Meteorologist, oceanographer, Python solvem probler, experimenter, Raspberry Pi tinkerer, geek dad, manager, space enthusiast, ham radio lover, science columnist, ADHDer, synthetic intellect curious, linux admin cosplayer.

Control Data Corporation: Institutional Founding, the CDC 1604, and the Soviet Sale

Research notes for a long-form NWP/HPC history blog post on michalbrennek.github.io. Scope: institutional/business history, customer deployments, the Soviet sale to JINR Dubna, and the Chippewa Falls relocation. Machine architecture and individual biographies are covered in sibling research files (Seymour_Cray.md, etc.).

Confidence flags appear inline. HIGH = multiple independent secondary sources agree. MEDIUM = single solid source or repeated lore. LOW = popular history that I could not corroborate to a primary or near-primary source. Items flagged LOW must be hedged or omitted in the final post.

ERA Origins 1946: From CSAW Codebreaking to St. Paul

Engineering Research Associates was incorporated in January 1946 in St. Paul, Minnesota. The technical core came directly out of the U.S. Navy’s wartime cryptologic effort. The relevant Navy unit was Communications Supplementary Activity – Washington (CSAW), also referred to as the Naval Computing Machine Laboratory in some accounts. CSAW operated from the Nebraska Avenue Complex in Washington and was the U.S. Navy counterpart to Bletchley Park, building special-purpose machines to attack Japanese and German cryptosystems.

The key founders were:

  • John E. Parker – the financier. A Naval Academy graduate working in corporate finance who had run the Northwest Aeronautical Corporation (NAC) in St. Paul, a wartime glider manufacturer. He raised the initial capital of approximately $220 000 (Ramsey County Historical Society). Confidence: HIGH.
  • Howard T. Engstrom – Yale-trained mathematician, former CSAW supervisor; later served as deputy director of NSA in the late 1950s. Confidence: HIGH.
  • William C. Norris – former CSAW lieutenant commander, codebreaker, electrical engineer. Confidence: HIGH.
  • Ralph I. Meader – former head of the Naval Computing Machine Laboratory. Confidence: HIGH (Wikipedia ERA, MIT Press Computers and Commerce).
  • Joseph Wenger – senior Navy cryptologist, listed alongside Norris and Engstrom in some accounts. Confidence: MEDIUM.

Around forty former CSAW colleagues were hired in early 1946 and moved into Parker’s idle NAC glider factory at 1902 Minnehaha Avenue West, St. Paul. The factory had finished its wartime work and the buildings were available, which is the prosaic reason ERA wound up in Minnesota rather than near Washington. Confidence: HIGH (Ramsey County Historical Society; Digital State exhibit, University of Minnesota).

The continuity from cryptanalysis is direct rather than metaphorical. ERA’s first contracts came from the same Navy that had paid CSAW. The first machine, Goldberg (completed 1947), used drum memory derived directly from CSAW work. Most senior staff held active Navy clearances. The Navy’s procurement officers in OP-20-G considered ERA effectively an extramural extension of their own laboratory.

Seymour Cray joined ERA in 1950 as a junior engineer, fresh out of his University of Minnesota MSc (1951). Frank Mullaney had joined in 1955 and would follow Norris and Cray to CDC. Confidence: HIGH (Cray Wikipedia; mncomputinghistory.com).

ATLAS and the ERA 1101: Codebreaking Goes Commercial

ERA’s flagship classified project was ATLAS, a stored-program computer built for NSA (or, more precisely, for the Armed Forces Security Agency, NSA’s predecessor) and delivered in December 1950. ATLAS was designated “Task 13” in the Navy contract paperwork. The commercial twin, the ERA 1101, took its model number from the binary representation of 13 (1101 binary = 13 decimal). The 1101 was unveiled in December 1951 as the first stored-program computer commercially available in the United States. Confidence: HIGH (Wikipedia ERA).

A larger successor, ATLAS II, was delivered to NSA in September 1953. Its commercial version was the ERA 1103, the first commercially successful scientific computer in the U.S. and the machine on which Cray cut his teeth. The design DNA from the 1103 – 36-bit word, parallel arithmetic, drum and core memory – ran straight into the CDC 1604 seven years later.

