Large L3 of desktop Ryzens is there to compensate for extra memory latency from chiplet design.I suspect the lower amount of L3 cache will only make a marginal difference in games / other software.
Monolithic APUs have lower memory latency and hence cache miss is somewhat less serious penalty for performance.
Though those workloads whose most of the time needed data fits to larger L3, but not this halved L3 are going to see clear difference.
It's all about memory usage pattern.