The Linux Kernel May Finally Phase Out Intel i486 CPU Support

[pWe00Iri3e7Z9lHOX2Qx]

A brief history lesson for you. The Pentium P5 wasn't released until March 1993, therefore the i486 was the literal state-of-the-art desktop CPU from intel 30 years ago, in 1992. How could it "stop being relevant", as you claim, when it was Intel's latest top of the line product? Your claim makes no sense. And as others have pointed out, it saw production use well into this century in various embedded and industrial systems. Linux is much more than an eye-candy desktop distro. It is used in a multitude of places you've likely never even heard of.

And just because intel stopped manufacturing i486 in 2007, doesn't mean folks stopped using them. In fact, the embedded and industrial markets typically run very long hardware refresh intervals. So its safe to assume industry was running i486 chips for many years after the official 2007 product eol. It is only now, here in 2022, where we can begin to have the discussion about removing support from the kernel.

Kids these days and their narrow distorted worldview, sheesh.

I think the important question isn't whether or not such systems exist in the wild, because they certainly do. The important bit is are these systems actually being updated? Right now 5.10 has the farthest projected EOL. Are vendors actually validating their solutions against recent LTS kernels? Or do they just live with whatever they shipped with for decades?

[wagaf]

486 computers are still being used and need to be maintained in the upstream Linux kernel so that these systems benefit from the latest optimizations and security fixes. There is no need to remove support for these systems. I never understood this destructive desire to break things that work.

Are they really being updated and profiting from latest kernel features? Most likely not. This kind of industrial system usually runs the same software from the beginning, in best case scenario with minor updates if strictly required. Even in the very uncommon case of major software updates, it's unlikely that the new software would require features from the very last kernel.

The only use case might be security fixes, in that case the last compatible LTS kernel would do a good job.

[jeisom]

So, if they drop an x86 based cpu every decade, I guess we can look to 2032 for them to drop the original Pentiums.

[Nille]

Such old System that are still used today are unlikely used with modern software. And there are still a bunch of long term kernels available on kernel.org that still support it. The 5.10 Series is a LTS kernel that gets still Updates the next 5 years and even after that date is the kernel still working well.

[Vistaus]

Oh noez! The Phoronix community won't like this as everyone still uses 486's.

[vsteel]

The industrial realm where almost all of these would be used don't change software unless they have to. Those cases fall into the if it isn't broke don't fix it. Doesn't matter what OS it is, they never change. The other part is most of the time the industrial applications have custom drivers to support special hardware. No one is going to spend time to update those drivers, or the people who knew how it all worked have long since retired.

Then we are left with the nostalgic people, most of the time they are happy running on old OS versions because it takes them back to when it was all new. Reliving childhood.

We are then left with the .0001% who are trying to daily drive these types of machines. They have to use old hardware because of the ports and standards supported are for the most part not around anymore, that greatly limits what you can do with them. Not to mention almost no developer has any hardware to even see if anything is still working. The people with this old hardware are not picking up the torch and saying they will take on support.

Security updates, you mean the ones for the hyper-threading, multi-core systems, and things like Wi-Fi, which your 486 doesn't have.

At some point, if you are going to keep running old hardware, you are going to have to run an old OS and realize the world doesn't owe you new software on decades old hardware. If they were simple to support and aligned with current hardware, the fact that it is old wouldn't be a problem. It is the fact that these older types of hardware are a headache to support.

[skeevy420]

The reasoning makes a lot of sense since most 32-bit distributions require newer CPU models anyways. Why support something that nothing modern even uses? And if your business really, really needs it you probably already hire a Gentoo, etc user that can build your environment for you.

I agree with the premise of accepting that sometimes old, museum, hardware just needs to stick to old, museum, software.

I've also thought that if there was such a great need for legacy hardware then the Linux Foundation or some other similar group would have created a Linux Legacy LTS forks of the kernel where goal would be to preserve hardware that gets dropped from mainline Linux and provide security fixes when applicable.

[zamroni111]

I never understood why people insist on supporting ancient hardware that stopped being relevant 3 decades ago.... I mean, what is going to happen if its gets removed? It is STILL GOING TO BE ABLE TO RUN 6.1 kernels forever..... I mean, why would you need to run the 6.2 kernel on a 486 and you are not satisfied with 6.1? Perhaps you need better support for the new Intel/AMD gpus with raytracing to acompany your 486 with 4mb ram? This is ridiculous, really. Removing support from future versions of software does not mean your ancient hardware can't run existing versions... This should have been such a huge non-issue if people were less childish about it....

Even i585 and other hardware with more than 10 years of end of sale should be removed too.

[nanonyme]

Even i585 and other hardware with more than 10 years of end of sale should be removed too.

In practical terms i686 hardware is already gone too. We're these days typically talking of SSE2-capable CPU's being run as 32bit.

[jrch2k8]

I would say nuke everything before SSE2 class cpus, since most of them simply get no benefit from such a new kernel and im not even sure latest glibc work properly on those anymore, let alone a desktop enviroment