The ERA-NSA relationship is the buried backbone of this story. The same engineers who broke Japanese ciphers in 1944 built scientific machines for NSA in the early 1950s, then walked out of Sperry-Rand to found CDC, then designed the supercomputers that would in turn be used by Lawrence Livermore and Los Alamos for nuclear weapons design. The lineage is unbroken from Pearl Harbor to the H-bomb to the 6600.

Sperry-Rand Acquisition 1952

Why ERA was sold. ERA in the late 1940s was undercapitalized and had generated political trouble for itself. The principal driver of the sale was John Parker’s exposure to a Senate investigation into how government cryptologic contracts had been awarded to a company in which Navy officers had financial stakes. Parker had also raised conflict-of-interest concerns by drawing investment capital from the same circles that had channelled wartime work to ERA. The Senate Naval Affairs Subcommittee aired these issues in 1949-1950. Confidence: HIGH (Wikipedia ERA; MIT Press Computers and Commerce, Norberg).

By 1951, Parker faced two pressures: (1) personal investors wanted liquidity, (2) the company needed working capital to compete with IBM’s emerging 700 series. ERA was sold to Remington Rand in 1952. Norris stayed; Engstrom stayed for a while.

In 1955, Remington Rand merged with the Sperry Corporation (the gyroscope and avionics firm) to form Sperry-Rand. The combined company lumped two distinct computer cultures into one division: the St. Paul / ERA team (drum memory, scientific computing, Navy contracts) and the Philadelphia / Eckert-Mauchly Computer Company team (UNIVAC, mercury delay-line memory, commercial transaction processing). Eckert and Mauchly had themselves been swallowed by Remington Rand in 1950.

The technical staff in St. Paul felt absorbed rather than acquired. Engineers who had run their own classified Navy projects were now division-level functionaries at a conglomerate whose CEO understood gyroscopes better than transistors.

Sperry-Rand Years 1952-1957: The St. Paul / Philadelphia Tension

During this five-year stretch, three things ground at the ERA veterans:

  1. Strategic neglect of scientific computing. Sperry-Rand bet the UNIVAC division on commercial transaction processing – the UNIVAC I and II had been hugely visible (election night 1952) but yielded narrow margins against IBM’s 700 series. Investment was directed to Philadelphia, not St. Paul. ERA’s drum-memory technology was repurposed inside the UNIVAC 1103 with attribution sometimes obscured. Confidence: HIGH.

  2. Philadelphia made the calls. Product cycle decisions were made by Sperry-Rand’s UNIVAC Division leadership in Philadelphia and New York. St. Paul’s proposals frequently stalled. The line attributed to several ERA veterans was that “we cannot ship anything” – meaning that even when St. Paul completed a design, corporate would not bless production. This phrasing recurs in CDC corporate folklore but I have not located a cited primary source for it; flag as MEDIUM.

  3. Norris was demoted. William Norris had been a vice-president at the UNIVAC Division. At some point in 1956-1957 he was demoted out of that role – accounts vary on whether it was reorganization or punishment. Confidence: HIGH on the demotion; MEDIUM on the exact mechanism.

Cray spent these years on classified Navy work. His major project at Sperry was the prototype shipboard computer for the Naval Tactical Data System, designated the AN/USQ-17 (also called the Univac M-460). It was Cray’s last design at Sperry. The prototype design was completed before he left in summer 1957; Sperry-Univac engineers then redesigned the entire machine using silicon transistors and won a $50 million Navy production contract for what became the AN/USQ-20 / CP-642. The CP-642 was the actual seagoing NTDS computer through the 1960s. Confidence: HIGH (Wikipedia AN/USQ-17; ETHW NTDS oral history).

Crucially: NTDS in production used UNIVAC CP-642 hardware, not the CDC 1604. Some popular accounts conflate Cray’s NTDS prototype work at Sperry with later CDC 1604 deliveries to the Navy. The two are different machines for different purposes; the CDC 160 (a small CDC peripheral computer, not the 1604) did sometimes appear in NTDS-adjacent shore installations, but the seagoing NTDS hardware was UNIVAC’s. Flag for the post: do not say “CDC 1604 ran NTDS.” Say “Cray designed the NTDS prototype at Sperry before walking out.”

