• Please review our updated Terms and Rules here

Looking for an older "non-PAE" Linux Antix version 14.4 (suitable for Pentium 1 systems)

P

Peter z80.eu

Experienced Member
Joined
Apr 21, 2016
Messages
88
Hi,
I am desperately looking for an older Linux Antix version 14.4, filename is MX-14.4-non-pae.iso ( originally downloadable from https://downloads.sourceforge.net/MX-14.4-non-pae.iso - but not existing anymore ).
The current SF page for older versions https://sourceforge.net/projects/antix-linux/files/old/ does *NOT* contain any "non-PAE" version, especially not the MX-14.4-non-pae.iso , not sure why.
Does anybody knows where to get it (even a working Magnet/Torrent link would be fine) ?

Thanks in advance for any hint regarding this old, last non-PAE-Version of Antix
Peter
 
Someone recommended this version for my Pentium system, he told me it "must be a non-PAE" version.
If the *_386.iso files with the higher version number would be exactly the same as "non-PAE", this would help a lot - but I do not know if this is true or not.
As far as I remember compiler options will exist for generating "pure" 386 code, for Pentium code, but also for 686 (Pentium Pro and higher) code ...
 
ldkraemer said:
Try Distrowatch - MX Linux
Click to expand...

Unfortunately, Distrowatch is linking back to the sourceforge page with the missing files.

Peter z80.eu said:
Someone recommended this version for my Pentium system, he told me it "must be a non-PAE" version.
If the *_386.iso files with the higher version number would be exactly the same as "non-PAE", this would help a lot
Click to expand...

It looks to me like the maintainers switched to new labels to indicate which .iso is which with version 15. Versions 14 and lower used -non-pae and -pae, while versions 15 and up use _386 & _x64 to mean the same things: -non-pae would be 32bit (386 and up) and -pae would be 64bit (_x64, 686 and up.)

It looks that way here:
https://sourceforge.net/projects/antix-linux/files/old/antiX-16/

and here:
https://sourceforge.net/projects/mx-linux/files/Final/
 
acgs said:
ersions 15 and up use _386 & _x64 to mean the same things: -non-pae would be 32bit (386 and up) and -pae would be 64bit (_x64, 686 and up.)
Click to expand...

Wrong.

PAE was an extension to the x86 architecture that was introduced with the Pentium Pro to allow 32 bit x86 CPUs to address up to 64 GB of memory in 4 GB pages. The original Pentium doesn't support PAE at all, and is limited to a 4 GB address space.

x64 means that the build requires an x86_64 CPU, such as the Athlon 64, or Pentium 4 with EM64T (Intel's implementation of AMD64.)

Back when these CPUs were still relevant, Linux distributions generally labeled themselves as i386-i586 for Pentium and lower CPUs, and i686 for Pentium Pro, Pentium II and later CPUs. Though as the 2000s came around and these CPUs became less relevant, i386 started to become the generic designation for "we run on 32 bit CPUs as long as they're not too old."
 
GiGaBiTe said:
Back when these CPUs were still relevant, Linux distributions generally labeled themselves as i386-i586 for Pentium and lower CPUs, and i686 for Pentium Pro, Pentium II and later CPUs. Though as the 2000s came around and these CPUs became less relevant, i386 started to become the generic designation for "we run on 32 bit CPUs as long as they're not too old."
Click to expand...

Thanks for the clarification of my post. I should have been more precise on a public forum; I wasn't trying to explain the details of Intel/AMD architecture, only the way the antiX/MX Linux .iso designations were changed. I didn't mean to imply PAE was only for 64bit. Also, this relabeling of .iso files took place in 2015, so I guess you could say that the change was high time in coming.

Posted on the antiX Linux Blog on October 28, 2022:

For those that like to live on the bleeding edge and build upwards from a small base, antiX has made available iso files based on Debian sid. Users are expected to be proficient.

We offer the following completely systemd-free and elogind-free flavours for both 32 and 64 bit architecture. User can download sysVinit or runit versions.

antiX-core (c500MB) – no X, but should support most wireless (libdbus-1-3 is installed).

antiX-net (c200MB)- no X. Just enough to get you connected (wired) and ready to build (no libdbus-1-3).

The 32 bit version uses a non-pae kernel.​

This would appear to verify that the labels were changed from "-non-pae" to "_386" and "-pae" to "_x64" for "simplicity." Based on this, I think that the OP should be able to use any of the "_386" versions on his Pentium 1.

Interestingly, I found another blog post from 2018 that talks about the “spectre” and “meltdown” security problems and suggests fixes for 13.x, 16.x, and 17.x versions only. That begs the question: why no fixes suggested for 14.x and 15.x?
 
I would consider a distro provider that disappears their older versions is possibly concerning. I would suggest you use Slackware with the non-PAE kernel. Slackware 14.2 is of the same vintage, and it is still being supported with patches. But Slackware 15.0 may also work. Just note that you may need to drop any Firefox after version 48 or find a suitable alternative.

Also, if you are running on a i586 class processor, I would need evidence that spectre or meltdown flaws are even possible... but if you don't need a browser, or not running any servers, you could possibly use pretty much any version of linux you want.
 
Testing with QEMU and the "-cpu" option, I was NOT able to run any _386 version of antiX with "-cpu pentium", but the older MX 14.3 non PAE version runs...
 
the3dfxdude said:
Also, if you are running on a i586 class processor, I would need evidence that spectre or meltdown flaws are even possible... but if you don't need a browser, or not running any servers, you could possibly use pretty much any version of linux you want.
Click to expand...

The Pentium uses speculative execution, branch prediction and other performance enhancements, it is most certainly vulnerable to spectre, meltdown and a host of other vulnerabilities.
 
"Debian i386 architecture now requires a 686-class processor" :cautious:

If it doesn't run on a 386, don't call it i386.
 
Peter z80.eu said:
Enlightment brings the following announcement: https://lists.debian.org/debian-devel-announce/2016/05/msg00001.html
Click to expand...

Good find! Thanks for uncovering the details I've been looking for.

... that means all MX Linux/antiX 386 versions newer than version 15.01 (or in other words, beginning from 16.0) will not run with these first, old Pentium processors.
Click to expand...

So, why can't 15.01 run under "QEMU -cpu pentium" then? It may be a flaw in QEMU's emulation (unlikely), a configuration problem where "Pentium" doesn't mean "Pentium 1" exactly, or a mislabeled 15.01 version? Would it be worth a try running on a real Pentium 1?

Plasma said:
"Debian i386 architecture now requires a 686-class processor" :cautious:

If it doesn't run on a 386, don't call it i386.
Click to expand...

100% agreement on that! Go back to PAE/non-PAE and start saying what is meant.
 
GiGaBiTe said:
The Pentium uses speculative execution, branch prediction and other performance enhancements, it is most certainly vulnerable to spectre, meltdown and a host of other vulnerabilities.
Click to expand...
Yes I know they have part of what is needed for the exploits. I asked for evidence the pentium is exploitable. These vulnerabilities have a specific sequence of events and each processor is a bit different. Exploitation depends heavily on threaded environments and task switching. The 1st version of atom processor is not vulnerable based on the originally published reports, and the atom was designed based on the i586 execution pipeline (while still being i686+). I have not checked the atom on the more recent exploit findings. I really expect these CPU vulnerabilities in practicality to be p4+ when hyperthreading was introduced. I would consider the p2 and p3 a maybe, if hyperthreading is not required. The pentium based on available information doesn't seem likely at all.
 
You must log in or register to reply here.
Back
Top