Because everything’s a trade-off, people optimize different systems for different things, if you have a real-time operating system that runs a power plant, it doesn’t matter if the interface is clunky as long as it hits its time targets for its tasks.

If you’re running a data center server, you’re probably worried more about total throughput over time, rather than immediate responsiveness to a terminal.

For a computer that does lots of machine learning and vector math, you might spend a massive amount of time making certain programs run a few percentage points faster by changing how memory is managed, cache allocation across CPUs, network access, you’re going to find your critical path and bottleneck of performance and optimize that.

When we’re talking about a general use desktop computer, we tend to focus on anything that a human would interact with minimize that loop. But because people could do anything, this becomes difficult to do perfectly in all scenarios. Just ask anybody who’s run Chrome for a while, without restarting, and has a thousand tabs open, because all the RAM is being consumed the computer starts to feel slow, because virtual memory management becomes more demanding…

TLDR, all of the operating systems are capable of being very performant, all of the kernels are really good, it’s all the extra stuff that people run at the same time that makes them feel different.