CDC Founding 1957: The Walkout and the Seven

The chronology that the contemporaneous record supports:

  • Early 1957: Sperry-Univac engineers Willis K. Drake, Arnold Ryden (former ERA treasurer), William Norris, and – in initial discussions – Seymour Cray formulate a plan for a new company. Confidence: HIGH.
  • 8 July 1957: Articles of incorporation filed in Minnesota by Fremont Fletcher, Abbot L. Fletcher, and D.P. Wassenberg (legal incorporators, not engineers). Confidence: HIGH (mncomputinghistory.com).
  • Mid-July 1957: Norris and roughly a dozen Sperry engineers resign. Confidence: HIGH.
  • 1 September 1957 (approximate): Operations begin. Confidence: MEDIUM on exact date. Some sources give early autumn.

The “seven founders” or “eight founders” lists vary across secondary sources. The most commonly named:

Name Role Confidence
William C. Norris President, CEO, principal organizer HIGH
Frank C. Mullaney Engineering, ex-ERA HIGH
Seymour Cray Chief designer (joined later in 1957/early 1958 after finishing AN/USQ-17 prototype work) HIGH
Arnold Ryden Treasurer, financial architect, ex-ERA HIGH
Willis K. Drake Co-organizer, used Minnesota Blue Sky law to enable secret stock sale HIGH
Henry Forrest Engineering, ex-Sperry MEDIUM (named in Murray The Supermen but I could not corroborate elsewhere; flag)
Robert (Bob) Kisch Engineering, ex-ERA MEDIUM (similarly Murray; some VIP Club sources confirm)
Robert Perkins Engineering MEDIUM (VIP Club)
Howard Sheckels Engineering MEDIUM (VIP Club)
Bill Keye Engineering LOW

Recommendation for the post: name Norris, Mullaney, Cray, Ryden, Drake, Forrest, Kisch as the canonical seven, and note that “the seven founders” is itself a tidied retrospective construction – in reality a dozen or more former Sperry employees followed Norris within months.

Cray’s delayed start. Cray formally signed on several months after the others. He was finishing the AN/USQ-17 prototype for Sperry under contract obligations. Norris declined to hire him until that contract was honourably completed – a small detail that captures Norris’s view of professional conduct. Cray joined CDC effective in early 1958. Confidence: HIGH.

Norris’s personal investment. The $75 000 figure cited in the user brief I could not corroborate to a primary source. Some accounts say Norris mortgaged his home; some say he put in considerably less. Flag as LOW. Safer phrasing: “Norris was the largest individual investor among the founders; reported figures vary.”

Initial location. Two locations show up in different secondary sources:

  • An old warehouse at 501 Park Avenue, Minneapolis (across the river from Sperry’s St. Paul plant). Confidence: HIGH (Wikipedia CDC).
  • The McGill Building in downtown Minneapolis. Confidence: HIGH (Norris Innovation Center).

These may be sequential – 501 Park Avenue first, then a move to the McGill Building as headcount grew – or the two names may refer to the same property. The post can simply say “an old warehouse in Minneapolis” without taking a position. CDC later (early 1960s) moved its corporate headquarters to Bloomington, Minnesota.

The IPO: 600 000 Shares at One Dollar

The famous IPO took the form of 600 000 shares of common stock at $1 per share, raising $600 000 in founding capital. Confidence: HIGH.

Three details deserve scrutiny:

  1. The Minnesota Blue Sky exemption. Drake and Ryden used a then-obscure provision of Minnesota securities law that permitted stock sales restricted to Minnesota residents to proceed without filing with the federal Securities and Exchange Commission. This kept the offering off the SEC’s radar – and crucially off Sperry-Rand’s radar – until the company was already capitalized. By the time Sperry-Rand learned what was happening, CDC was a going concern. Confidence: HIGH (mncomputinghistory.com; Digital State exhibit).

  2. Kalman & Company of Minneapolis is widely cited as the broker. The 1957 Commercial and Financial Chronicle (FRASER, St. Louis Fed) confirms Kalman & Co. as an active local underwriter. Confidence: MEDIUM-HIGH on Kalman as broker; LOW on whether they were the only broker.

  3. Oversubscribed within hours / days. This is part of the standard popular telling. Confidence: MEDIUM. The IPO was certainly fully subscribed; “within hours” is the kind of detail that gets sharpened in retellings. Safer phrasing: “the offering sold out quickly.”

