Skip to content

What we have, and what is still missing

Written 2026-08-27, after gathering sixteen sources and checking them against the manuals in this repo. This is the honest state of the research, not a plan.

Updated 2026-08-27 (later the same day): ten more ndwiki pages fetched, the sintran.com mirror searched, and several conflicts closed. See OCR-WANTED.md for what to run through OCR next.

The one structural finding - now confirmed by Ronny and by a wiki quoting ND

Four generations, one pattern. A 16-bit machine runs the operating system, with a faster, wider compute engine attached to it. Ronny confirmed this directly: "nd5 needs nd1, nd50 needs nd10 and nd500 or nd5000 needs nd100/110 or 120", and "the 5xxx machines are co-processors for the 1xxx machines".

Host Attached engine Evidence
NORD-1 NORD-5 "The operating system is contained in the NORD-1 computer" - ndwiki NORD-5, from ND-NYTT Sept 1972
NORD-10 NORD-50 ND-06.005.01 NORD-10 - NORD-50 Communication System and ND-60.116.01, both held here
ND-100 ND-500 "relied on a ND-100 to do housekeeping tasks and run the operating system" - ndwiki ND-500
ND-100/110/120 ND-5000 "uses a ND-100/ND-110/ND-120 CPU as I/O processor" - ndwiki ND-5000 family

And it is also how the company died. The ND-5000 page states the failure directly: as the 5000 line got faster, the dual architecture became bottlenecked because all I/O had to pass through the ND-100. English Wikipedia says the same thing from the other end - the ND-120 was nearly called ND-1000, and the mixed 16/32-bit architecture was increasingly seen inside ND as the bottleneck.

The ndwiki NORD-50 page puts it in ND's own terms - the NORD-50 is "a total slave to the NORD-10/S", with no I/O system and no interrupt system of its own, all I/O done through the NORD-10/S, which runs SINTRAN III/VS and the NORD-50 monitor and "submits jobs to NORD-50 for execution in batches". To the NORD-10 the NORD-50 "looks like any else device and is controlled via standard IOX instructions".

This is the spine of the whole document: one architectural idea, carried for twenty years, that eventually became the trap. Five NORD-50 manuals sit in the mirror waiting to prove it from primary sources - see OCR-WANTED.md.

Closed on 2026-08-27

  • What the NORD-50 was. Settled: a special-purpose array processor and total slave to the NORD-10/S, agreeing with the primary ND Design Goals document. The three secondary sources calling it a standalone 1975 supermini are repeating an error. Dead unless a primary document revives it.
  • NORD-42 is not a separate machine. It is an OEM version of the NORD-12, built for Noratom-Norcontrol, used as the processor in their DATABRIDGE system and in maritime simulators in Norway, China and Mexico.
  • NORD-20 is barely a separate machine either. "Almost identical with the NORD-2B ... electrically the same but the CPU is contained on six boards instead of ten." Released before the NORD-10; first installed January 1972; 43 systems by August 1974. It also carries the same description our NORD-2U entry came from, so NORD-2U and NORD-20 may be one machine.
  • NORD-2B has substance now: a simplified, cheaper NORD-1 with 4-16 kW core, a substantial number at CERN, and one at Norway's National Centre for Epilepsy processing live EEG to detect epileptic seizures. Hardware Manual dated December
    1. Serial 05 survives at Telemuseum.
  • The founding date. mirror/history/ND-names.html gives both: 7 July 1967 as a personal company, 19 September 1967 as an AS. English Wikipedia's "August 8, 1967" matches neither.
  • Where SINTRAN came from. Not ND. The original was released in 1968 by the Department of Engineering Cybernetics at NTH with SINTEF, and the name is a portmanteau of SINTEF and FORTRAN.
  • The image provenance. The three ND-5000 photographs are byte-for-byte the size of history/nd-5000.jpg, nd5800cpu.jpg and nd5850cpu.jpg in the mirror. Source confirmed; permission still not asked.
  • A whole afterlife nobody had. Nine more company names between 1992 and 2004, from an eyewitness. See sources/norsk-data-com-nd-names.md.

