The Linux Kernel May Finally Phase Out Intel i486 CPU Support

[rmoog]

You know what CPUs are coverage by i586 and i686, right?

That's a rhetorical question.
i586 is a brief glimpse in the x86 family of architectures as it started and ended with the OG Intel Pentium.
i686 started with Pentium Pro and it remains to this day a common target for binary x86 distribution. Its end is marked with introduction of SSE2.
How can I help?

I have 4gb of swap on my lenovo n22 and its been used like that since it was new, i still use the machine lots to this day, and it uses swap quite often. I have a few kingston 128gb SSDnow from 2011 in the same situation. flash storage is no where near as fragile as most people make it out to be. swapping on flash is perfectly fine

Keeping a Portage workspace on flash however is still pretty bad and out of respect for flash memory, because it does such a great job, I try to keep that entirely in RAM.

[mangeek]

Isn't MC 68000 still supported?

I was running kernel 5.15 on an emulated 68040 last week! More to just check it out and do some other emulator stuff, but... yes.

I'd say the same thing about that as I do about the 486, though. I'd be fine if newer kernels dropped support for ancient platforms. I don't mind running 'museum kernels on museum hardware', and the places where 486 and 68k CPUs might still be in use aren't use cases where new kernels are needed anymore.

[kneekoo]

I have 4gb of swap on my lenovo n22 and its been used like that since it was new, i still use the machine lots to this day, and it uses swap quite often. I have a few kingston 128gb SSDnow from 2011 in the same situation. flash storage is no where near as fragile as most people make it out to be. swapping on flash is perfectly fine

This is strictly a matter of write cycles the flash memory allows before they become "read-only". If you're lucky, you get storage with good memory cells. If you're double-lucky, the storage device manufacturer also over-provisions the storage with some extra GB that are hidden from the user/OS, and are put into use as some of the cells wear out. Not every device comes with that kind of luck, though.

But whatever the case, it only makes a difference in how much sooner or later they reach their write limit and become unusable. Swapping constantly on flash memory brings that moment closer faster. Also, people's use cases are very different. What's "normal usage" for me or yourself is nothing like others use their hardware. So combining that with the flash memory type, some people will have a good experience, others not.

[Svyatko]

Swapping on flash memory is a terrible idea, especially on cheap hardware, which is the target of those RAM-limited laptops. And again, the point I was making there is that not everyone knows how to shop for hardware, and some people actually have little money to spend on computers. And even today, people can buy this kind of hardware, brand new.

Here's one example that costs $200: https://www.amazon.com/10-1Inch-Grap...dp/B086LB6FVW/

It's a cheap and fully functional PC, you can do a lot of things with it, but the RAM limitation is pretty bad.
CPU: Intel Atom x5-Z8350 quad core 64-bit, Secure Boot, AES-NI and VT-x (you can even run retro VMs)
RAM: 2GB maximum
Storage: 32GB eMMC

Apart from the write cycle limit that flash memory has, eMMC is also not great because the more you use it, the slower it becomes, so swapping both shortens its lifetime and performance. And running a desktop OS and browsing the web requires some RAM. The Raspberry Pi OS would be a great choice for that kind of a PC, maybe even the 64-bit version, but if you have to run multiple programs and you want to make sure you don't easily end up swapping, 32-bit software takes the RAM usage down a few notches and you can do almost every kind of task you need, especially for such a low-priced hardware.

With Linux we have zram, zcache, zswap, MGLRU, and le9 patch https://www.phoronix.com/news/le9-Linux-Low-RAM.
User can place swap to additional device - microSD card or something attached to USB.

[Svyatko]

Talks about which x86 CPU introduced speculative execution:

For x86 only, not for all others: https://en.wikipedia.org/wiki/Superscalar_processor

https://en.wikipedia.org/wiki/Pentium_(original)

Sorry guys, I was wrong.

Pentium Pro - November 1, 1995 https://en.wikipedia.org/wiki/P6_(microarchitecture) | https://en.wikipedia.org/wiki/Pentium_Pro

First x86 CPU with modern architecture: NexGen Nx586 CPU - March 1994

https://halfhill.com/byte/1994-6_cover-nexgen.html

The Nx586 is a fascinating design. It demonstrates that the 80x86 chip's microarchitecture hardly needs to resemble Intel's offerings. On a range of common benchmarks, it performed comparably to the Pentium. It might even presage some ideas that Intel has planned for the P6 and P7.
...
Nx586 Microarchitecture. The Nx586 combines a CISC front end with a RISC-like processing core. It employs dynamic branch prediction and speculative execution to keep its deep pipelines busy.

