Skip to content

History of the Norsk Data machines

The machines. What each one was, how it worked, what changed between generations, and how they relate to each other.

Ronny's scope, 2026-08-27: "i am mostly interrested in the history of the machines, not the business of people". So this is hardware history. The company story - founders, listings, acquisitions, the collapse, the afterlife - is gathered in sources/ where it turned up, and stays there. It gets used only where it explains a machine: why a model was renamed, why a line was cut, why the ND-505 was built narrow to clear an embargo.

This is a research collection, not a finished document. We are still gathering. Files are named after machines, not numbered, because the shape and order of the final document is not knowable until the material is in - how many machines there turn out to be, which belong together, and which are variants rather than generations.

Gathering has already changed what any structure would have to be: the NORD-5 is a compute module for a NORD-1 host rather than a separate generation; the NORD-50 is an array processor, not the supermini one timeline calls it; a NORD-2U turned up in a primary ND document, in none of the timelines being used; and an image shows a machine badged ND-5700 Compact, another model in none of them. Any ordering fixed before that work would already be wrong.

This folder is our own writing. Official scanned manuals stay in Reference-Manuals; operating-system detail stays under SINTRAN. Every claim needs a source citation, and anything unproven is marked ASSUMPTION: or UNVERIFIED:.

Start here

  • machines/ - one page per machine, 46 of them, all to the same shape: header table, plain-English paragraph, specification, what was new. Start here if you want to look something up rather than read a story.
  • MACHINE-TIMELINE.md - the main document: every machine with its year, how important it was, and the six threads along which the hardware evolved - logic technology, control store, speed, caching, memory and protection, and interrupts. Plus the one architectural idea that shaped the whole range and then ended it.
  • WHAT-WE-STILL-NEED.md - the state of the research: what is covered machine by machine, what is settled, and every conflict still open across the sources.
  • OCR-WANTED.md - machine documents sitting as PDFs in the local sintran.com mirror that are not yet in this repo. Nothing needs finding; it needs OCR and importing.

Written so far

File Covers State
machines/ 46 spec sheets, one per machine and configuration One page each, same template throughout. Images to be added later via Claude Design.
MACHINE-TIMELINE.md Every machine, 1962-1994 Chronology, importance and hardware evolution. Every claim marked primary, secondary or disputed. The ND-500/5000 half is thinner than the 16-bit half.
NORD-1.md NORD-1, and the NORD-5 compute module Checked against every primary document in Reference-Manuals/1/ - the Feb 1970 Reference Manual, ND-01.004.01 Hardware Manual vols I and II, ND-01.005.01 Connectors. Eight open questions recorded, none faked shut.

The machines to gather, and the order to do it in

The order below follows two things at once, which happen to agree: the machines run roughly chronologically, and the primary material in this repo gets denser as you go down the list, until the last two groups where it thins out again. Each group is a batch because its machines share documents - you cannot sensibly do a host without its slave, or a CPU without its variants.

The list roughly doubled on 2026-08-27 when the two Wikipedia articles were added. Between them they name models that appear in nothing else we hold: Nord-2B, Nord-9, Nord-20, ND-100/CE, ND-Satellite, ND-505, ND-510/CX, ND-520, ND-530, ND-550, ND-5400, ND-5900-2, ND-5900-3, ND-5904, ND-88000, and the Uniline 33 and Uniline 88 Unix ranges. Both pages are uncited, so most of those are currently names with no documentation behind them - working out which are real machines, which are variants and which are errors is itself part of the gathering.

Two whole subjects also arrived with the English article and have no home in the list below yet: the pre-company machines at FFI Kjeller (SAM, and SAM 2 also called FLINK), and the Unix years - DIAB, SCO System V, NDIX, the Motorola 68030 and 88000 ranges, and Dolphin Server Technology.

1. The NORD-10 family - NORD-2U, NORD-2B, NORD-9, NORD-10, NORD-10/S, NORD-12, NORD-20, NORD-42, NORD-50

Do this next. It is the immediate successor to the work already done, and Reference-Manuals/10/ is a rich folder: the ND-06.008.01 NORD-10/S Reference Manual, ND-06.010.01 microprogram, the Design Goals document, the operator's guide and panel description, drawings, and verification programs.

