FreeBSD Continues Push Toward Deprecating 32-bit Platforms

[energyman]

those who are complaining: you can always volunteer to do the work to support 32bit. Noone is stopping you. If FreeBSD itself is not willing to accept your work, they won't stop you forking and doing your own thing. Isn't it beautiful?

But demanding that volunteers increase their workload, just for you and a handful of others like you - that is just selfish and wrong.

[phoenix_rizzen]

the last 32-bit-only intel was netburst series which was end of sale in 2010.
even if the hardware still alive, i dont think it was still used as daily primary computer by 2020 and moreover installed with latest ubuntu

Incorrect. NetBurst, aka Pentium 4, started as 32-bit only but ended its run supporting EM64T aka Intel's not-quite-compatible-with-amd64 take on 64-bit x86. The 64-bit version was released in 2004 and the P4 was discontinued in 2008.

The follow-up to the P4, Pentium-M, was 32-bit only. The follow-up to that, the original Core architecture, aka Core Solo and Core Duo, were also 32-bit only. These were released in 2006.

The Core 2 architecture that followed was the first designed to support 64-bit from the get-go, also released in 2006.

The Intel Atom N200-series was released in 2008 that had some models that were 32-bit only, and others were 64-bit.

[Jumbotron]

Meh...way of the world and natural evolution. Why not the same for silicon and hardware arch?

Sooner or later at the rate that ALL kernels, Windows, MacOS, Linux, BSD's are expanding to be capable of handling the increased computational need at scale and performance you have to cut the cruft. Arborists the world over know this as the practice of pruning,.

As an example, this just came out over at The Register where users ( peasants ) on the Microsoft plantation will soon wake up to this if their kit is of a certain, shall we say, antique vintage....

https://www.theregister.com/2024/02/13/microsoft_might_have_just_pulled/

[AmericanLocomotive]

Who are FreeBSD's primary users? How many of them are still using 32 bit processors? Of those, how many care about running the latest versions of FreeBSD?

[slacka]

I'm all for progress, but without X32 as a viable option, people here are missing a big benefit of 32-bit. For work, we had hundreds of 32-bit micro VM's. I investigated the tradeoffs of upgrading to 64-bit. At the time the 64-bit VM's required at least 30% more memory. The benefit of lower power costs from improved CPU efficiency did not outweigh the cost of purchasing more RAM. X32 not only reduces apps memory footprint by 10-30%, but consistently performs better than x86-64, in some cases as much as 40%. People here cheering for the death of 32-bit are cheering for larger, bloated apps. 64-bit pointers just waste memory unless your app requires more than 4GB of RAM.

Also, Intel introduced the (Z6xx) of Atom processors 2011, so people here claiming 32-bit x86 CPU haven't' been sold since 2010 are wrong.

Finally, our software team compiles for multiple architecture targets that we don't necessary support. This helps uncover latent undefined behavior bugs and improved portability for future platform support. For BSD, they may be too developer constrained for this the trade-off to make sense. But for the Linux kernel, and compliers like gcc and clang, supporting these platforms is a win for code quality. On the distro level, probably best to reserve 32-bit for niche distros as long as the kernel and compliers continue to support them.

[Weasel]

crazy and stupid

Your post? Yeah.

[SoongVilda]

It's funny to see how everyone dropping 32-bit support, I'm not surprised. Why would you want to have 32-bit CPU? It's useless, slow, limited RAM, hard to hold any usable load. I have Core Duo Q9400 overclocked from 2,7GHz to 3,3GHz and it's still so slow, hard to use just for web browsing - I already compiled kernel and browser to this CPU and situation is still same, okay... load of CPU dropped from 95% to 88%, but it's still crap, just what the heck want to use 32-bit CPU in the 2024?!

32-bit packages, kernels and so on must be a huge work to maintain. That maintenance eat resources, man power for more important things, which were actually used in the real word.

Stop everyone burning time on 32-bit bullshits, it's history, no one want to brings back, no one to use and margin error in the population should sense that and upgrade their systems. I believe some crazy people running important things on 32-bit machines, which is pretty dangerous, crazy and stupid and we shouldn't support that behavior. Otherwise they can fork it and maintain 32-bit version things itself.

[chocolate]

Haiku is looking really good on x86, especially now that it's got an Emacs port, but its USB support is unfortunately hit and miss, so the live environment for installation doesn't boot on some PCs (e.g. I recently tested on a Sony Vaio P 1st gen. which would make for a nifty "focus writing" machine every now and then). Support of some kind for x86 will never really die; major GNU/Linux systems and FreeBSD waving goodbye is understandable, even though losing some cool machines is a pity.

[stormcrow]

Every distro should make one last 32bit LTS release and then just drop it. In 6 years or so when it becomes EOL there is certainly no need to still use such processors in production. And hobbyist projects can then support them as long as the kernel supports it.

FreeBSD isn't a Linux distro. It's BSD which has a direct lineage from the early days of Unix. They aren't related nor are Linux distros and FreeBSD developed and packaged in the same way. Your comment is irrelevant to an article on FreeBSD. Also, the last line is tone deaf, there's a hell of a lot of 32 bit processors still out there, still being produced, still in use, and still being designed (ex the enormously popular R Pi Zero, most PtP wireless gear uses ARM32, RISC-V is both 32 and 64 bit, etc). However, embedded systems isn't really FreeBSD's wheelhouse like it is Linux & NetBSD so all those low powered devices aren't really relevant to FreeBSD's future.

FreeBSD as a cooperative is reasonably ascertaining their resources, degree of developer interest, the goals of the project, and the dwindling hardware availability for hardware they support and need for 32 bit releases of their operating system and are giving time for people to make plans if they still use FreeBSD on 32 bit hardware. However, FreeBSD has been a server oriented operating system for a very long time. How many 32 bit servers have you seen in the past 5 years that are actively receiving upgrades regardless of the installed OS? The last one under my care, an old PowerEdge tower from the P4 era, was sent to the recyclers a few years ago. Good riddance, it was like burning a 95 Watt incandescent light bulb 24/7.

[varikonniemi]

Yeah, so drop it even sooner, or now?

I got annoyed when in the distro i use, manjaro, someone saw that 32 bit support on the install media takes a large amount of space, and no installed package needs it. So why ship it?

It's time to stop 32bit burden like multilib by default. Of course offer it as installable package.