Moving into a single 8-core CCX design will certainly allow more granular reduction in core counts, but to be honest TSMC's yields are allegedly so good I don't actually see the point, unless it's a move to up core counts across the range.
Ryzen 5: 8 cores
Ryzen 7: 10 cores
Ryzen 9: 12 and 16 cores, maybe even 14 too.
The trick with that is the Ryzen 5 requires a fully-working chiplet, but it doesn't have to be the best clocker. Or you could do 4+4 for a non-X and 8+0 for the X, the latter benefitting from having no cross-CCD latency, but then you run the risk of the Ryzen 5X actually outperforming the Ryzen 7 because that will have a latency penalty the 5X doesn't have.