The NORD-50 belongs in this batch rather than on its own: ND-06.005.01 NORD-10 - NORD-50 Communication System and ND-60.116.01 NORD-10 - NORD-50 Operator's Guide describe a host-and-slave pairing, the same arrangement the NORD-1 and NORD-5 had.

NORD-20 is no longer a bare name: the ndwiki history has development starting in 1971 and the machine introduced in 1972. It also adds a NORD-42, completed 1974 with MOS memory on 4Kb chips, which appears in no other source at all.

Gaps going in: we hold no NORD-12 Reference Manual (a wiki transcription is the only route to it); NORD-2U rests on one line in the Design Goals document; and NORD-2B and NORD-9 are still bare names. Whether NORD-2, NORD-2B and NORD-2U are one machine, two or three is unknown.

Two systems rather than machines belong with this group: NORDIC, developed 1971 and completed 1972 - almost certainly the multicomputer installation at the Meteorological Institute that ND-NYTT wrote about in September 1972 - and NORDCOM, with a NORDCOM-74 revision and a graphics variant.

2. The ND-100 line - ND-100, ND-100/CX, ND-100/CE, ND-100 Compact, ND-110, ND-120/CX, ND-Satellite

The best-documented machine in the repo, and the one the rest of the project already knows most about - SINTRAN/, the emulator work and several skills all describe ND-100 behaviour, so claims can be cross-checked against working knowledge rather than only against paper. The ND-120CX "Delilah" arrives 1985 per the timeline; ND-100 Compacts were still being delivered in 2001. /CE and the /CE comes from Norwegian Wikipedia and needs confirming. ND-Satellite is confirmed: the ND-COSMOS article describes it as a small computer ND had recently released, and says the COSMOS name was an internal pun on it - and Hardware/3D-Models/ already holds a measured ND100-Satellite model.

3. The ND-500 line - ND-500, ND-505, ND-510/CX, ND-520, ND-530, ND-540, ND-550/560, ND-570/CX, ND-570/CXA, ND-580/CX

Reference-Manuals/500/, plus Reference-Manuals/ND-05.009.4 EN ND-500 Reference Manual.md and the existing SINTRAN/ND500/ analysis. This is where the "architecture that is not von Neumann based" claim and the 1983 world's-fastest claim need testing, and where the ND-100/ND-500 interface work already in the repo becomes directly useful.

The ndwiki history says the ND-500 was split into ND-520, ND-560 and ND-570 in 1982 - and marks that entry with its own "(?)". Note ND-560 against Norwegian Wikipedia's ND-550; they may be the same machine misremembered. It also adds the ND-570/CXA in 1983 and, in 1986, the ND-580/CX models 20, 30 and 40 - "systems based on two to four ND-570 CPU:s and one ND-110/CX CPU", a multiprocessor line no other source mentions.

One entry here is worth the trip on its own: ND-505, a cut-down ND-500 built to clear the CoCom embargo on the Eastern bloc - a deliberately narrowed export model. If true that is a real story. It is uncited in both places it appears, and the two disagree on the number that is the entire point of the machine: Norwegian Wikipedia says 28-bit, English Wikipedia says 29 bit addresses.

4. The ND-5000 family - ND-5000/Samson, ND-5400, ND-5700, ND-5800, ND-5850/Rallar, ND-5900-2, ND-5900-3, ND-5904

Reference-Manuals/500/ also holds the ND-5000 Hardware Description, Hardware Maintenance, Microprogram Guide, Memory Management System and SAMSON Expected Behaviour. Images for three of these are already in images/nd-5000/. The ND-5700 appears on a machine badge in one of those images and in both Wikipedia articles, but in no timeline. English Wikipedia groups 5400, 5700 and 5800 under the ND-5000 name, and calls ND-5900-2, ND-5900-3 and ND-5904 dual-, triple- and quad-CPU machines of the same series - which, if true, means several of these "models" are configurations rather than designs. The sources also disagree on when the ND-5850 arrived: 1987 in both Wikipedias, 1990 in the norsk-data.com timeline.

5. Server 88, the Unix years, and ND-88000

Server 88 gave the TpServer and ES platforms in 1988. Around it sits the Unix push: the DIAB collaboration from 1987, the 1988 Santa Cruz Operation agreement for System V on ND's Intel PCs, NDIX (ND's own Unix for its own architecture), the Uniline 33 on Motorola 68030 in 1989, ND-88000 and the Uniline 88 built by Dolphin, and the Data General Aviion reselling deal.

