Coz the issue of IPC was raised in the Zen 5 thread.
Something you should consider in relation to what was raised in the Zen 5 thread: performance per clock of a certain architecture is affected by the different memory layers it has access to. You can see this in memory sensitive applications, where much faster DRAM or a much bigger cache can massively increase performance. One easy to pick example is gaming, where 3D v-cache AMD CPUs manage to outperform the vanilla SKUs while running lower clocks too. This is important because the cores themselves are the same, but they are fed with increased efficiency and work closer to their theoretical potential.
Another scenario where we can witness this type of memory PPC scaling is when we normalize for ISO clocks. If we take a core that is built to work at ~6Ghz and downclock it to ~3Ghz, the relative speed of the memory subsystem increases. CPU wastes less clock cycles waiting for memory to deliver the goods. In memory sensitive workloads a 50% drop in CPU clocks will lead to less than 50% drop in performance per clock. This can happen because of DRAM memory alone, but can also be impacted by bus/L3 clocks (depending on architecture).
In the Zen 5 thread you took the GB6 score as a performance per clock indicator and attempted to normalize for ISO clocks. AFAIK GB is memory sensitive, so this means that normalization is not as easy as applying the rule of the thirds.
You'll notice I used PPC instead of IPC. Performance per clock is much closer to describe what we really mean when we colloquially use "IPC" in this forum and in massmedia. This meaning is not really compatible with the pure engineering term (read this thread if you want ot know why). As consumers we talk about IPC as an average obtained from a battery of tests (or obtained from something like SPEC but then also reflected in a battery of workloads to see distribution of gains across the spectrum). The more different the architecures are in terms of clocks and scope, the less relevant these IPC comparisons are, to the point where it is much saner to just compare
performance, power and cost as the trifecta that really matters.