I have an old Windows 7 laptop and the newest versions of the Chromium browser (Supermium fork, for legacy Windows compatibility) run far faster than any versions of Firefox or Internet Explorer ever did.
Adding features makes it slower and is inevitable pain of progress. Suggestion we should stop improving things is ridiculous.
I doubt website loads faster. Statistics show that modern websites load slower on modern hardware than old websites used to their respective hardware. I don't see why it would be different on Windows 7.
It depends on the software usage. If you're not using cpu-demanding tasks like rendering videos in Adobe Premiere, Blender 3D, etc, then very old pcs will continue to work fine.
The desktop computer I'm typing this comment on is a 10-year old Intel i7-5820K 3.3GHz pc. Back in 2014, I maxed it out at 64 GB RAM but I took half out and reduced it to 32GB RAM. I use it daily for VS2022, VSCode, VMware, MS Excel.
I also help maintain a desktop for my 80-year old friend. Her computer is a 15-year old i7-950 3.06GHz. That computer from 2009 runs Windows 10 and she uses it daily for Chrome browsing, Youtube videos (including 4k), Amazon shopping, and Mozilla Thunderbird email.
It's possible that Windows 11 with its TPM requirement may finally force a hardware upgrade of those dinosaurs but I read there are hacks to get around that.
I could definitely see how buying a new high-end pc today will last ~15 more years for typical consumers. On the other hand, the power users who want to run the latest LLM locally with 600-watt graphics cards that will be obsolete in a year will be a different story. Today's NVIDIA 5090 with 32GB RAM may be too small to run the next latest & greatest LLMs for those who want to stay on the bleeding edge.
EDIT REPLY: >Why did you take half the RAM out?
It was a long story that I left out. The motherboard was unstable with all 64GB of RAM in it. It would lock up with RAM corruption after a few hours. Finding the root cause of this this took several days of trial & error with swapping the 8 RAM sticks and running MEMCHECK on multi-hour scans. After testing and going the process of elimination, it turns out that none of the RAM sticks had defects. The defect was the motherboard itself. Take any of the 4 out of 8 RAM sticks so it's 32GB RAM and everything is super stable.
I was just mentioning the 32 GB RAM without all that backstory to emphasize that I've gone 10 years without being at the more "future-proof" 64 GB.
This is what killed a lot of computers in my company's laboratories.
I mean, depends on what software you use. There's a pretty sizeable and growing ecosystem of people who put a lot of thought into performance. Just look at tooling like ripgrep, some of the newer terminals people have been working on, recently I came across a pretty nice neovim plugin where someone had written their own custom SIMD fuzzy string matcher (https://github.com/saghen/frizbee). There's some pretty admirable effort people put into performance these days.
I think speed of your setup is mostly limited by how willing you are to look for better alternatives.
Having said that my 8GB MacBook Air runs the unit tests for my current project four times faster than my 2018 i7 Mac. I will upgrade within a couple of years.
That is grinding to a halt. Chips are making only modest performance gains with each new fabrication bode, and the time between nodes is stretching to 3 years. Not only that but it looks like GAA FETs at 1-2 nm (marketing name) is close to the end of the road.
Software is going to have to stop getting more bloated, and may have to become more efficient as people want to run it on smaller devices.