Closed on 2026-08-27, second pass

  • The host-and-slave architecture is now PRIMARY, not a wiki claim. Reference-Manuals/10/ND-06.005.01 NORD-10 - NORD-50 Communication System.md (August 1975): the NORD-10 "may regard the NORD-50 as an I/O device", and "the NORD-10 has complete control, and the NORD-50 is regarded as a slave to the NORD-10". Communication is by IOX instruction.
  • The bottleneck is visible in 1975. The same manual: all transfers are 16 bits wide, so "the NORD-10 must use two IOX instructions to transfer a NORD-50 word". The NORD-50's registers are all 32-bit. The constraint that the ND-5000 documentation complains about in 1987 is present in the first pairing.
  • MOVEW, TSET and RDUS are CX, not CE. The primary ND-100 Reference Manual: "The CX-option consists of improved CE-instructions (Commercial Extended) plus the following instructions (CX only): MOVEW ... TSET ... RDUS ... SINTRAN III: segment-change instructions." ndwiki was right and English Wikipedia was wrong. Opcodes: MOVEW 1431xx (xx 00-08), TSET 140123, RDUS 140127.
  • ND-550 and ND-560 are both real and distinct, from ND's own ADP/WMIPS ladder (510/CX 1.0, 530/CX 1.6, 550/CX 2.1, 560/CX 3.8, 570/CX 5.9).

Closed on 2026-08-27, third pass

  • The ND-5000 is not a new architecture. ND-05.009.4 EN ND-500 Reference Manual: "This manual is valid for both the ND-500 and the ND-5000 computer systems." One architecture, 1981-1990, in four implementations - ND-500/1, ND-500/2, Samson, Rallar. Even the '87 extensions run on the older CPUs.
  • The ND-500 and ND-5000 host relationship is now primary too. The same manual defines the term: "Whenever the I/O processor is mentioned, this means the ND-100 or the ND-110 processor." Three of the four host/slave pairings are now proved from ND documents; only NORD-1 with NORD-5 is still secondary.
  • What kind of machine the ND-500 was: dedicated local-variable-base and record-base registers, top-of-stack with low and high limit traps, string instruction and subroutine-entry chapters, and own/child/mother trap enables. A design built around compiled high-level languages.

Closed on 2026-08-27, fourth pass

All from Reference-Manuals/10/ND-06.008.01 NORD-10-S Reference Manual.md, which had not been opened before this pass.

  • The 18-bit memory word is an ordinary ND convention, from a primary manual: "Memory modules with 18 bits word length provide one parity bit per byte, while 21 bit modules are used for memory error correction." 16 data + 2 byte-parity =
    1. This was previously reachable only through a wiki quoting the NORD-12 manual. It still does not prove the NORD-1 case, but it makes the explanation ordinary.
  • The NORD-10 register count explained: "8 general registers for each program level and two scratch registers for each level to be used by the micro-processor" - 16 x 10 = 160 total, 16 x 8 = 128 program-visible. The wiki numbers were right; now we know the arithmetic.
  • Context switching improved between models: 1.5 us on the NORD-10 (Design Goals) and 1 us on the NORD-10/S (Reference Manual). Two ND documents, two machines - not a contradiction.
  • The ring system, in full, with ND's own recommended layout: ring 0 user programs, ring 1 compilers and assembler, ring 2 operating systems, ring 3 kernel. Both protection systems must pass, not either.
  • Paging detail: four page index tables of 64 words each, in high-speed 16-bit registers - which is where the wiki's "256 word block" comes from, 4 x 64.

Closed on 2026-08-27, fifth pass

From Reference-Manuals/10/ND-06.010.01-NORD-10-S-MICROPROGRAM.md, previously unread.

