Skip to content

What Amix Is

Amix is Commodore's port of AT&T UNIX System V Release 4 (SVR4) to the Motorola 68030 Amiga. ✅ The kernel platform string is m68k-cbm-sysv4. It is a monolithic SVR4 system with no AmigaOS compatibility layer ✅: it drives the Amiga's native chipset through its own kernel drivers (the bitplane console, and the /dev/screen screen manager the stock mono X server draws on), never through AmigaOS libraries, and otherwise treats the machine as a generic 68030 Unix workstation ✅. If you know SVR4 on any other 68k or 3B2 machine, you already know most of Amix.

This page is the one-screen orientation. For the deeper mechanics, follow the cross-links to hardware, the boot process, the kernel architecture, and the full version reference. For end-user history and install media, the authoritative community resource is amigaunix.com.

Lineage: ported from the AT&T 3B2 codebase

Amix was a direct port of AT&T's 3B2 (WE32x00) SVR4 codebase, not an evolution of a pre-existing 68k Unix. 🟡 (amigaunix.com states this but hedges — "it appears that … 3B2 codebase"; corroborated by EAB/datagubbe, but no primary citation.) Commodore chose the 3B2 source for licensing-cost reasons rather than starting from a native m68k tree, which is why several defaults (for example the System V s5 filesystem default — see filesystems and disks) look like 3B2 inheritances 🟡 (exact rationale unconfirmed). The port was led by Michael Ditto, "Unix Systems Software Architect" at Commodore from 1988 to 1991. ✅ Contemporaries describe the effort as "quick and dirty." 🟡 (community-reported)

The opening line of Ditto's own driver paper calls Amix "a direct port of the AT&T Unix System V operating system … essentially identical to … System V Release 4." ✅ That paper — Writing Amix Device Drivers, presented at the 1990 European Amiga Developer's Conference — is the single most authoritative primary source on how the system is built, and it grounds the driver model.

🟡 Sun Microsystems twice explored OEM-selling the A3000UX as an entry-level workstation; both deals fell through.

What Amix is — and what it is not

Amix is:

  • ✅ A monolithic SVR4 Unix for the 68020/68030 Amiga, identified as m68k-cbm-sysv4.
  • ✅ A standards-era SVR4: STREAMS networking, TLI plus BSD sockets, POSIX.1, init / /etc/inittab / run levels, virtual consoles on Alt+F1..F8, and SVR4 packaging (pkgadd / pkgmk / pkgtrans). See kernel architecture.
  • ✅ A real workstation OS with X11, TCP/IP, and NFS — see networking and X11 and the desktop.

Amix is not:

  • ✅ Not AmigaOS-compatible. There is no Amiga binary compatibility, no AmigaDOS, no Workbench. Dual-boot between Amix and AmigaOS is handled at the ROM level (the "Superkickstart" bootstrap), not by any in-OS shim — see the boot process.
  • ✅ Not a user of AmigaOS's chipset libraries. Amix reaches the native chipset through its own Unix drivers: the console draws on the native bitplane display, and the /dev/screen screen manager hands chip-RAM bitmaps to programs such as the stock mono X server ✅. The stock kernel.c also carries an audio driver at character major 46 ✅.
  • ✅ Not loadable-module based (stock Amix; one third-party exception). Drivers are statically linked into the kernel; adding one means editing kernel.c and relinking /unix. See the driver model.
  • ✅ Not runnable on every Amiga. The stock kernel needs a 68020/68030 with a real MMU and a hardware FPU, its stock drivers are Zorro II only, and it does not run on the 68040/68060 (so the A4000 cannot officially run stock Amix) — the public 68040/68060 port lifts the CPU limit and runs the FPU-less 68LC060 with a software FPU, metal-proven on a real 68LC060 (A4000T + Z3660, 2026-09-25) ✅; a driver that page-maps its board, such as the A4091/A4092 SCSI driver, reaches Zorro III ✅ (see the Zorro III correction). The hard limits are detailed on the hardware page and the quirks checklist; the 040/060 situation has its own status page.