What this IPO actually is, in business-history terms: it is one of the first significant engineer-led equity exits before the term “venture capital” existed. The pattern that became normal in Silicon Valley in the 1970s – a group of engineers walking out of an established firm, raising friends-and-family money locally, going public early on a thin prospectus – ran first in Minneapolis in 1957. The 8-page prospectus is itself a document that would be illegal under modern securities law.

Building the Company 1957-1960

The first year was hard. Norris cut salaries – including his own – to keep the company alive. By 1958 contracts started arriving from Navy and Air Force customers who had been ERA clients. CDC’s strategic decision under Norris was clear and unromantic: focus on scientific computing, leave commercial transaction processing to IBM and Sperry. The customer profile would be national labs, military research installations, and high-end engineering firms. Margins on each unit would be high; volumes would be small.

Cray, hired in early 1958, was given a small team and told to build a machine. Cray’s preferred size was three to five engineers reporting to him personally. The first product was the CDC 1604, a 48-bit transistorized machine that was effectively a re-implementation of the ERA-Sperry 1103 – the same architectural lineage, but in transistors rather than vacuum tubes. Confidence: HIGH.

Industry expectations of CDC in 1958-1959 ranged from indifferent to dismissive. IBM’s announcement of the 7090 in late 1958 was perceived as the dominant fact in scientific computing. Trade press treated Norris’s outfit as one more upstart.

CDC 1604 Launch 1960

Announcement date: 16 October 1959. Confidence: HIGH (Computer History Museum).

First delivery: January 1960 – Model #1, Serial #1 – to the U.S. Naval Postgraduate School in Monterey, California, installed in Room 101A of Spanagel Hall. Cray personally supervised the installation. Confidence: HIGH (NPS official history).

Pricing: approximately $1 030 000 per system (1964 figure, with 192 KB of memory). Earlier configurations were listed at “around one million dollars”. The user brief’s “$1 million” is in the right ballpark. Confidence: HIGH.

Why Monterey first: Cray lobbied for it. NPS was a credible academic-military customer; a successful Monterey install would be visible across the Navy. The Bureau of Ships had contracted for ten 1604 systems in 1958; NPS got Serial #1, the rest went to Navy operational control centers. Confidence: HIGH (CHM “Weather by Computer”; NPS 70 Years).

Production figures: “By 1964, over 50 systems had been built” (Wikipedia CDC 1604).

Variants: 1604 (original), 1604A, 1604B (small upgrades). The CDC 924 (1961, $180 000) was a 24-bit cut-down version for smaller customers; NASA was a 924 customer.

1604 Customer List

What I can verify, ranked by confidence:

High confidence

  • U.S. Naval Postgraduate School, Monterey – Serial #1, January 1960. Used by Navy meteorologists; this 1604 became the operational machine of the Fleet Numerical Weather Facility founded in 1959, which is the relevant detail for an NWP-history blog post.
  • U.S. Navy Bureau of Ships – 10 units total contracted in 1958, including operational control centers (e.g. FOCCPAC, Fleet Operations Control Center, Pacific in Hawaii). FOCCPAC ran a cluster of CDC 160As that supervised four 1604s.
  • Pentagon / DASA (Defense Atomic Support Agency) – one 1604 used during the Cuban Missile Crisis (October 1962).
  • Marathon Oil Company, Findlay, Ohio – 1604 used for a text-mining application called Masquerade in 1960.
  • University of Minnesota – 1604 installed alongside the surviving ERA 1103. Confidence: HIGH (UMN Archives).
  • Minuteman I ICBM ground stations – two redundant 1604 systems per station, late deployment. Confidence: HIGH (Wikipedia CDC 1604).

Medium confidence

  • East Germany – a 1604 was discovered in working condition in East Germany after reunification in 1991. Reported in CDC’s Contact magazine and on Philipp Hachtmann’s CDC history site. The route by which a 1604 reached the GDR is undocumented in open sources. Confidence: MEDIUM on the discovery; LOW on how it got there.

