and AMD (pretty old figures could have been debunked), I suspect Sony paid its a fair share, hence my comment. Reuse of functional units is common, ARM, Intel, AMD everybody does it, for example we know that Jaguar use Llano divider.
Zen is a great but 8 physical cores is a big amount of silicon and they've yet to launch, putting an APU altogether for fall 2017 sounds like a impossible target. Hence I start to entertain the idea of a follow-up to Puma cores, I've always been against that idea because AMD needs such cores and if they were working on them it would make sense to inform investors as they are not exactly going through easy times. A possibility could be for the console manufacturers MSFT or SONY (a real OR not XOR) to be the main financial contributors to the project, so the all venture would be for now under the radar (NDA, the detail of the accountability of the project is hidden behind the aforementioned agreements, etc.).
Jaguar and Puma are outdated porting them untouched to 14/16nm would be one hell of a miss-investment though we've seen others.
I don't think reworking Zen is an option or a good idea.
Well you can run the calculation, it just depend on the number of physical cores and the clock speed.
As for PR anything strictly greater the 5.5 should be rounded to 6, I don't expect such a gap but I don't expect 6,000 GFLOPS or 6,000,000 MFLOPS. I do expect stupid reactions all over the web though to whatever the final precise figure is.