The 5800x is a 105 watt part(!!!), still gets starched by the M1 in single core, and is only about 40% faster in multicore despite using all that power to drive double the performance core count.
But by all means, go off to cherry pick more niche benchmarks...
I did not cherry pick any benchmarks, I have picked the only 2 benchmarks that have been published for both Apple M1 and for Zen 3, and that were also freely available, so I could also run them on my own computer, to verify that the published benchmark results are correct.
They are not niche benchmarks. Geekbench 5 covers many common application domains and its only flaw is that the GB5 tests are very short so they finish before the CPU overheats and slows down, so this benchmark is much more favorable for Apple M1 than for a desktop CPU.
Even if GB5 is advantageous for M1, it still shows that Zen 3 is faster.
gmpbench tests a fundamental feature of any CPU, its ability to compute operations with large numbers. That may not matter for you but it is definitely one of the important criteria for measuring the single-thread speed of a processor, so when Apple M1 scores 6422 and a Ryzen 7 5800X (not the fastest Ryzen) scores 7816, there is no doubt about which starches which.
Moreover, the multi-thread results that I have seen for Geekbench 5, which very strongly favors Apple M1 (because in a real application M1 would overheat soon and it would slow down) show a 60% higher speed for 5800X, not of 40%, e.g. 12339 for 5800X vs. 7716 for Apple M1.
If you want to live with the illusion that Apple M1 is faster in single-thread execution than Zen 3, even if all the real numbers from benchmark results contradict you, that is your privilege.
The 3990X is the most powerful desktop part on the market. It gets merely a 1.4x speedup over the 5950X despite having four times the core count. (Granted, the 5950X core is somewhat better).
Geekbench isn't terribly useful at the best of times. It's particularly bad when comparing across architectures and operating systems. And, as should be clear, its multithreaded test does not scale linearly with core count. You should never take the ratio between two scores to mean the ratio between performance, especially when comparing multithreaded scores with different core counts.
The 5800x is a 105 watt part(!!!), still gets starched by the M1 in single core, and is only about 40% faster in multicore despite using all that power to drive double the performance core count.
But by all means, go off to cherry pick more niche benchmarks...