ETA Systems

ETA Systems
Company type	Subsidiary
Industry	Supercomputing, Computer hardware
Predecessor	Control Data Corporation (CDC)
Founded	1983
Founder	Neil Lincoln
Defunct	1989
Fate	Operations discontinued by CDC
Headquarters	United States
Area served	High-performance computing market
Key people	Neil Lincoln (chief architect and founder), William Norris (CDC CEO)
Products	ETA-10 supercomputer series
Parent	Control Data Corporation (CDC)

ETA Systems emerged in the early 1980s as a supercomputer enterprise spun off from Control Data Corporation (CDC) with the objective of reclaiming a position in the high-performance computing arena. Although it produced the ETA-10 supercomputer, persistent financial losses led CDC’s leadership to eventually discontinue the operation.

Historical development

Following Seymour Cray’s departure from CDC in the early 1970s—prompted by disagreements over funding for the CDC 8600 project—CDC continued advancing the CDC STAR-100, while Cray independently developed the faster Cray-1, effectively sidelining CDC from the supercomputing sector.

CDC’s CEO, William Norris, concluded that restoring competitive strength required creating an independent entity free from corporate limitations. In 1983, ETA Systems was founded with a goal to deliver a 10 GFLOPS system by 1986, emphasizing nimbleness and innovation.

By April 1989, CDC terminated ETA Systems operations, retaining only minimal internal functions. At shutdown, ETA had sold seven liquid-cooled and twenty-seven air-cooled ETA-10 units. Despite achieving an industry-leading price/performance ratio and overcoming initial software obstacles, CDC exited the supercomputer market, redistributing remaining ETA units to high schools through the SuperQuest competition.

Products

The exclusive offering from ETA Systems was the ETA-10, which evolved from the CDC Cyber 205 architecture and maintained compatibility with it. Distinct from vector machines by Cray, the ETA-10 implemented pipelined memory operations and incorporated a shared-memory multiprocessor design supporting up to eight CPUs alongside eighteen I/O processors.

A notable innovation was its adoption of liquid nitrogen cooling in select models, allowing CPU cycle times near 7 nanoseconds and delivering roughly 9.1 GFLOPS—approaching but not reaching the 10 GFLOPS target. Liquid-cooled variants included the ETA-10E and ETA-10G, whereas more economical air-cooled versions marketed were the ETA-10Q and ETA-10P ("Piper").

ETA Systems prioritized expanding main memory capacity beyond what Cray offered, aiming to serve sectors requiring extensive data processing such as oil reservoir modeling and aerodynamic simulations. Nevertheless, difficulties with operating system software limited wider adoption. Purdue University notably installed UNIX on an ETA-10P; however, further development ceased due to budget cuts following CDC’s downturn.

A follow-up model targeting 30 GFLOPS performance, designated ETA-30, never materialized because of ETA’s closure.

Software

The software environment for the ETA-10 series encountered considerable obstacles. Rather than porting the CDC VSOS operating system from the Cyber 205, CDC and ETA opted to develop a new OS tailored to exploit their hardware’s capabilities fully. However, when the initial ETA-10 E units shipped in 1986, no operating system was yet operational; users had to load programs individually through an attached Apollo Computer workstation, with each run requiring a reboot.

Despite UNIX gaining popularity within supercomputing circles, ETA created its proprietary EOS operating system, which was unavailable at product launch. A UNIX System V–based OS version became accessible in 1988 after Canadian company HCR Corporation completed porting efforts, enhancing usability and convincing some customers who had delayed payment to accept delivery.

The decline of ETA was not solely attributable to operating system shortcomings. The Fortran compiler (ftn200) remained largely unchanged from that used on the CDC205 and depended on vendor-specific features (Q8* subroutine calls) at a time when code portability was increasingly important. Compiler optimization efforts lagged behind competitors such as NEC, Cray Research, and emerging minisupercomputer producers.

In general, hardware vendors during this period—including CDC—underinvested in developing robust software ecosystems comprising libraries and commercial or open-source applications critical to maintaining user bases and market presence.

Naming the firm

Neil Lincoln, the chief architect and founder of ETA, clarifies that "ETA" neither serves as an acronym nor relates to "estimated time of arrival." He deliberately distanced the company from any connection to the Basque separatist group ETA or Japan’s Eta social minority, considering Japan’s significance as a key market. Lincoln noted that the name was suggested by his sons, inspired by the sequence of letters most commonly used on a linotype machine.

An alternative explanation associates the name with Douglas Hofstadter's influential 1980 book Gödel, Escher, Bach, which highlights letter frequency patterns such as ETAOIN. Another proposed origin is that ETA represents Engineering Technology Associates, positioned as a successor to Engineering Research Associates (ERA); however, company leaders have refuted this claim.

These varied naming theories illustrate deliberate branding choices made by senior management to strategically position the firm within the competitive landscape of supercomputing.

External links