This is where the architecture story ends - ND leaving its own CPUs. Thin everywhere except one uncited Wikipedia article, so it goes after the groups with manuals.

6. Before the company - SAM and SAM 2 / FLINK

The founders came from Forsvarets forskningsinstitutt at Kjeller, where they built the SAM and the SAM 2, also called FLINK, said to be the basis for the NORD-1. Not Norsk Data machines, but the machines Norsk Data came out of. FFI, not ND, would hold the documents.

Not a group: the company story

Parked by Ronny's scope decision. The material is gathered - fifteen annual reports and 103 company newspapers sit in the mirror, and sources/norsk-data-com-nd-names.md has an eyewitness account of every name change to 2004 - but it is not what this document is for. Pull from it only when it explains a machine.

Two things sit outside this order and can be picked up whenever they become interesting: terminals and peripherals (the ND-246 terminal, the Technostation, NOTIS), and the software lines (SINTRAN II and III, Nord TSS, SIBAS, MAC, NPL, PLANC, QED, PED, LED, FOCUS, NOTIS, NORTEXT, Technovision, BIBDIA, and the networking pair NORDNET then ND-COSMOS), which mostly belong under SINTRAN/ rather than here.

Two books nobody here has read

  • Heradstveit, Per Oeyvind (1985), Eventyret Norsk Data, Stenersen, ISBN 8272010402
  • Steine, Tor Olav (1992), Fenomenet Norsk Data, Universitetsforlaget, ISBN 8200215016

How sources are handled here

  • Primary = an ND document (manual, annual report, drawing, ND-NYTT). Those live in Reference-Manuals/ and are cited by their ND number.
  • Secondary = someone else's retelling. Copied verbatim into sources/, each with a header saying where it came from, when it was fetched, and what is known to be wrong in it.
  • The notes say which of the two each claim rests on. A claim two secondary sources agree on is still secondary.
  • Contradictions get recorded where they are found, both figures and both sources named. They do not get quietly resolved.

Sources collected so far

File What it is
sources/norsk-data-com-timeline.md The year-by-year company timeline from norsk-data.com (Sintran Data). Norwegian, verbatim. Its own sources are ND annual reports and the 1987 anniversary book. Carries a list of its known errors.
sources/ndwiki-nord-1.md ndwiki's NORD-1 article - specs, panel, surviving machines.
sources/ndwiki-nord-1-cpu-detail.md ndwiki's card-level CPU description - which register sits on which board.
sources/ndwiki-nord-1-boards.md ndwiki's board list, board number to ND drawing number to name.
sources/ndwiki-nord-1-serial-47.md One surviving machine documented card by card, with a restoration log and a period price list.
sources/ndwiki-nord-5.md ndwiki's NORD-5 article - the compute-module architecture and its timings.
sources/ndwiki-nord-12.md ndwiki's NORD-12 article - largely transcribed from the NORD-12 Reference Manual, which we do not hold. Carries ND's 16-data-plus-2-parity = 18-bit memory word.
sources/no-wikipedia-norsk-data.md Norwegian Wikipedia's Norsk Data article. Uncited. The FFI/SAM pre-history, the full list of managing directors, and a wide machine list. Conflicts listed in its header.
sources/en-wikipedia-norsk-data.md English Wikipedia's Norsk Data article - nearly three times the Norwegian one and not a translation of it. Uncited, but the only source here covering the Unix years, Dolphin, the Telenor purchase, the UK afterlife, and Lithuania. Conflicts listed in its header.
sources/en-wikipedia-nd-cosmos.md English Wikipedia on ND-COSMOS. Short and uncited, but partly corroborated from inside this repo: NORDNET is confirmed by a primary ND manual, and the Ethernet and HDLC link layers by our own measured XMSG work.
sources/ndwiki-history-of-norsk-data.md ndwiki's year-by-year history, 1967-1989. The best-sourced secondary we have - twenty numbered references, several to real published work. Adds NORD-20, NORD-42, NORDIC, NORDCOM, ND-560, ND-570/CXA, ND-580/CX, "Samsonite", SINTRAN-IV, and named installations including SCANNET and JET. Its installation counts are ambiguous - see its header.
sources/ndwiki-nd-100.md ndwiki on the ND-100 line. Two of its claims verified against our primary manuals - the CX instruction mnemonics and the 39.3216 MHz ND-110 oscillator. Adds the NORD-10/M origin, the ND-125/CX, the RMIC/BUFALU/RMAC gate arrays and the Samson/Delilah joke.
sources/en-wikipedia-nord-100.md English Wikipedia on the Nord-100. Shares an ancestor with the ndwiki page, so mostly duplicate - except the ND-110CX 1986 renewal and the whole Butterfly-110 / Teamstation workstation story, which is here and nowhere else.
sources/ndwiki-nd-500.md ndwiki on the ND-500. A stub, but it states the host-and-slave arrangement plainly and maps model numbers onto four implementations. Its ND-505 question is settled from a product sheet we hold.
sources/ndwiki-nd-5000-family.md ndwiki on the ND-5000 family. Largely redundant with SINTRAN/ND5000/ND5000-FAMILY-MODELS-REFERENCE.md, which is primary-based and better - but a compact statement that the models are mostly one design differentiated by CPU type, a clock jumper and enabled caches.
sources/ndwiki-sintran-iii.md ndwiki on SINTRAN III. Mostly redundant with SINTRAN/Release-Documentation/SINTRAN-III-Release-History.md, which is better and primary-based - but adds versions A to I, the SINTRAN I/II and NORD-TSS predecessors, and what the system felt like to use.

