Updated: Aug 10, 2022
For the past few years, the hype around AMD "taking down" Intel has been very real. To this day, I have no idea how even some of the most respected enthusiasts in the industry are buying into it. Great marketing? Sure. Buying off reviewers? Probably. Snowballing bandwagon mentality? Gotta be. But there are very reputable sources such as LinusTechTips out there, heaping praise onto AMD the past few generations that surpass the sentiment that they're "finally becoming relevant again" but into absurd stratospheres that claim AMD "crushes" Intel.
I'm by no means an Intel or AMD fanboy- personally, I believe that for the past 2 decades now, CPU's and their real position in terms of overall PC performance and capabilities, has become less and less relevant in reality. Architecturally sure, they (traditionally) handle all the computing tasks between each peripheral and component in a computer that delivers an end user or application experience. Again though I say 'traditionally'.
Since the advent of General Purpose GPU (GPGPU) compute implementations (such as Nvidia's CUDA or Opensource OpenCL), we've gotten to offload compute-intensive applications from the CPU to the GPU and realized massive, massive increases in efficiency in performance across the board. Sure, a CPU itself is still comprised of more robust architecture that allows for far more complex instructions and tasks to be handled per execution core that, when compared 1:1 against a compute core in a GPU, has far more power and potential. However, even low end GPUs contain hundreds to THOUSANDS of compute units to perform against each single CPU core.
Plus, when you think about it, a GPU itself has a directly-connected, purpose-built, self-contained entire on-board figurative 'northbridge' of a traditional computer, whereas a CPU requires a compatible motherboard, RAM and underlying chipset, and then general purpose routes to the compute units through standardized interfaces and protocols like PCIe, USB, SATA, etc (which is something a GPU doesn't have to deal with, but is extremely necessary for general computing- so that is one win/need for CPUs).
AMD is always catching up with Intel and Nvidia standardizations and innovations in the PC world (for consoles, they actually get very innovative consistently- so point for AMD there). They seem to prefer stacking more and more simplified, less-capable components onto boards to beat them out with raw numbers to deliver their horsepower. For example, AMD's 'threadripper' and bulldozer architectures both marketed themselves on having more cores per CPU than Intel. However, check out their TDP numbers, check out their die sizes, check out their rollouts of technologies that admittedly do sometimes show up before Nvidia or Intel get there (such as fabrication process shrinks or implementations like PCIe 4.0 this generation, DirectX versions, HBM memory, etc) but also fall behind with refining and making sure their implementations are adoptable and compatible with the rest of the industry.