Disruptive Compute Alternatives to COTS

Abstract

In recent decades, the defense community has largely adopted a COTS (commercial off-the-shelf) approach to high performance electronic and computerized weapons systems. The reasons were compelling, as the scale of investment of commercial industry was able to drive innovation in semiconductor process, architecture, IP and software eco-systems more quickly than custom defense programs, which has resulted in higher compute density and system capability using COTS devices. Furthermore, the volumes typical in defense programs do not generally provide the economies of scale for a custom chip development, with the required high engineering design and verification effort, as well as high manufacturing costs. Custom chip costs and transistor counts continue to increase at an exponential rate as semiconductor processes advance. By definition, COTS semiconductor products are used in commercial applications, and so can be easily obtained by adversarial nations. A major drawback of COTS is this technology is also in use in our adversarys weapons systems, which may even adapt the latest semiconductor products at a more rapid rate than US defense programs. Another trend is the near ubiquitous fabless semiconductor model, where semiconductor companies design their chip products, but outsource manufacturing to a small number of foundries which provides the economy of scale for large investments necessary. This fabless business model has also supported the creation of many IP companies, which have created a wide variety of innovative architectures and tools which can be customized for specific applications. Many of these smaller companies are also more nimble, and can focus on optimizing a solution for specific problems, in contrast to large semiconductor vendors which generally are developing products for widest possible application. An alternative to COTS semiconductors is to develop heterogeneous, programmable ASICs, which could be used for multiple defense applications in

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 25, 2019
Accession Number
AD1075600

Entities

People

  • Michael Parker

Tags

Communities of Interest

  • Advanced Electronics
  • Cyber
  • Electronic Warfare
  • Energy and Power Technologies
  • Materials and Manufacturing Processes
  • Weapons Technologies

DTIC Thesaurus Topics

  • Algorithms
  • Artificial Intelligence
  • Communication Systems
  • Computer Programming
  • Computer Vision
  • Computers
  • Deep Learning
  • Digital Signal Processing
  • Dynamic Range
  • Frequency Domain
  • Image Processing
  • Machine Learning
  • Multiple Input Multiple Output
  • Neural Networks
  • Radar
  • Semiconductors
  • Signal Processing

Readers

  • Government Contracting/Procurement.
  • Integrated Circuit Design and Technology.
  • Systems Analysis and Design

Technology Areas

  • Microelectronics