| sources/ndwiki-nord-50.md | ndwiki on the NORD-50. Settles what it was - a special-purpose array processor and "total slave to the NORD-10/S", agreeing with the primary ND Design Goals against three secondary sources. Also names the ND 1100/S - ND 1400/S systems. | | sources/ndwiki-nord-10.md | ndwiki on the NORD-10. 24 boards, 160 registers, 1K ROM microprogram, 16-to-18-bit paging, and the four protection rings. Warns it is drawn from an ND introduction document we hold. | | sources/ndwiki-nord-10-s.md | ndwiki on the NORD-10/S. Short; an index to the primary manual we hold. | | sources/ndwiki-nord-20.md | ndwiki on the NORD-20. Near-identical to the NORD-2B, six CPU boards not ten, released before the NORD-10. Price and cycle-time comparison against the NORD-10. | | sources/ndwiki-nord-2b.md | ndwiki on the NORD-2B. A simplified, cheaper NORD-1; many at CERN; one processing live EEG for epilepsy research. | | sources/ndwiki-nord-42.md | ndwiki on the NORD-42. An OEM NORD-12 built for Noratom-Norcontrol, not a separate design. | | sources/ndwiki-sintran-ii.md | ndwiki on SINTRAN II. SINTRAN is not an ND name - NTH and SINTEF, 1968, and the name is SINTEF plus FORTRAN. | | sources/ndwiki-nord-tss.md | ndwiki on NORD-TSS, the timesharing half of the pre-SINTRAN-III world. | | sources/ndwiki-nord-pl.md | ndwiki on NORD PL, the language SINTRAN III was written in. | | sources/ndwiki-nd-notis.md | ndwiki on ND-NOTIS. Matters here because COSMOS and the Butterfly existed partly to run it. | | sources/norsk-data-com-nd-names.md | norsk-data.com's company-name table, 1967-2004, with an eyewitness commentary. Company history, parked - but it settles the founding dates and is where most of our photographs come from. |

Note on fetching ndwiki: as of 2026-08-27 the live site sits behind an Anubis proof-of-work gate, and both the article URLs and api.php refuse any script. The copies here were taken from the Wayback Machine, and each file names its snapshot date.

Images

Pictures live in images/, one sub-folder per machine. Every one is listed in images/CREDITS.md with its author and licence.

Corrections made to the rest of the repo

  • Reference-Manuals/1/README.md listed "Binder NORD-1 ITT-1600" as a NORD-1 peripheral binder. It is not - it is Norwegian study notes on the ITT-1600, a different computer, bound with TI datasheets. Fixed, and the file marked "do not cite for any NORD-1 fact". It is the only file in that folder mentioning parity, so a grep makes it look like the answer to the NORD-1's 18-bit core question. It is not.

Where the manuals are

Most ND manuals sit at the top level of Reference-Manuals/. Three machine folders break that out: