I get the impression from your post you design and manufacturer bleeding edge GPU's from your garden shed then?
I get the impression from your post that you lack the understanding of the subject matter at hand and as such decided to mock what you didn't understand rather than ask questions and get answers.
Tahiti/GCN1.0 is a design that was locked out and done architecturally some 3 years ago. We have had several GCN 1.1 designs since then, without checking I would guess the 7790 is at least 18 months old. But you believe AMD will spend what will likely be in the region of 3-4million on the masks and maybe 20-30million on 12-18 months of work to tape out a "new" GCN1.0 design with two features added?
You think you would take the basic building blocks of GCN1.1 with true audio which you have, and randomly add two of the features to the old architecture rather than merely use the ALREADY FINISHED work of using the GCN 1.1 architecture. It's pure nonsense, it would take MORE work to get trueaudio and other features added to an old design, than use the work done on GCN 1.1 in which trueaudio is already a feature.
The idea of adding a couple of GCN 1.1 architecture features to the old features is quite literally absurd, it's beyond the realms of possible. You would literally be talking about peeing millions of dollars down the drain to add features to an older less good architecture when the newer architecture is already done and has those features.
There is zero chance of Tonga having features from newer architectures yet being based on Tahiti... none at all, literally 0%, not 0.3%, not 5%, there is no chance.
The worst case is that Tonga is a GCN 1.1 architecture product with superior power efficiency and performance, a chip that is between the 7700/290x(both GCN 1.1) and aimed at roughly 280 performance levels.
Ideally the new chips would be GCN 2.0 but it all depends where they've done what work. They've been working on GCN2.0 chips for 20-14nm designs and the processes are delayed so they are releasing something new at 28nm. Is it cheaper for them to port their 20nm gcn2.0 design to 28nm, or just take the building blocks and work they have with GCN 1.1 on the 28nm process and build up a midrange and maybe new high end chip towards the end of the year.
It depends when work on Tonga started, it might have been started 3-6 months before GCN 2.0 was ready and the parts coming maybe 3-6 months after Tonga are GCN 2.0 based.