I've actually been pleasantly surprised by how well most Linux stacks work in resource constrained environments.
I have a first-generation Framework - the ones with the shitty Intel CPUs that don't support proper S3 sleep, only the awful "modern standby" - and I often throttle the clock to 400 MHz to save battery. GNOME's performance doesn't degrade at all; the only place it feels "un-snappy" is when starting heavier apps.
I think I may be misremembering, but I got S3 working quite well in both Linux and Windows. This comment from Nirav Patel suggests it should work quite well (although unsupported by Intel as with all other modern Intel CPUs): https://news.ycombinator.com/item?id=31435132
Right now, I have enough small but annoying hardware issues with my Framework laptop that I'm not bothering to improve anything for myself, either software or hardware related. I currently use it only as a desktop, and don't have anything but Windows installed at the moment.
In the past I've always recommended Framework, and I still feel I can't just leave my negative tone above sitting without further comment.
<Start of rant>
All the issues I've had would be trivial fixes, except that I'm not in a country that Framework ships to, and they can't send to anywhere outside that list, or allow freight-forwarding. Even for simple, inexpensive parts. I don't know the reason for this, but I do genuinely trust that it is because of a real problem for them, not just a lack of effort.
Excepting that, I've always had very high-quality, prompt responses. I'm currently waiting until Framework expand their supported countries, I move country again, or I can arrange to send parts to someone else I know who can forward parts on to me.
From soon after purchase, I was silently enduring a rattling fan, and the 1st gen 1TB expansion card overheating issue. Recently, the 1st-gen backup battery design issue has suddenly made itself known to me, and now I'm more pessimistic. It seems all it takes is moving to a non-supported country to turn what was designed to be a repairable laptop, into something less convenient to fix than my previous preferred choice (any of the more repairable, common enterprise laptops).
For anyone living in a supported country, I would still suggest a Framework laptop, but with slightly more hesitation than before. <End of rant>
Linux desktops certainly handle low-resource environments well, particularly compared to Windows, but even so it has some omnipresent elevated latency relative to late 90s/early 00s commercial operating systems, which can be felt even on powerful hardware.
It’s hardly a dealbreaker and not even really a problem (which is probably why it’s still there) but making software instantaneous does wonders for improving how it feels to use.
Slow CPUs and GPUs work much better on Linux than on Windows, but when the problem is a shortage of RAM, Windows will hold out much longer before grinding to a halt.
In an era of soldered-on RAM, this is becoming a rather annoying problem. Unfortunately, Linux doesn't offer the same APIs Windows does to take preventative action before running out of RAM. Windows' auto-growing page file also makes for a less crashy experience than Linux' static swap partitions (or constant-size pagefile). Plus, Windows comes with memory compression (zram/zswap) out of the box, configured to use both memory compression and disk swap to catch these situations.
I wish Windows wasn't such a slog on the CPU and GPU because making Linux work in low-memory situations is an absolute pain.
> making software instantaneous does wonders for improving how it feels to use.
When you phrase it like that, I'm actually more surprised Linux suffers from the extra latency. When most software is written to scratch an itch and optimization beyond "works on my machine" is for fun, you'd expect latency to evaporate over time
It’s beyond my realm of expertise so I can’t speak confidently, but my hunch is that many of the responsiveness papercuts are cross-domain in nature, which makes them more difficult to track down and fix (not to mention makes who’s responsible for fixing them more murky).
Another big chunk likely comes down to the tradeoffs all the big DEs have made in favor of making development easier or improving DX. This is understandable but at the same time it would be nice if at least one took a hardline stance towards commiting performance reductions and placed responsiveness as a chief concern, much as the operating systems of yesteryear did.