Big.little has easily won this race. Apple/Arm does this well,
Again, not a clear cut winner. The big.LITTLE approach usees less silicon area than P-cores with SMT2, but - at least in theory with everything else being eqal (i.e. not comparing different processors with different ISAs made on different generations of silicon) - the fat P-cores have higher per thread performance over a larger number of threads, and still have very competitive throughput when the secondary SMT threads are added in later.
Apple silicon is looking good mostly because Apple is always exclusively on TSMCs best node; the gap has narrowed significantly when Ryzen mobile 7000/8000 jumped to TSMC 4. (Then there are all the different design trade-offs / targets, where AMD and Intel would rather want to melt their CPUs when Apple would rather ditch active cooling.)
It's just like the good ole' times when Intel won by default by having the best transistors. Now Apple has played that game for a few years, but that won't last forever. Intel can afford TSMC's best, too, and has swallowed their pride. Even AMD is finally rich enough to buy TSMC's best. But we customers might not actually be rich enough, at least not in large masses, so AMD will use those cutting edge wafers for Epyc and AI accelerators. Without Apple's reality distortion field, there might not be enough customers willing to overpay twice the price for half the RAM and half the mass storage.
Lunar Lake will be interesting in that it is a serious attempt to aim an 'x86 design to a very different product target. I am not sure if Microsoft can do their part on the software side to really challenge Apple, but we'll see.