https://cpumuseum.jimdofree.com/museum/nexgen/nx586

The Nx586 CPU introduced in 1994...
...
Introduction 03/1994
...
AMD purchased NexGen when AMD's K5 chip failed to meet performance and sales expectations.

https://en.wikipedia.org/wiki/NexGen

The company was best known for the unique implementation of the x86 architecture in its processors. NexGen's CPUs were designed very differently from other processors based on the x86 instruction set at the time: the processor would translate code designed to run on the traditionally CISC-based x86 architecture to run on the chip's internal RISC architecture.^[3] The architecture was used in later AMD chips such as the K6, and to an extent most x86 processors today implement a "hybrid" architecture similar to those used in NexGen's processors.

It went public in 1994, and was bought by AMD in 1995 for $850M. The technology forms the platform architecture for all of AMD’s current microprocessors.

...

AMD purchased NexGen when AMD's K5 chip failed to meet performance and sales expectations. Some NexGen customers were even given free AMD K5 CPUs with motherboards in exchange for sending in their NexGen hardware.

Development of AMD's internal K5 successor was halted in favor of continuing from NexGen's Nx686 designs, eventually becoming K6.

[kneekoo]

With Linux we have zram, zcache, zswap, MGLRU, and le9 patch https://www.phoronix.com/news/le9-Linux-Low-RAM.
User can place swap to additional device - microSD card or something attached to USB.

Yup, zram is great and easy to install by a newbie. But in the case of USB, 3+ is recommended for simultaneous read/write operations. USB 2 is only good to spare the flash storage in the mobile device, but otherwise it's underwhelming. :P

[Developer12]

Minor nit - I believe the original Pentium is superscalar but does not use speculative execution.

Pentium Pro was the first x86 with speculative execution IIRC. (fixed per Svyatko's comment below)

hmm. revising things, it may be possible that the AMD K5 is the first. it predated the pentium pro.

[oiaohm]

With Linux we have zram, zcache, zswap, MGLRU, and le9 patch https://www.phoronix.com/news/le9-Linux-Low-RAM.
User can place swap to additional device - microSD card or something attached to USB.

MMCe and MicroSD you run into write cycle issues. MMCe is about 10000 cycles per cell and your responsibility priced MicroSD cards are also only about 10000 cycles. This 10000 cycle limit is good enough for lots of uses but swap on a low memory computer is not one of them. You can in fact run into trouble with MicroSD and Linux distribution logging eating though the write cycles.

M.2 and other proper SSD solutions that wear leveling controllers with dram cache this gets you out to 100000 cycles+. This gets in the range of reasonable. Do note the dram cache that some of the swap operations don't end up using flash memory write cycles because they go into the cache and get disposed of before being written out. This is a feature missing from the MMCe cards and MicroSD in fact all SD cards. This is a space/size limit that there is simple not the space for the dram.

http://www.supertalent.com/products/stt_usb_detail.php?type=USB%203.0%20Express%20Dram%20Disk

The annoying part is we don't have competitive market for the above product type. Yes 64Gb of ram in a 3.1 USB drive form factor. This is a ideal swap device for expanding system without enough ram other wise. No write cycle limit issue. Yes the device is ram when you power off system it goes blank as well. Yes they are a true pain to find on sale anywhere.

The reality for swap storage you are looking for dram in some form to be used either SSD dram cache or a dram device so it has decent life span.

Random flash device into swap storage is really playing bad game of dice.

[Clive McCarthy]

From what I read the 486 was discontinued in 2007. It seems it should have happened before then ? In that case ALL cpus that were discontinued before 2007 should also be dropped.

I understand the fun of keeping old hardware running but there is no need for an up to date Kernel. Hardware is cheap by comparison with maintaining support for archaic cpus and gpus. If you want to run old hardware run old software on it.

[oiaohm]

From what I read the 486 was discontinued in 2007. It seems it should have happened before then ? In that case ALL cpus that were discontinued before 2007 should also be dropped.

https://www.datasheetarchive.com/whats_new/ebe36b70ff1e4e36b37c2a628ef63d18.html

That not the correct date. Last know 486 clone production date that I can find is 2013. Yes its the ZFx86 not the intel 486.

Yes so new systems for industrial usage made though to 2015 could have the ZFx86 inside them with only a ISA 486 instruction set

486 ISA as a CPU platform has not been out as production as much as the intel end of 486 production would suggest.

Yes the 486 ISA as a platform is only next year coming up to a decade past the last known production date.

If you applied your logical to the real last date of the 486 ISA only chip everything 10 years should be removed.