Items the user brief speculated about but I could NOT corroborate

  • CERN (Geneva) – No CDC 1604 at CERN. CERN computing in the 1604 era was IBM 709 (1961), then IBM 7090 (1963). CERN’s first CDC machine was the CDC 6600 in January 1965, then a CDC 3800 in October 1966. Confidence: HIGH that CERN never had a 1604 (CERN IT timeline).
  • University of Illinois, University of Wisconsin, University of Pennsylvania – I found no specific evidence of 1604s at these institutions. They had other contemporaneous machines (ILLIAC, etc.). Treat as not-confirmed for the post.
  • Naval Research Laboratory – plausible given Navy contract pattern but I could not find a specific cited 1604 installation. LOW.
  • NSA – plausible (NSA had bought ERA’s ATLAS series), but I could not confirm a specific 1604 there. LOW.
  • Lawrence Livermore National Laboratory – LLNL was a CDC customer but for the 6600 (delivered August 1964), not the 1604. I found no record of a 1604 at LLNL. Treat as not-confirmed.
  • DESY (Hamburg) – I found no 1604 record at DESY in the search window. DESY did acquire later CDC machines in the 6000 series.

Recommendation for the post: confine the named 1604 customer list to NPS Monterey, the Navy Bureau of Ships fleet, DASA / Pentagon, FOCCPAC, Marathon Oil, University of Minnesota, Minuteman I ground stations, and JINR Dubna (see next section). Resist the urge to pad the list with universities I cannot verify.

The Soviet Sale to Dubna

This is the most politically interesting part of the story and also the part where the user brief’s framing needs adjustment.

What is well attested:

  • A CDC-1604A (sometimes written CDC 1604A) was installed at the Joint Institute for Nuclear Research (JINR), Dubna, north of Moscow.
  • It was used in the JINR Computing Centre alongside Soviet machines (M-20, KIEV/MINSK-22, BESM-3M, BESM-4, BESM-6) for high-energy physics data processing – particularly bubble-chamber photo analysis.
  • The JINR developers writing the FORTRAN translator for the BESM-6 were tasked with ensuring software compatibility with CDC-1604A FORTRAN programs, indicating the 1604A was a working production machine on which user code already existed.
  • A CDC-6200 was added to the JINR complex in 1972 as a partner to the BESM-6.

Sources (HIGH confidence):

  • V. P. Shirikov, Distributed Systems for Data Handling (IFIP SoRuCom 2006 paper) – explicitly names CDC-1604A at JINR.
  • Wikipedia Talk:BESM-6 – editor Avivanov76’s note that “the USSR bought the CDC-1604 in 1968” and that BESM-6 FORTRAN compatibility with 1604 programs was a development goal.

What is less clearly attested:

  • The exact year of acquisition. Wikipedia talk-page lore says 1968. The user brief speculated 1965-1966. I could not find a primary source pinning down the year. The Shirikov paper places the broader “sequence of new computers” in the late 1960s-early 1970s. The 1968 date is the most commonly stated and the most plausible: by 1968, COCOM rules on civilian-use scientific computers were just permissive enough at the 1604’s by-then-modest performance level to make a licence achievable. Best estimate: 1968. Flag as MEDIUM.

  • The export-licence narrative as told in the user brief – “interagency battle: State (favored), Defense (opposed), Commerce (split); licence eventually granted under conditions including no military use and limited follow-on support” – this is plausible Cold War export-control practice and matches the general template of US-USSR computer-export decisions in 1968-1972, but I could not locate a specific declassified document or NYT article describing the CDC 1604 to Dubna decision. The closest documented case is the CDC 6400 to the Institute of High Energy Physics at Yerevan, considered under COCOM clearance during the Nixon administration (Heritage Foundation summary; FRUS 1969-76). The 6400 case shows precisely the pattern the user brief describes – but it is a different machine and a different end-user. Flag as MEDIUM-LOW that the same political pattern applied to the 1604 sale, and HIGH that the broader licensing context in 1968 worked roughly as described.

  • “Stolen from West Germany” claim. A user comment on Wikipedia Talk:CDC 1604 asserts the Soviets stole a 1604 from West Germany; this is unsourced talk-page lore and conflicts with the Shirikov primary account of a working CDC-1604A integrated into JINR’s documented computing infrastructure. The “stolen from West Germany” version is almost certainly Cold War legend conflating the East German 1604 (discovered 1991, see above) with the JINR 1604A. Recommendation: do not repeat the West Germany / theft story. Use the export-licence framing.

