Computer History Exhibits

1
Computer History Exhibits Signs and
Placards master copy on Haring
First floor Stanford CSD history Basement
Technology timelines Floor 2 Early computing
Floor 3 The sixties Floor 4 The seventies Floor
5 Galaxy game
2
Computer History Exhibits
case f 2
case f 1
case2
case3
case1
case4
case5
Basement Timelines
First floor Early Stanford CSD history
2nd 50sUnivac Whirlwind 3rd 60s IBM 360
DEC PDP-6 4th 70s Aple II Cray
Fifth floor Galaxy game
A joint project of . Stanford Faculty,
Staff, The Computer Museum History Center
Questions to Gio_at_cs or look at
http//www-cs.stanford.edu (museum)
3
Computer History Exhibits
Opening Talks in room B1, Nov. 5th, 530
pm Donald Knuth George Forsythe and the
Development of Computer Science Gordon Bell
Values Issues in Preserving Historical
Computer Artifacts
Serra street
Exit to outside
B1
Entrance to Basement Lecture Hall
First floor
Basement
Campus Drive
4
(No Transcript)
5
Platter from General Precision Librascope L 4800
head-per-track Disk Unit Stanford AI Lab DEC
PDP-6, November 1967 Storage capacity per side
ca. 1,120,665 words of 36 bits Capacity per unit
(10 inner sides of 6 platters)
11,206,650 words or ca. 48 M bytes. Total 5484
heads (and tracks). Total weight 5200
lbs Rotational speed 900 rpm, Avg. access time 35
msec. Transfer rate 1.6 m sec/word or 2.7 M
byte/sec Startup current 300 amps, Startup time 5
minutes, thermal stabilization 2 hours Cost
300,000 (1,420,000 today) The photograph shows
the unit with the disks and the electronics bay
(2000 lbs) removed.
Courtesy of Martin Frost
6
Total Tracks (and Write-Read heads) 5484
(includes 300 spares) Bits/Track 80,256
Bits/Sector 66 Sectors/Rev 1216 based on CPI
1997 159.1 159.6 160.0 160.2 160.1 160.3
160.5 1967 32.9 32.9 33.0 33.1 33.2 33.3
33.4 33.5 33.6 33.7 33.8 33.9 33.4
7
SONY Corporation 3.5 Floppy disk drive
ca. 1991 High density, double sided
one head per side Capacity/floppy 1.4
Megabytes With disk in protective,
low friction carrier.
Courtesy of SUMEX
8 Floppy disk first use ca. 1965 Single
sided disk Capacity/floppy ca. 150 Kilobytes
Courtesy of Vaugn Pratt
8
SONY Corporation 3.5 Floppy disk drive
ca. 1991 High density, double sided
one head per side Capacity/floppy 1.4
Megabytes With disk in protective,
low friction carrier.
Courtesy of SUMEX
5 Floppy disk drive Shugart
Corporation first use ca. 1977 Single sided
disk Capacity/floppy 360 Kilobytes
Courtesy of Vaugn Pratt
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
Polya Hall Home of the Stanford
Computer Science Department 1963- Oct.1979 Named
for George Pólya (1887-1985) Prof. of
Mathematics
Polya 13 Dec.1887- 7 Sep.1985 Don Knuth
15
(No Transcript)
16
IBM Card Programmed Calculator (CPC) A CPC was
Stanfords computer from 1953-1956. The tall box
is the arithmetic unit, which used 1500 vacuum
tubes and had 8 registers of 4 digits and 1
register of 5 digits. Digits were represented by
4 bits each, requiring 2 vacuum tubes per
bit. The box on the right contained 4 mechanical
accumulators of 12 digit words and 2 of of 16
digits, and 48 words of mechanical storage.
Mechanical storage was implemented in the form of
wheels, which were positioned by solenoids, and
had contacts for readout. Instructions were read
from cards, placed into the center unit, at a
rate of up to 150 per minute. Through wiring a
plug board placed in the arithmetic unit certain
cards could be skipped, giving some control over
program flow. The CPC was not yet a von
Neuman machine architecture. The central unit
also had a printer, which could print 120 columns
of numeric output at 150 lines per minute (lpm),
but only 40 columns of letters at 100
lpm. Results could also be punched on the
rightmost unit, on up to 50 cards per minute.
Another wiring board selected the card columns.
17
(No Transcript)
18
Data processing cards were invented by Hermann
Hollerith of the U.S. Bureau of the Census.
