ON LIMITS IN COMPUTING POWER
Abstract
At one time or another you have probably all heard of the growth figures quoted for the computing industry in the double decade of1955-1975; these figures are part history and part extrapolation, but to the extent that history has progressed since the estimates were made, the extrapolations are valid. In these twenty years the size of the computer has decreased 10,000-fold for equal computational capability. The unit cost of calculation is down by the startling figure of 200,000-fold, while speed has increased 40,000-fold. Also, there has been an explosive growth of installed capacity, which over the double decade of 1955-1975 has increased 160,000-fold. The T70s have been extensively analyzed and projected, and by 1975 or so machines ought to operate close to 10 operations per second. This morning I thought it would be more exciting to move on into the T80s to see what limits might set a ceiling on computational capability. These thoughts do not reflect original research on my part; rather I have tried to extrapolate from the work of others. There is not universal agreement about the arguments on which I draw, so my conclusions must be considered as "ballpark" guidelines. Although I recognize that we can conceivably get increasing capability from software improvements, or from better numerical analytic techniques and better mathematics, I want to avoid these issues today and talk about (1) the hardware, particularly with respect to component speeds and the limitations imposed by the laws of physics; (2) the logical arrangements used to implement arithmetic; and (3) the overall machine architecture.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 14, 1969
- Accession Number
- AD0694745
Entities
People
- Willis H. Ware
Organizations
- RAND Corporation