Not to much opinion can be made till we can get some specifications and/or benchmarks.
It also depend on if the software written for heterogeneous architecture, if it is at the operating system level or if it can be taken advantage of by current programs. If it is adopted it may just be the boost AMD need on the CPU side to become competative again.
It also depends on if they make a hexa-core model.
It also depends on the IGPU. If they can get performance closer to the r7 260x then it is an extremely compelling SFF chip.
At this point it all boils down to speculation until we get some more information. Kaveri was far less impressive than it should have been on the enthusiast side. Richland was really the sweet spot. If they can get a jump similar to the jump from trinity-richland as apposed to the richland-kaveri they may really have something. It would also help a lot if it is based on 20nm.