Jump to content

Seymour Cray

From Alfawiki
Seymour Cray
Born
Seymour Roger Cray

Template:Birth date
Chippewa Falls, Wisconsin, U.S.
DiedOctober 5, 1996(1996-10-05) (aged 71)
Occupation(s)Electrical engineer, computer architect
Years active1950–1996
Known forArchitect of supercomputers, founder of Cray Research

Seymour Roger Cray (September 28, 1925 – October 5, 1996) was a pioneering American electrical engineer and architect of supercomputers who developed a succession of the fastest machines globally and established Cray Research. Often hailed as "the father of supercomputing," Cray is recognized for founding the supercomputer sector. Joel S. Birnbaum, former Hewlett-Packard CTO, remarked that numerous features now common in high-performance computing were once considered highly ambitious when Cray first envisioned them. Larry Smarr, former head of the National Center for Supercomputing Applications, referred to him as "the Thomas Edison of the supercomputing field."

Early life

Cray was born in Chippewa Falls, Wisconsin, where his father’s profession as a civil engineer inspired him. Demonstrating an early talent for engineering, he constructed a device at age ten that translated punched paper tape into Morse code. His family’s basement served as his personal workshop.

Following his graduation from Chippewa Falls High School in 1943, Cray served as a radio operator during World War II, participating in operations across Europe and the Pacific focused on breaking Japanese naval codes. He subsequently earned a Bachelor of Science in electrical engineering in 1949 and a Master of Science in applied mathematics in 1951 from the University of Minnesota.

Engineering Research Associates

In 1950, Cray began working at Engineering Research Associates (ERA) located in Saint Paul, Minnesota. ERA originated from a former United States Navy laboratory dedicated to codebreaking devices and undertook various engineering endeavors. Cray gained distinction as a digital computing expert by designing the ERA 1103, recognized as the first commercially successful scientific computer. He remained with ERA through its acquisitions by Remington Rand and later Sperry Corporation, where ERA became part of the scientific computing division within the UNIVAC enterprise.

Control Data Corporation

Unhappy with ERA’s direction under Sperry Rand management, Cray joined forces with William Norris to establish Control Data Corporation in 1957. By 1960, he had completed the design of the CDC 1604, an enhanced and cost-efficient iteration of the ERA 1103 that delivered strong performance relative to its price point. While CDC began production of the 1604, Cray shifted his attention toward other projects. Initially involved with the CDC 3000 series, which targeted business data processing applications, he preferred focusing on developing high-performance computers rather than lower-cost commercial systems. After starting work on the 3000 series, he delegated its completion and concentrated on creating the CDC 6600.

Although not built with cutting-edge hardware components, the CDC 6600 was engineered to maximize overall speed by optimizing system-wide performance factors such as input/output bandwidth to avoid processor bottlenecks. Cray famously observed, "Anyone can build a fast CPU. The trick is to build a fast system." The CDC 6600 became the first commercially available supercomputer, outperforming all competitors by a wide margin despite its significant cost. It surpassed rivals like IBM’s IBM 7030 Stretch, which encountered difficulties due to unresolved technical challenges such as "imprecise interrupts." To overcome this, Cray integrated ten peripheral processors—small computers dedicated to managing all I/O operations—thereby eliminating conventional I/O interrupts and improving efficiency. Its successor, the CDC 7600, achieved speeds approximately five times faster.

In a 1963 Business Week article introducing the CDC 6600, Cray expressed a principle that is often mistakenly attributed to Grosch's law:

CDC's Chippewa Falls Facility

During this timeframe, Cray became increasingly discontented with the interference from CDC's management. He insisted on a tranquil workspace with minimal managerial involvement; however, as CDC grew, numerous middle managers frequently disrupted his work, often leveraging him to impress potential clients.

To preserve his concentration, Cray opted to move from St. Paul back to his hometown, Chippewa Falls, where he possessed property. This relocation, endorsed by Norris, aimed to minimize distractions while still permitting attendance at board meetings. Additionally, concerns about nuclear conflict influenced the decision; Cray's residence near the new laboratory featured a substantial fallout shelter.

The Chippewa Falls lab was inaugurated midway through the CDC 6600 development project, which was ultimately delivered successfully. Subsequently, the CDC 7600 was developed there, achieving peak performance levels ten times greater than the 6600. However, the subsequent CDC 8600 project encountered major obstacles and effectively ended Cray’s prosperous tenure at CDC in 1972.

Despite the success of both the 6600 and 7600 models, their development nearly led CDC to financial ruin. The 8600 project faced similar difficulties; Cray wished to pursue a fresh start, but Norris preferred concentrating on the STAR-100 initiative. Unwilling to operate under these limitations, Cray departed from CDC.

Cray Research

The separation occurred on good terms; within a year, Cray established Cray Research on the same Chippewa property, with Norris contributing $250,000 as initial funding. Research and development remained in Chippewa Falls, business operations were based in Minneapolis, and unlike CDC, manufacturing was also situated in Chippewa Falls.

Initially uncertain about its strategic direction and facing financial hurdles, Cray Research gained substantial backing from Wall Street due to Seymour Cray’s esteemed reputation. This support facilitated the creation of their inaugural product, the Cray-1, which launched in 1976.

