Abstracts - faqs.org

Abstracts

Engineering and manufacturing industries

Search abstracts:
Abstracts » Engineering and manufacturing industries

Expert opinion: traditional mainframes and supercomputers are losing the battle

Article Abstract:

The days of traditional emitter-coupled logic (ECL)-based mainframes and supercomputers may be numbered, as system architectures based on multiple CMOS and bipolar CMOS processors offer greater performance and lower cost than ECL-based systems. The first two quarters' worth of sales of Digital Equipment Corp's VAX 9000 mainframe in 1991, for example, were less than one-tenth of original forecasts. Conversely, the massively-parallel Thinking Machines CM-5 supercomputer, which employs single-instruction, multiple data (SIMD), multiple instruction, multiple data (MIMD) and vector processing in a single unit, may be a landmark in the new trend. The CM-5 can be configured with 32 to 16,384 processing nodes, each based on a standard reduced-instruction-set computer (RISC) processor, for prices ranging from about $1 million to over $100 million. The CM-5 reflects a move to defining classes of computers by the number of CMOS or BiCMOS processors included in the hardware.

Author: Patterson, David A.
Publisher: Institute of Electrical and Electronics Engineers, Inc.
Publication Name: IEEE Spectrum
Subject: Engineering and manufacturing industries
ISSN: 0018-9235
Year: 1992
Semiconductors and related devices, Product information, Design and construction, Mainframe computers, Trends, Outlook, Mainframe Computer, Computer Design, CMOS, Thinking Machines Corp., Massive Parallelism, BiCMOS, Thinking Machines CM-5 (Supercomputer)

User Contributions:

Comment about this article or add new information about this topic:

CAPTCHA


Massive parallelism rivals supercomputers

Article Abstract:

Visualization is a new supercomputer characteristic that lets a user instantly see the results of a computation in graphical form. A single-user workstation capable of visualization is priced at $100,000. Visualization is expanding into higher-level computers. Multiple instruction, multiple data architecture emerges as the standard for parallel processing machines, although a few companies offer single-instruction, multiple-data, massive parallelism machines. There is a growing need for high-speed networks to tie high level computers together.

Author: Wallach, Steven
Publisher: Institute of Electrical and Electronics Engineers, Inc.
Publication Name: IEEE Spectrum
Subject: Engineering and manufacturing industries
ISSN: 0018-9235
Year: 1989
Parallel processing, Critique

User Contributions:

Comment about this article or add new information about this topic:

CAPTCHA



Subjects list: Supercomputers, Supercomputer
Similar abstracts:
  • Abstracts: Value stream mapping: this tool puts you and your customers on the same page
  • Abstracts: Redeveloped algorithms wring more out of supercomputers. Looking Ahead in Space. The Challenge to U.S. Supercomputers
  • Abstracts: Parallelization of an event driven simulator for computer systems simulation. Solartron Introduces New Marine Radar Simulator
  • Abstracts: Simulating a JIT/kanban production system using GEMS. Modeling Human Operators on a Microcomputer: A Micro Version of SAINT
  • Abstracts: A prototype knowledge-based simulation support system. Visual interactive fitting of bounded Johnson distributions
This website is not affiliated with document authors or copyright owners. This page is provided for informational purposes only. Unintentional errors are possible.
Some parts © 2025 Advameg, Inc.