Yeah I don't think they have much choice on how large they can go till 7nm EUV is ready for prime time starting next year, I guess it's also questionable whether it's worth releasing a big die version of 1st gen RDNA at all given there are some obvious features missing against competitors that 2nd gen RDNA will close up, that will become far more relevant with titles in Q1 2020, though 2nd gen RDNA would also be reliant on the node improvements from EUV from the looks of it(And its node would have been set in stone a long time ago with the design rule changes for EUV).