The size difference is not significant. This is about the maintenance burden. When you need to change some of the code where CPU architecture specific things happen you always have to consider what to do with the code path or the compiler flags that concern 486 CPUs.

Here is the announcement by the maintainer Ingo Molnar where he lists some of the things he can now remove and stop worrying about: https://lore.kernel.org/lkml/20250425084216.3913608-1-mingo@kernel.org/

The Linux kernel is well over 30 million lines of code (lots of that is drivers).

This change shrinks the kernel by about 15,000 lines. That is not nothing, but it hardly moves the needle.

It is just one less thing to have to worry about and one less constraint to limit flexibility in the future.

It’s probably less about making the kernel smaller and more about security and reviewing code. The less code you have to maintain, the fewer vulnerabilities even if it’s old code.

I would doubt almost 20 year-old code is taking up a lot of space or presenting new vulnerabilities. And it’s obviously open source so if anyone needs it, they can always use an older kernel or maintain it. Sometimes, your oldest code is insane. I wish there was a budget for every company and government to pay retirees part time to go back over their oldest code that’s still in use. A lot of retired programmers would do it for fun and nostalgia. And to be horrified something they wrote 20 years ago hasn’t been updated or replaced.

Probably not a lot of space savings, but certainly a reduction in complexity, which helps programmers keep everything together and frees their time to work on the newer stuff