The Cray-1 prioritized balanced system performance over mere processor speed and surpassed nearly all competitors, including CDC’s STAR-100. Only the specialized ILLIAC IV matched its capabilities under limited scenarios. Serial number 001 was loaned to Los Alamos National Laboratory in 1976; the first complete unit sold for $8.8 million to the National Center for Atmospheric Research. Initial sales forecasts anticipated around twelve units sold; however, over eighty units were ultimately purchased, generating significant profits for Cray Research and earning Seymour Cray the moniker "The Wizard of Chippewa Falls."

Subsequent achievements proved more demanding. While Cray developed the Cray-2, other teams introduced the dual-processor Cray X-MP and later a quad-processor variant—both commercially successful. Released after six years of development, the Cray-2 offered only slight speed gains over the X-MP due to its sophisticated memory architecture but experienced lower sales because its intricate pipeline complicated programming efforts.

As work commenced on the Cray-3, Cray again encountered distractions from daily management tasks and resigned as CEO in 1980 to concentrate exclusively on design as an independent contractor. In 1988, he moved the Cray-3 project to Colorado Springs.

By 1989, challenges with the Cray-3 reemerged amid budget limitations and competing priorities. Company leadership decided to focus on enhancing an upgraded X-MP model utilizing high-speed memory from the Cray-2 platform, resulting in the commercially successful release of the Cray Y-MP.

Cray Computer Corporation

The Colorado Springs laboratory was separated from Cray's main operations to establish Cray Computer Corporation, which assumed responsibility for the Cray-3 initiative.

The Cray-3, operating at 500 MHz, marked the second significant setback for Cray. To realize a tenfold boost in performance, Cray opted to construct the system using gallium arsenide semiconductors, diverging from his traditional reliance on established technologies. This approach necessitated the in-house development of every element, including the semiconductor chips themselves.

Despite numerous obstacles, the team successfully delivered the initial Cray-3 unit to NCAR on May 24, 1993. The system functioned primarily as a prototype during its installation phase for debugging purposes. By this time, multiple massively parallel computing platforms had entered the marketplace, offering better cost-to-performance ratios that surpassed those of the Cray-3. In reaction to this competition, Cray launched the design of the Cray-4, aiming for a 1 GHz clock speed and enhanced power output regardless of expense.

Nevertheless, no additional Cray-3 units were sold by 1995, and with the Cold War's conclusion diminishing demand for top-tier supercomputers, prospects for the Cray-4 appeared grim. Consequently, Cray Computer Corporation exhausted its financial resources and filed for Chapter 11 bankruptcy protection on March 24, 1995.

SRC Computers

Initially skeptical of massively parallel architectures, Cray maintained that a single high-speed processor outperformed many slower processors working in parallel. However, by the mid-1990s, improvements in compiler capabilities and evolving market requirements challenged this perspective. To address these changes, Cray founded SRC Computers, focusing on creating a massively parallel system optimized for communication and memory efficiency—critical limitations in parallel computing designs. Following Cray’s untimely death in an automobile accident, SRC Computers continued advancing these technologies with a specialization in reconfigurable computing.

Technical approaches

Two core design philosophies guided Cray’s engineering: efficient thermal management and accurate signal timing.[1] His machines incorporated sophisticated cooling techniques such as integrated coolant channels within mainframe structures and thermal interfaces connected directly to circuit boards. Additionally, safety interlocks were employed to ensure operation only occurred when cooling systems were fully functional, thereby protecting hardware components.

To mitigate clock skew, Cray meticulously matched all signal path lengths electrically by fine-tuning circuit board trace layouts. He applied principles derived from Maxwell's equations to factor in radio frequency phenomena that could alter signal speed and effective path length.

Notably, rather than relying on advanced CAD tools, Cray preferred working with basic materials—utilizing faintly ruled 8½″ × 11″ graph paper for his designs.[2]

List of computers

Cray contributed to the development of the following systems:

Personal life

In 1947, Cray wed Verene Voll; they had three children before divorcing around 1978. Voll was a nutritionist and daughter of a Methodist minister—a background shared with Cray’s mother. Later, he married Geri M. Harrand. Among his descendants is LGBTQ rights advocate Andrew Cray, his grandson.

Known for shunning the spotlight, Cray pursued hobbies such as skiing, windsurfing, tennis, and notably excavating a tunnel beneath his residence—a project he credited with sparking creative insights. Former Cray Research CEO John A. Rollwagen recounted stories illustrating Cray’s unconventional management style and preference for focusing on ambitious objectives rather than detailed progress reports.

Cray passed away due to injuries sustained in a rollover accident near the United States Air Force Academy in Colorado on October 5, 1996, shortly after celebrating his 71st birthday.

Posthumous Recognition

Since 1997, the IEEE Computer Society has presented the Seymour Cray Computer Engineering Award to acknowledge groundbreaking achievements in high-performance computing systems that embody the pioneering spirit and influence of Seymour Cray within the field.

See also

References

External links

  1. ↑ Customer presentation by Seymour Cray, c1979
  2. ↑ Science, 1978
Retrieved from "https://www.alfawiki.org/index.php?title=Seymour_Cray&oldid=22059"