Allison J Parent
Your unaware of many systems of the 60s that were
large on performance
and transistors. Your one transistor per cubic inch is far to low.
I've worked with cordwood modules that were more in the 6 per cubic
inch. Even flat boards were fairly dense.
No, one transistor per cubic inch is on the high side (as I describe
in a previous posting). Sure, you can make cards that have a dozen
discrete transistors per cubic inch (without even going to SOT-23
packages).
But you can't build a *system* with that overall density.
IBM didn't achieve densities better than one transistor per cubic
inch until they switched from SMS cards (discrete components on small
printed circuit board modules, similar to DEC R and B series modules)
used in the 7000-series computers to SLT (hybrid integrated circuits on
ceramic substrates, first used in System/360 in 1964).
References:
_IBM's Early Computers_
Charles J. Bashe, Lyle R. Hohnson, John H. Palmer, Emerson W. Pugh
MIT Press, 1986 (out of print) ** This book is a must-have **
_IBM's 360 and Early 370 Systems_
Emerson W. Pugh, Lyle R. Johnson, John H. Palmer
MIT Press, 1991
_Memories That Shaped an Industry_
Emerson W. Pugh
MIT Press, 1984
_Building IBM: Shaping an Industry and Its Technology_
Emerson W. Pugh
MIT Press, 1995
_Planning a Computer System: Project Stretch_
Werner Buchholz, editor
McGraw-Hill, 1962 (out of print)
_Computer Architecture: Concepts and Evolution_
Gerrit A. Blaauw and Frederick P. Brooks Jr.
Addison-Wesley, 1997