But system performance improvement is real. There are ongoing enormous advances of compute and I/O power which allow new and improved media handling, more complex UI and workloads, and overall freer tradeoffs between memory capacity and time required to complete work.
In other words, the system is always getting faster, but you're not, so a more powerful system allows more work to be done within your attention window.
So upon this realization, you might become curious as to what the relationship between system performance and your perceptions of speed. Without explaining in any detail, you might first suspect that the relationship is non-intuitive, much in the way that speed of transportation and perception of distance traveled are not intuitive. A walk to the corner market by every measure is shorter than a flight across country, yet your perception of scale does not lead to the direct apprehension that the flight is 500 times faster than your stride. Same thing with computer power
Another unintuitive aspect of performance is that speed relationships within computing systems are subject to a time vs. distance tradeoff with interactions generally slowing with distance. If your sense of performance comes from interacting with web sites, while the parts of your PC go really fast, the stuff you're accessing is arbitrarily far away, and subject the sum of all the delays over that distance, only some of which are affected by the speed of local processing. There's a famous law, similar to Moore's law, coined by Cray architect Gene Amdahl, that observes that no matter how many things a system can do at once, the time to complete a job are limited by the time required to those things that must occur in a sequence. For your PC the wifi connection is a tube through which all the stuff you want to access is ordered sequentially. Even if your PC and the computers you access were infinitely fast, it still takes time to get the data through that tube. So you might suspect that your experience accessing the web is generally tube limited, not PC limited.
Gone are the days of a two year later CPU bringing 3x or 5x the performance to a new system.
Any latency you experience comes by the many layers beyond your devices’ capacities.
I believe the hardware is pretty flashy and we’re better erring on economizing power consumption than generative predictive web preloading.
Fast-forward to today, and Apple is arguably in the same boat as Intel. They have mediocre year-over-year density improvements (only a 1.5x-2x density improvement most years) and can't radically change their IPC with ISA updates. They pay hand-over-fist for TSMC's best nodes but lose out on profit margins to Nvidia's datacenter products. Adding insult to injury, Apple arguably has the worst modern desktop GPU architecture, which compounds the acceleration issues ARM has with SIMD and vector workloads. Apple Silicon is backed into a corner, from a design perspective. Ye olde "I'm a RISC machine that wishes I had CISC architecture" problem.
Still a great chip for browsing the web or editing video. But again, people cautioned back at the M1's launch that ARM is hardly a silver bullet. We still have x86 laptops because for >90% of PC use-cases, ARM isn't worth the licensing cost.
x86 licensing costs are complicated. The basic old 16-bit architecture is very well-documented, and while it's proprietary it's still implemented in hundreds of products and clones. Modern 32-bit and 64-bit implementations have a lot more ISA extensions that would be costly to license and hard to develop on your own. As a result, modern x86 designs can really only be made by licensing Intel or AMD technology.