  • The NORD-10 control store, exactly: "1k x 32 bits Read Only Memory". Four microinstruction types - ARITHMETIC, INTERBLOCK, JUMP, LOOP - chosen by ROM bits 31 and 30. Hardware-generated entry points, a readable microprogram counter giving simple subroutines, entry points two apart.
  • What else lived in the ROM: the instruction repertoire, the operator panel driver, MOPC (operator communication in stop mode), the bootstrap loader and a memory check. Confirms the secondary claim and names the parts.
  • ND's own reason for microprogramming, contrasting with "a large and complicated Time Counter/Cycle Counter" in a non-microprogrammed machine - which is precisely the NORD-1, whose card list has card 123 Cycle counter and card 151 Time counter.
  • The 32-bit vs 48-bit floating point option was a microcode choice on the same hardware - the flexibility was a product, not just a design argument.
  • NORD-42 is primary-confirmed as NORD-10 family: the ROM covers "NORD's 10/S, 10, 42 instruction repertoire".

Closed on 2026-08-28, sixth pass

From Reference-Manuals/ND-06.015.02 ND-100 Functional Description.md, previously unread. The ND-100 rows were the last major ones still resting on wikis.

  • "ND-100 is a 16-bit general purpose single board computer" - the single-board claim is now primary, not a Wikipedia phrase.
  • Bit-slice, quantified: "a 4-bit subsection of the 16-bit wide ALU and register section" - so four slices, which is what the ND-110's BUFALU later replaced.
  • The control store widened: 2K x 64 bits on the ND-100 against the NORD-10's 1K x 32 bits.
  • The CX-option is a bigger PROM: "By expanding the microprogram PROM to 4 K by 64 bits, a number of instructions are introduced. These instructions comprise what is known as the CX-option." CE and CX were microcode, not silicon.
  • A writable control store existed on the ND-100 as an option - 256 words by 64 bits - years before the ND-110 made one standard.
  • Address space: 64 KW without the memory management system, 16 MW (32 MB) with it.
  • Address arithmetic was microcode on the ND-100 and hardware on the ND-110 (RMAC). Both ends of that trade are now documented.
  • Context switching is register-file selection, done by microprogram - the ND-100 manual states the architecture the NORD-10 introduced.

Closed on 2026-08-28, seventh pass

From Reference-Manuals/ND-06.026-1-EN ND-110 Functional Description.md. This was the last major machine whose evolution claims rested on a wiki.

  • All three gate arrays confirmed and defined by ND: RMAC the address-arithmetic array, BUFALU the 16-bit ALU array which "also contains the current register set", RMIC the microinstruction sequencer. The secondary sources' claim that BUFALU absorbed the register blocks is corroborated by ND's own wording.
  • What is on the one card: CPU, MMS, cache, control store, interrupt and trap handlers, timing and real-time clock, operator panel interface, terminal 1 serial interface, register file, cycle controller and bus interface.
  • The control store is 8K deep by 64 bits, in four 16-bit groups - so the progression is 1K x 32 (NORD-10), 2K x 64 (ND-100, 4K for CX), 8K x 64 (ND-110).
  • Independent confirmation of last pass's finding: "the microprogram control store is writeable (optional on ND-100)" - stated in the ND-110 manual, matching what the ND-100 manual said from the other side.
  • A new instruction worth noting: TRR CILP, opcode 150113, inhibits individual pages in cache. Also VERSN, opcode 140133, reads print and microprogram version numbers.

Closed on 2026-08-28, ninth pass

From SINTRAN/ND500/ND500-ND5000-INTERFACE-COMPREHENSIVE-GUIDE.md, the repo's own carve work from NPL source and ND manuals. This closes the gap named in the eighth pass - the ND-100-to-ND-500 mechanism.