The BESM-6 connection. The Soviet supercomputer BESM-6 (entered production 1968, ITMVT/Lebedev) was not a hardware clone of the CDC 1604. It shared the 48-bit word size and similar performance class but had a different instruction set, different floating-point representation, and different index register architecture. The Soviet design team consulted translated IBM Stretch documentation more than the 1604. What was real, however, is that the BESM-6 FORTRAN system was deliberately written to run programs originally developed for CDC-1604A hardware, because by then there was a Russian user community with 1604 code. The 1604 left a software footprint in the USSR even if not a hardware lineage. Confidence: HIGH (Wikipedia BESM-6 talk; Shirikov).

Why this matters for the post. The CDC 1604 in Soviet hands is a small but precise instance of the West-East scientific computer trade between detente openings and the COCOM ratchet. It was not a smuggling story. It was a licensing story. By 1968 the 1604 was no longer near-state-of-the-art – the 6600 had shipped in 1965 – so granting an export licence to a known scientific institute (JINR was an explicitly multinational body, with Soviet, East European, and even occasional Western collaboration) was politically possible. The 6600 itself was firmly export-controlled; CDC’s later attempt to sell a 6400 to Yerevan in 1969-1970 became the actual flashpoint. Recommendation: pitch the post around the 1604 licence as the precedent that opened the door, with the 6400 / Yerevan case as the unresolved sequel that hardened the policy.

The Chippewa Falls Move 1962

By 1962, CDC headquarters in Bloomington had grown enough that Cray found the corporate environment intolerable. He demanded his own laboratory in his hometown of Chippewa Falls, Wisconsin – about 80-100 miles east of Minneapolis-St. Paul. Confidence: HIGH (Markoff obituary; Wikipedia CDC).

Norris agreed. CDC built the Chippewa Falls Laboratory. From 1962 onward this was Cray’s domain: it was where the CDC 6600 (delivered to Livermore in September 1964), the CDC 7600 (1969), and the cancelled CDC 8600 were developed. Cray ran the lab as effectively a private fiefdom. Visitors – including senior CDC executives – needed his explicit invitation. Norris is reported by Markoff to have rarely visited himself.

The “leave me alone” quote. The user brief flags the often-quoted line “If they want me to design a million-dollar computer, I want them to leave me alone.” I could not source this to a contemporaneous attribution. Flag as LOW. A better-attested Cray-Norris exchange is the five-year plan: when Norris asked Cray for a five-year strategic plan, Cray’s reply (per Markoff and multiple subsequent sources) was:

“Five-year goal: Build the biggest computer in the world. One-year goal: Achieve one-fifth of the above.”

Confidence: HIGH on the five-year-plan quote. Use this rather than the unsourced “leave me alone” line.

The decentralization principle was Norris’s, not Cray’s. Norris had observed at Sperry-Rand exactly what happens when a successful technical group is folded into a conglomerate – and his explicit organizational doctrine at CDC was to let star engineers run their own labs with minimal interference. Markoff’s obituary frames this as “Norris valued Cray’s technical vision enough to grant him operational independence.” Norris’s principle, in his own later writing (New Frontiers for Business Leadership, 1983): “tolerance for entrepreneurial failure and technical failure” combined with “zero tolerance for sloppiness.”

CDC Corporate Culture and Decentralisation

Three points worth making in the post:

  1. Decentralization was a deliberate organizational strategy. Norris had multiple semi-autonomous laboratories: the main Bloomington-Minneapolis engineering operation, Cray’s Chippewa Falls lab, smaller specialised facilities, and later international subsidiaries. Each was given budget autonomy and engineering freedom unusual for the era. This is the model that Bell Labs had pioneered for research; Norris extended it to product engineering.

  2. The “Cray Lab” was the extreme case. No other CDC engineering manager had Cray’s degree of independence. By 1968 he ran what was effectively a separate company-within-a-company in Chippewa Falls.

  3. The drift in the 1970s. As CDC grew past 25 000 employees and Norris turned increasing personal attention to social-responsibility programs (PLATO computer-aided instruction; inner-city plant siting; the prison-education initiative), the decentralization principle eroded under bureaucratic pressure. Cray’s frustration grew. He left CDC in 1972 to found Cray Research. This was, in retrospect, the moment the CDC corporate culture changed from engineer-led to manager-led – and the moment supercomputing leadership exited Minneapolis for Chippewa Falls under new corporate paper.

