8 May
2001
8 May
'01
2:09 a.m.
A photomicrograph is a picture of the die, in this context. Looking at a map of
the device, in the form of an aerial photograph, which is what this is, gives a
general notion of the relative size of the structures on the die, and, in some
cases, allows for recognition of their functions. I wouldn't bet my paycheck on
guessing what a given structure is, just based on its apppearance on the typical
photomicrograph, however.
Intel used posters consisting of photomicrographs of their products, back in the
late '70's-early '80's in their marketing effort. Some of these were
sufficient, if one collected them all, to allow for claims of reverse
engineering some of their prominent and popular parts, e.g. the '186, and the
8751. It was certainly possible to recognize some structures, like EPROM cells,
RAM cells, etc, based on their geometry. I suppose one could have learned a
great deal with enough time and effort, along with a complete, layer-by-layer,
set of pictures. However, I rather imagine that folks smart enough to
reverse-engineer the devices from the pictures, probably could have done the job
without them.
Dick
----- Original Message -----
From: "Jim Battle" <frustum(a)pacbell.net>
To: <classiccmp(a)classiccmp.org>
Sent: Tuesday, May 08, 2001 12:54 AM
Subject: Re: How many transistors in the 6502 processor?
At 10:48 AM 5/6/01 -0600, you wrote:
I doubt it was really 68K transistors. If you
multiply out the number of
transistors needed to implement the registers and counters, then add 50%
for the
ALU and double the whole thing for the control logic, you'll probably have as
good a count.
Don't forget that the 68K was microcoded; lots and lots of transistors were
used just for that. In fact, I recall reading that it was described as
"68000 potential transistors" because otherwise the transistor count would
have been dependent on the number of 1's or 0's in the microcode store.
I found transistor counts for a few other processors from a book I have on
my shelf:
MC6800 -- 6000, 57.5%
MC68000 -- 60000, 69%
IAPX 432 -- 150000, 84% (two chips)
MC68020 -- 192000, 70%
IAPX 386 -- 275000, 70.5%
(% is portion of chip dedicated to control logic [vs datapath])
From "The Architecture of Microprocessors", Fracois Anceau, p. 107
This book is in large part based on reverse engineering microprocessors
using a microphotographs (or is that photomicrographs?)
-----
Jim Battle == frustum(a)pacbell.net