  • The host link changed exactly twice in fourteen years: IOX instructions (NORD-10 to NORD-50, 1975), DMA over PCB 3022 with a bank of 16 IOX registers (ND-100 to ND-500, 1981), then octobus message passing (ND-5000, 1987).
  • The ND-500 interface is the NORD-50 idea unchanged in kind - LMAR5/RMAR5 for the memory address register, LSTA5/RSTA5 status, LCON5/RCON5 control, master clear, terminate, TAG in and out, limits, lock and unlock. A 16-bit host driving a 32-bit machine one 16-bit register at a time, six years later.
  • The dependency is literal in the hardware: the ND-500's status register carries a five-bit STOPREASON, and reason 1 is MOCALL, a monitor call. The 32-bit processor stops and asks the 16-bit machine to do the work.

Closed on 2026-08-28, tenth pass

From Reference-Manuals/500/ND-05.020.01 EN ND-5000 Hardware Description.md.

  • Each ND-5000 model ships with a named I/O processor: ND-5200 with an ND-110, ND-5400 and ND-5500 with an ND-110/CX, ND-5700 and ND-5800 with an ND-120/CX. The 16-bit line's generations are stepped inside the 32-bit range.
  • CPU type to model mapping confirmed from primary: type 1 = model 2 (ND-5200), type 2 = models 4, 5, 7 (ND-5400/5500/5700), type 3 = model 8 (ND-5800, ND-5900).
  • Type 1 is type 2 with the cache and AAP baby modules removed, and on the ND-5200 "floating-point operations are performed by the microprogram" - confirming what had only been a wiki claim.
  • Type 3 is not type 2 plus a layer: unique mother board and cache/IDA baby module, plus an IDAC "booster" module used only in the ND-5800 and ND-5900.
  • Clock speeds confirmed, including the counter-intuitive part: the ND-5400 and ND-5500 run slow (156 ns) while the cheaper ND-5200 runs at 70 ns.
  • Caveat: our OCR of the model/cache table has merged columns, so which caches each model enables is not readable line by line from this copy.

Opened on 2026-08-27

  • ND 1100/S, ND 1200/S, ND 1300/S, ND 1400/S - system-level product names for NORD-10/S plus NORD-50 combinations, per the ndwiki NORD-50 page. In no other source.
  • NORD-4 - the external ND-library in the mirror has a folder named 01-NORD-1 NORD-4. No source we have gathered mentions a NORD-4.

A correction from Ronny, 2026-08-28

The host-and-slave dependency runs one way only. The NORD-10 and the ND-100/110/120 were also delivered without the coprocessor - they were complete, standalone products, and the volume business. The NORD-5, NORD-50, ND-500 and ND-5000 are the ones that cannot run alone.

The document had been stating the pattern in a way that could be read backwards, as if the 16-bit machines existed to host the 32-bit ones. Fixed in MACHINE-TIMELINE.md section 3, with the installed-base ratio as evidence: 62 NORD-10 against 3 NORD-50 in 1975, 83 against 3 in 1976, 114 against 7 in 1977.

This sharpens rather than weakens the argument: the expensive half was optional and the cheap half stood alone, which is why the architecture survived twenty years - and why the eventual bottleneck hit the high-margin product while the volume machines carried on unaffected.

Coverage, machine by machine

Legend: P = primary ND document held in this repo. S = secondary source gathered. W = written up in History/.