Successors: 924, 3600, 6600

The 1604 lineage runs cleanly through the 1960s:

  • CDC 924 (1961) – 24-bit, smaller and cheaper at $180 000. NASA was a customer.
  • CDC 1604A / 1604B – minor variants of the original 1604.
  • CDC 3600 (1963) – larger sibling, 48-bit, more memory, faster. NCAR (National Center for Atmospheric Research) installed a 3600 in early 1964 as its first machine, providing the foundation for atmospheric general-circulation modelling at Boulder. Confidence: HIGH (CISL/UCAR supercomputing history). The 3600 also shipped to several universities and research labs.
  • CDC 3800 (1965-1966) – variant of the 3600 family with newer memory.
  • CDC 6600 (delivered September 1964 to LLNL; serial #7 to NCAR in December 1965) – the breakthrough. Three million instructions per second; ten times the IBM 7090. Cray’s 34-engineer team in Chippewa Falls had built the world’s fastest computer. The famous IBM internal memo from Thomas Watson Jr complaining that “thirty-four people, including the janitor” had outpaced his 1100-engineer effort dates from this period.

The architectural step from 1604 to 6600 is enormous (RISC-style instruction set, ten parallel functional units, peripheral processors), but the institutional step is simple: the same man designed both, in the same Wisconsin lab, three years apart.

Sources

Primary / archival

  • Charles Babbage Institute, University of Minnesota – Control Data Corporation Records; William C. Norris papers (Collection 279); Erwin Tomash collection. https://archives.lib.umn.edu/repositories/3/resources/279
  • Computer History Museum – William Norris oral history (CHM Oral Histories collection); CDC 1604 catalogue entry; Selling the Computer Revolution CDC brochure series.
  • Hagley Museum and Library – Sperry Rand Corporation, ERA Division records.
  • University of Minnesota Archives – Engineering Research Associates – Remington Rand – Sperry Rand records (Collection 292).
  • Wilson Center, Foreign Relations of the United States (FRUS) 1969-76, vol XIII – documents on US-USSR computer-export decisions including the CDC 6400 / Yerevan case (the closest cousin to the 1604 / Dubna question).

Secondary – canonical

  • Charles J. Murray, The Supermen: The Story of Seymour Cray and the Technical Wizards Behind the Supercomputer (Wiley, 1997). The standard popular history.
  • Arthur L. Norberg, Computers and Commerce: A Study of Technology and Management at Eckert-Mauchly Computer Company, Engineering Research Associates, and Remington Rand, 1946-1957 (MIT Press, 2005). Academic; covers the 1946-1957 backstory in detail.
  • William C. Norris, New Frontiers for Business Leadership (Dorn Books, 1983). Norris’s own account.
  • IEEE Annals of the History of Computing – multiple CDC articles, including “Control Data Corporation: the Norris era” (vol. 17 no. 4, 1995).
  • V. P. Shirikov, “Distributed Systems for Data Handling,” IFIP SoRuCom 2006 – naming CDC-1604A at JINR Dubna.

Web sources consulted

Items the post must hedge or omit

  • Norris’s $75 000 personal investment – LOW confidence, no primary citation. Hedge.
  • CERN as a 1604 customer – INCORRECT. CERN had IBM 7090 then jumped to CDC 6600 in 1965. Do not include.
  • University of Illinois, Wisconsin, Pennsylvania as 1604 customers – not corroborated. Do not include.
  • Soviet sale year 1965 or 1966 – almost certainly 1968 by the most plausible Wikipedia talk-page lore, but no primary citation. Hedge.
  • “Stolen from West Germany” Soviet 1604 story – Wikipedia talk-page rumour, conflicts with the Shirikov primary account. Do not include.
  • “They want me to design a million-dollar computer, leave me alone” quote – LOW confidence. Use the five-year plan quote instead, which is well attested.
  • “We cannot ship anything” Sperry quote – MEDIUM confidence as Cold War CDC folklore but no primary attribution. Use cautiously.
  • Detailed export-licence interagency battle for the 1604 / Dubna decision – I could not find a specific FRUS or NYT cite. The general 1968 COCOM environment is well documented, but the specific 1604 / Dubna decision papers I did not surface. Either (a) frame the licence story generically with the 6400 / Yerevan case as the documented analogue, or (b) hedge that “specific declassified records on the 1604 / Dubna decision are scarce in open sources.”