a) CPU cores:
Even Intel is being limited by current architecture. x64 was not big boom as anybody ever expected (well lets say that we can handle much more memory natively).
What intel is doing last few years is adding more memory channels to its CPUs. its a nice way how to improve CPU performance... and sell much more memory chips
AMD did not followed this step.
b) Compute cores
They are here for years now, but the apps able to utilize their potential are mostly in gaming business. Both Intel and AMD have processors with ondie integrated graphics. But clearly, in rough performance AMD graphic cores/compute cores are far better as anybody can expect.
program desiged to bruteforce passwords able to run via CPU and GPU. CPU can generate up to 500 passwords per second. GPU can generate up to 10500 passwords. (These are numbers I measured on my own PC, sorry for no screenshot).
Using HW potential which is now being used only for entertainment would improve most of currently used software.
So yes I agree that computing/graphic cores can be used as a measurement for theoretical performance of APUs.
Basically no current software (except games) is able to utilize such HW potential. Most software even dont bother to support other than x86/x64 architectures. I believe its a matter of fact that AMD has processors (with computing cores) with higher computing power than Intel, but this potential is not used due different architecture.
Is it possible to write OS which would use GPU core instead of CPU? How complicated it would be?