Machine P S W Note
SAM, SAM 2 / FLINK - thin - Pre-company, built at FFI Kjeller. One sentence in each Wikipedia. FFI would hold the documents, not ND.
NORD-1 yes, 4 docs yes, 4 yes Done and verified. Eight open questions recorded.
NORD-2B - yes - Simplified, cheaper NORD-1, 4-16 kW core. Hardware Manual Dec 1970. Two manuals exist in the Tingo collection, neither held.
NORD-2U / NORD-20 - yes - Very likely one machine. Near-identical to the NORD-2B, CPU on six boards not ten, released before the NORD-10. 43 installed by Aug 1974.
NORD-4 - name only - New 2026-08-27. A mirror folder is named 01-NORD-1 NORD-4. Nothing else anywhere.
NORD-5 assembler only yes in NORD-1 Compute module; needs a NORD-1 host.
NORD-9 - name only - A bare name in Norwegian Wikipedia. No ndwiki page exists. May not be real.
NORD-10, NORD-10/S yes, rich yes, 2 in timeline ND-06.008.01 now read: rings, paging, register block and context-switch times all verified from primary. Ready to write as its own notes.
NORD-12 no manual yes - The manual exists in the mirror - heim.bitraf.no/tingo/.../NORD-12_Reference_Manual_ocr.pdf. Needs importing.
NORD-42 - yes - Not a separate design: an OEM NORD-12 for Noratom-Norcontrol.
NORD-50 yes, 2 comms docs yes, settled - Array processor, total slave to the NORD-10/S. Five more manuals sit in the mirror unimported.
ND 1100/S - ND 1400/S - one line - New 2026-08-27. System names for NORD-10/S plus NORD-50 combinations.
ND-100, /CE, /CX yes, rich yes, 2 in timeline ND-06.015.02 now read: single board, bit-slice width, 2K/4K x 64 control store, WCS option and address space all primary.
ND-110, /CX, PCX yes, 2 docs yes in timeline ND-06.026-1 now read: all three gate arrays, the one-card contents, the 8K x 64 writable control store and CILP verified from primary.
ND-120 / Delilah test programs only yes - No dedicated ND-120 manual found here.
ND-125 - one line - Appears nowhere else in this repo. Cited to sintran.com ECO 100-786, 1994.
Butterfly 110 / Teamstation - en.wp only - Ericsson PC/AT host, ND-110PCX on two ISA cards. One source.
ND Satellite range 3D model yes - Real: the COSMOS name was a pun on it, and there is a measured model in Hardware/3D-Models/.
ND-500 family yes, rich stub - Model numbers map to four implementations, not four designs.
ND-505 yes, product sheet yes - 29-bit user addressing settled. The 31-bit / snipped-pin-27 story is unverified.
ND-5000 family yes, rich yes partly, elsewhere SINTRAN/ND5000/ND5000-FAMILY-MODELS-REFERENCE.md covers this from primary manuals.
ND-5830 / 5850 Rallar some yes - KUSK and GAMP gate arrays. Date disputed, 1987 or 1990.
Server 88 / TpServer / ES thin one line - 1988. Almost nothing.
ND-88000 / Uniline / NDIX - en.wp only - The Unix years. One uncited source for the whole subject.
The company 1967-2004 15 annual reports in the mirror yes, 4 - Reports unimported. Plus an eyewitness account of every name change to 2004.

What we should get next, in order of value

1. Import from the mirror - it is already on this disk

Everything below is a PDF sitting in E:\Dev\Ronny\mirror-sintran-com, most of it already OCR'd once. Nothing needs finding. Full list and file paths in OCR-WANTED.md. In order:

  • Five NORD-50 manuals plus its maintenance manual and assembler - they prove the host-and-slave architecture from primary sources
  • Fifteen ND annual reports (1977-1992) - the primary basis for the whole company story, which currently rests on wikis
  • 103 company newspapers - 51 issues of FRND through the crisis years, plus ND-NYTT and the English-language ND News
  • The NORD-12 Reference Manual - the last big machine-manual gap
  • ND-06.019.01 ND-10 device programming instructions, and earlier editions of the NORD-500 Reference Manual and NORD-100 Functional Description

2. Documents named by our sources that we do not hold

  • "Norsk Data - hva gikk galt?" ("Norsk Data - what went wrong?"). The ndwiki history page cites this eight times, for the PLANC date, the SIBAS numbers, the ND-100/CE delivery, the ND-110 and ND-5000 introductions, SINTRAN-IV and Uniline. It is the single most-cited source behind our best secondary, and we have never seen it.
  • ND-NYTT No 5, September 1972 specifically. The mirror has three ND-NYTT issues, all 1981-82. The September 1972 issue is cited by two separate ndwiki articles - for the NORD-5 performance figures and for the NTNU machines that include NORD-1 serial 47 - and it is not in the mirror.
  • Annual reports for 1967-1976, 1981 and 1982 - the mirror has the other fifteen.
  • The two NORD-2B manuals - Hardware Manual Volume 1 (December 1970) and I/O System (March 1971), linked from ndwiki as being in the Tingo collection, not confirmed present in the mirror.
  • Cambridge Memories EXPANDACORE 18 documentation - the last route to the NORD-1 18-bit core question.
  • The two books: Heradstveit, Eventyret Norsk Data (1985), ISBN 8272010402; and Steine, Fenomenet Norsk Data (1992), ISBN 8200215016.

