Abstract

The rate of growth in modern CMOS technology seems to be matched only by our ability to translate it into more and more ambitious microprocessor chip designs of higher and higher density and design costs. Do these designs match conventional wisdom? How effective are these designs? How efficient are they? The first two can be answered by looking at the ever increasing benchmark performance numbers. The third, however, deals with how well we are extracting this performance from the silicon and design investments we are making in their fabrication, and is not a subject of intense scrutiny today. This paper attempts to initiate such a discussion by looking at our "historical records," that is the huge number of data sheets on different microprocessor designs, and drawing some relatively straightforward, but very enlightening, conclusions from them. 1.0 Introduction The microprocessor began with chips containing a few thousand transistors running at a few hundred kilo Hertz. Today we have ...

Citations