History at a glance

  • 🟡 First public demo: Uniforum, Dallas, January 1988, on an A2500UX — at that point still SVR3, not SVR4. (The 1988 Uniforum Dallas demo is documented; the machine and month are community-reported, not in primary sources.)
  • ✅ Commercial window: roughly 1991–1992.
  • ✅ Support ended 1993. Commodore filed for bankruptcy in April 1994.

The SVR4 product line proper begins with the 1991 releases and ends with the 2.1 retail release of February 1992 🟡 (the month is community-reported — sources confirm only the year), after which only an unofficial patch series continued. The full timeline is below; the versions reference carries the per-release detail.

Version matrix

The releases, with confidence tags carried straight from the research brief:

Version Date Notes Tag
SVR3.x precursors 1988–89 A2500UX demos (68020 → 68030); proprietary windowing, not yet SVR4 🟡
1.1 1991 First widely-referenced SVR4 release; mono X reported "slow as molasses" 🟡
2.0 / 2.01 / 2.03 1991 Color X via the A2410; archive.org carries 2.01 and 2.03 installers 🟡
2.1 Feb 1992 🟡 Last retail release ✅ (month community-reported). Ships pre-formatted SVR4 man pages (no nroff source; the Amiga sections are source) ✅
2.1 patch 2a → kernel 2.1c post-1992 Unofficial but considered definitive; inet / NFS fixes (no Y2K content — quirks §7) ✅

Notes on the matrix:

  • ✅ The locally analysed install media is 2.1. The boot, root, and patch floppies are named amix_21_boot.adf, amix_21_root.adf, and amix_21_patch.adf; see their anatomies under boot-disks and the install walkthrough.
  • ✅ The patch disk self-identifies as "Patch Disks 1 and 2 for International, USA-Only, 2-user, and Unlimited-User Amiga UNIX System V Release 4.0 Version 2.1." It greps the live uname -v for ^2\.1.* 08004..$ before applying, which is why the patch path lands on kernel 2.1c. See the patch disk anatomy.

Two things that do not exist (do not propagate the lore):

  • 🔴 "Amix 2.2" does not exist in any primary source. It is most likely confusion with kernel 2.1c. Do not cite a 2.2 release.
  • 🔴 The "2.1c, 1994" date seen on some wiki pages is almost certainly wrong — support ended in 1993.

Where to go next

See also

Sources

  • The 2026-09-29 Y2K correction ✅ — date 092912002026 accepted on a stock 2.1c box; the stock setclk boot run ending in date: bad conversion; the patch disk's 103-member cpio carrying no setclk (quirks §7).
  • Research brief §1 (identity, lineage, history; version matrix) — sources/research-brief.md.
  • Research brief §2 (the stock 68020/030 limit) and the 040/060 status page.
  • The A4000T metal run: Amix 2.1c reaches login: about 60 s after power-on on a real 68LC060 (the FPU-less 68060) on a Z3660 in an A4000T, running 060-direct (the card's CPU boot mode, not its 68030 emulation) with 128 MB of card RAM — first-party record, 2026-09-25 ✅.
  • The native-chipset display and audio facts: research brief §21 (the screen manager's DisplayScreen()/NewBitmap chip-RAM bitmap on the native display) ✅; jusii/amix-z3660rtg docs/X-INSTALL-RTG.md (the console login on the chipset screen; the stock startx on the native chipset) ✅; the stock cdevsw[] rows 10 screen and 46 audio on the major-number registry ✅.
  • The "quick and dirty" characterisation: research brief §1, untagged there and sourced to an EAB thread and datagubbe.se (community, 🟡).
  • The stock-only scope of the loadable-module, FPU and Zorro II limits: the kernel architecture page's "Loadable modules" section (linked above), the 040/060 status page (the 68LC060 software FPU) and the Zorro III correction.
  • Michael Ditto, Writing Amix Device Drivers, 1990 European Amiga Developer's Conference (opening line on direct-port lineage) — see bibliography.
  • amix_21_patch.adf analysis via tools/inspect-adf.sh (the Version 2.1 self-identification and the ^2\.1.* 08004..$ uname -v check).
  • amigaunix.com and its history page (community-reported timeline; 🟡 items).
  • en.wikipedia.org/wiki/Amiga_Unix and /wiki/Amiga_3000UX (background; corroborating the demo and commercial dates).