3. Questions answerable from manuals already on this disk

Nobody needs to find anything for these. They are sitting here unread.

  • When was the NORD-10 dropped from SINTRAN III? ndwiki says version H was the last for the NORD-10; our own SINTRAN/Release-Documentation/SINTRAN-III-Release-History.md says NORD-10 was dropped at J. If H was last, the drop was at I.
  • ND-60.116.01 NORD-10 - NORD-50 Operator's Guide is still unopened. ND-06.005.01 has now been read - see "Closed" above.
  • What separates the ND-500/1 from the ND-500/2, and what each of the four ND-500 implementations actually changed. Reference-Manuals/500/ is unread.

4. Image licences

Three ND-5000-family images in images/nd-5000/ have no established licence and are presumed to come from norsk-data.com, which grants none. They are fine for research here and must not be published without asking. See images/CREDITS.md.

Conflicts still open across all sources

Kept in one place so they are not quietly resolved. Details and citations are in each source file header, and the NORD-1 ones in NORD-1.md.

Question Positions
Founding date 7 July 1967 (timeline, Broennoeysund; Norwegian Wikipedia) vs 8 August 1967 (English Wikipedia)
How many founders Three (both Wikipedias) vs four, adding Terje Mikalsen (timeline)
First NORD-1 delivery 1967 to Norcontrol (Norwegian Wikipedia) vs nine sold in 1968 (timeline) vs three installed 1968 (ndwiki history)
When Skaar became MD 1977 (Norwegian Wikipedia) vs 1978 (timeline). Kolbjoern Johansen 1972-77 appears in one source only
What the NORD-50 was 1973 array processor (primary ND Design Goals) vs 1975 second-generation supermini (three secondaries)
NORD-1 built "at least 142" (1974 magazine survey) vs ndwiki history annual counts that pass 160 by 1974
ND-100 data channel rate 16 Mbit/s (Reference Manual 2.2) vs ~20 Mbit/s (same manual, 1.2)
NORD-1 interrupt levels 15, then 256, then a flat 16 - all in the same primary manual
ND-5850 launch 1987 (both Wikipedias) vs 1990 (timeline)
ND-5000 introduction 1985 series (ndwiki history) vs 1987 family (everywhere else)
Stock listings London 1981 + New York/Nasdaq 1982 (both Wikipedias) vs London 1981 + Stockholm 1983 (timeline)
~~MOVEW/TSET/RDUS~~ SETTLED: CX only, from the primary ND-100 Reference Manual. English Wikipedia is wrong
ND-505 width 28-bit (Norwegian Wikipedia) vs 29-bit user addressing (settled - our own ND-505CX product sheet) vs 31-bit physical (ndwiki, unverified)
Tandberg bought 1979 and kept (timeline) vs bought 1979 and independent again autumn 1980 (ndwiki history)
First CERN contract 1972 (timeline) vs 1974, Lab II, NORD-10 (ndwiki history)

One thing to be careful about

Several of our secondary sources are not independent of each other. The ndwiki ND-100 article says it began as a copy of the English Wikipedia article; the ndwiki history page says it began as a copy of a NODAF wiki article; the ndwiki NORD-1 article says it began as a copy of English Wikipedia. When two of them agree, that may be one source counted twice. Agreement between wikis is not corroboration.

The primary manuals in Reference-Manuals/ do not have that problem, and this repo holds a lot of them. Most of the questions above are answerable from documents already on this disk.