• Please review our updated Terms and Rules here

Operating System compatibility on early x86-32 chips

Geri

Geri

Experienced Member
Joined
Apr 24, 2019
Messages
147
6xesjhK.png


http://maker4d.uw.hu/osc_compatibility.png <--- Click for updated chart.

Currently i'm gathering information on OS compatibility with early 32 bit x86 chips.
I will update this table regularly. You can help me filling the holes if you have information.

Please note: Its 2020, and your mind could play trick with you, and maybe you misremembering something from 20 years ago!
Therefore, please only suggest new information if you can verify it, or if its from verifiable source.

Mod Edit: Added the link under the table as per OP's request.
 
Last edited by a moderator:
The last version of Windows to support "lesser" CPUs is Windows 2000.

Windows XP will not install on anything older than a Pentium because it requires CPUID and CMPXCHG8b. This rules out Cyrix's 5x86 and anything based on it. You can install Windows XP on a 486 system if you have a Pentium Overdrive, either the 63 or 83 MHz version. I've done it and it works, it's just painfully slow.

The Cyrix 6x86 core is very buggy and Microsoft officially dropped support for it in Windows XP.

I had an AMD GeodeGX 400 some years ago and Windows XP would not install because Windows thinks it's a 486 class CPU.
 
Thankyou for the informations, however in certain cases even if the os refuses to install, once if its installed, it works with the questioned cpu. Therefore i am currently canot red out the 6x86 and 5x86 and geode cpus based on that.
 
It seems i cant edit the original post any more, so here is the latest updated version
BoYaq8p.png
 
Geri said:
Thankyou for the informations, however in certain cases even if the os refuses to install, once if its installed, it works with the questioned cpu. Therefore i am currently canot red out the 6x86 and 5x86 and geode cpus based on that.
Click to expand...

The requirement in Windows XP is concrete, it's required by the kernel, HAL, boot loader, etc.

https://reactos.org/wiki/Supported_Hardware/CPU

They tested the 6x86 and they confirm CPUID and CMPxCHG8b are missing, which means that the 5x86 at minimum will also not work. I know a Geode will fail too because I've tested it.

Interestingly, it looks like later core revisions of the 6x86MX and MII added at least some support for those instructions, but it doesn't appear they work correctly.
 
FWIW, in the Debian 7 column: the last version of Debian that ran on original-vanilla 386 CPUs was Debian Sarge, circa 2005. Etch and later need a "486". (Their justification was a series of library and GCC issues that cropped up specifically on the 386, and of course the Linux kernel proper depreciated 386 support in 2012.) Whether they'll work on some of those mutant CPUs like 486SLCs, etc, is possibly an open question.
 
Win7 runs on a Pentium? Wow, that's news to me! I always assumed that Win7 uses i686 instructions because I never could get it to run or even install on two different P233-MMX systems w/ 512MB of RAM. However it worked just fine on a PII-400 (same amount of RAM).
 
Eudimorphodon said:
I know I've seen Windows 95 running, or at least booted, on a 16mhz 386sx, so it would surprise me immensely if any of these CPUs couldn't run it.
Click to expand...
I've done it too. A 386DX is claimed as minimum spec for Windows 95 but it will install and run on a 386sx. If you have a 386 with more than 4MB of RAM it does actually run at a usable speed. The problem comes with trying to add any software to it, as almost everything requires a 486.

Also, no stock version of Windows 95 will reliably boot into GUI mode on a K6-2 faster than 350Mhz - once in a blue moon it will load, but more often than not the system will lock up with 'WINDOWS PROTECTION ERROR'. Microsoft did release a patch to fix this, but rather bizarrely decided to package it in a Windows-based installer rather than a DOS one given that the problem is that the GUI will not load! Doesn't make it incompatible as once you manage to get the patch installed it will be fine, but probably worthy of a special note.

Incidentally - anyone know what the patch actually does to fix the problem? Once installed the system appears to pause for a few seconds at the end of the MS-DOS portion of the boot before the GUI appears. This delay is not there on systems without the patch. What is it doing during these few seconds? Did the delay itself fix the problem?
 
GiGaBiTe said:
...
Click to expand...
Well i have a computer to test it actually, just didnt had time for that for now. The old Cyrix 6x86 (the non-mx version) is gating down a lot of instructions in certain cases, such as it disables eve the CPUID instruction under certain cases. What reactos thinks and what windows thinks, is a very different question, but as i have a 6x86L to play around, i will try to find time and try what the XP installer says about it.

You can run Windows 98SE on a 386 by bypassing the system checks, it's just painfully slow.
Click to expand...
What do you mean by bypassing? You need a specific boot switch?

Eudimorphodon: Ok, then i pull out both of the Debians from the 386.

cwathen: interesting, i will add a side-note.
 
You didn't include NetBSD. I'd add it because NetBSD is supposed to support the widest variety of platforms.

As far as Debian 9 supporting the C7 cpus, that may well be correct; however you may well find that Debian 9 isn't supported on VIA C7 systems that also use the VIA VT8237 Southbridge chipset member. I believe support for that was dropped sometime around Lenny, so you may not be able to install anything later that Debian 4.0 on a C7 thin client. It's one of those things where you go through the trouble of installing Debian and then find that it won't boot. Very irritating.
 
Last edited:
Chuck(G): good idea, maybe later i will add netbsd, however i know nothing about that system, so its currently beyond my view.

Debian 9 runs fine with my C3, i have thrown out my C7 years ago due to inability to repair it so i cant test it. I have only added the green stamp due to the assumption it will work (if c3 works).

btw i forgot to add amd k5, fixed

QPG5fn6.png
 

Attachments

  • QPG5fn6.png
    QPG5fn6.png
    57.6 KB · Views: 1
Last edited:
One thing to keep in mind when talking about Via "C3s": the earlier Samuel/Ezra and later Nehemiah cores are significantly different, and an important distinction is the latter core supports the CMOV instruction. Support for that is what makes-or-breaks most Linux distributions that require an "i686 or later" CPU. If you tested Debian on a Nehemiah be aware it may well not work on a Samuel/Ezra. In fact, here's confirmation that it won't.

[1] The following processors, supported in jessie, are now unsupported:

* AMD K5, K6, K6-2 (aka K6 3D), K6-3
* DM&P/SiS Vortex86, Vortex86SX
* Cyrix III, MediaGX, MediaGXm
* IDT Winchip C6, Winchip 2
* Intel Pentium, Pentium with MMX
* Rise mP6
* VIA C3 'Samuel 2', C3 'Ezra'
Click to expand...

Your table shows green for Pentium and Pentium MMX, are you sure about that? (* Edit: Same question Re: K6-2 and a few others.)
 
Last edited:
Agreed about the CMOV thing--it stymied me on my K6; I didn't even bother to try with a P1 or P1MMX. May also be an issue for the Cyrix 686. It's easy to check--just boot the Debian installation disk--it'll tell you right off if the CPU support is lacking.

Of course, NetBSD will run on about anything; though I understand that 386 support has been dropped.

I guess there's no interest in OS/2 either.
 
I have an 800 Mhz C3 handy, which should make it either a Samuel II or Ezra, I guess I could probably perform the experiment. (Assuming its BIOS will boot off a USB key, it currently doesn't have any kind of removable media. And I'm way too lazy to set up a PXE server just for this.)
 
You must log in or register to reply here.
Back
Top