Commonly known as IBM cards they were used for
data and program storage from 1890 up to the
1980s. They had 80 columns, and up to 4 holes
out of 12 positions could be punched out per
column, allowing first 12, later 64, and
eventually 256 distinct characters codes per
column. More holes weakened them. The size of
the card was based on the dollar bill of that
time, so that they might be carried in standard
wallets. Dollar bills are now smaller in size
and in value. Silver certificate dollar bill
from 1920 courtesy of Voy and Gio Wiederhold
19
Early Computers at Stanford
Type arrived-retired Location speed(/x)
Memory Prim.language msec
Words/bytes IBM CPC Mar.1953-56 Elec.Lab.
760K/13M 48 wired board IBM
650 Jan.1956-62? Elec.Lab. 2.2K/19K 2KW
SOAP Burroughs 220 Jun.1960 Encina
200/3300 10KW Balgol shared
with First National Bank of San Jose (overnight
check processing) IBM 7090 Feb.1963?-67 Pine
Hall 4.4/25 32KW
FORTRAN Burroughs 5500 Mar?.1963-68 Pine Hall
Algol DEC PDP-1 1964 - Pine Hall
5/18bits 64KW DEC PDP-6 Aug.1965 AI lab
4/36bits LISP
IBM/360-50 Jun.1965 SLAC 4/16
256Kb IBM/360-50 Dec.1965-7x Med.Sch. 4/16
1.128Kb PL/1 subset IBM/360-67 May
1967- Pine Hall 1.5/6 500Kb Algol
W, installed as an IBM/360-65 because of an
inadequate timesharing system
FORTRAN IBM/360-75 SLAC 0.75/3
1Mb FORTRAN IBM/360-91 1968 SLAC 0.2/0.4
2Mb FORTRAN DEC PDP-10 1969?
-85? AI lab LISP, SAIL DEC system
2040 1976-1977 LOTS 1.0
128KW DEC system 2050 1977-19
LOTS 0.5 256KW
20
Early Faculty at Stanford 1953 Jack Herriot,
Alan Peterson, codirectors computation center
21
Remington-Rand Univac Flip-Flop Assembly Model
1818A, serial 001348. Manufd for the U.S. Navy,
Oct.1960.
Courtesy of David Hermreck, Potomac, MD. Two?-bit
highly reliable plug-in electro-mechanical memory
unit. It uses relays, composed to form flip-flop
storage cells, similar to the exposed AEC unit
shown. The access time was about 1/2 sec. To
avoid corrosion, all joints were soldered to be
airtight, and then the unit was filled with
nitrogen gas, through the valve on the side.
All contacts are gold plated.
Similar flip-flop units, but not sealed, were
used for the IBM CPC (Card-Programmed Calculator)
shown above, used at Stanford from1953 to 1956.
The CPC could hold 9 words of 4 4-bit digits in
vacuum tube circuits, and 48 words of 10 digits
in relay storage. The CPC was hence not a
von-Neumann machine architecture programs
remained external. Computation was driven by sets
of cards, fed through a card reader at up to 2.5
instructions/second.
22
Primary Programming Languages Taught at
Stanford ltTentative Draft, tell us what you
knowgt Language years compiler ma
chine Board wiring 1953-56 none IBM CPC
Assembler 1956-60 SOAP II IBM 650 Algol
58 1960-65 Balgol Burroughs 220 FORTRAN 1963-6
7 FORTRAN II IBM 7090 Algol 60 1963-68 Algol
Burroughs 5500 Algol W 1968-75
Wirths IBM/360 FORTRAN 1975 FORTRAN
IV IBM/370 ALGOL 60 1976-77 SAIL DEC
10 PASCAL 1978-91 LOTS DEC-10 C 1991-today
Apple Macintosh Java? future? Information
courtesy of Claire Stager, Eric Roberts, ...
23
(No Transcript)
24
DEC-10 system Memory Controller Board,
modified for LOTS, the Stanford Low-Overhead Time
Sharing System, 1977 By 1976 semi-conductor
memory prices had dropped to the extent that
large number of display terminals could each have
their own buffer in a timeshared system. The
buffersizes were adequate for 40 lines of 80
3200 characters each, requiring about 320, 000
bytes for 100 terminals. This was more than
provided for in the original controller design,
so that boards for LOTS were modified to allow
high-order addressing. On PCs and workstations
today, the entire display image is buffered,
omitting the need for a hardware charcter
generator, but requiring up to a Megabyte per
display.
Courtesy of Ralph Gorin
25
ACME system status panel, 1966
Designed by Robert Flexer and Klaus Holtz For the
time-sharing and real-time data acquisition
system in The Medical school, ACME, status
indicators were provided on each of the 30
terminals, to reduce user frustration. The white
ACME IS ON light was pulsed periodically, so that
it would decay if the system went down. YOU ARE
ON signaled each time slice allocated. The
WAITING FOR YOU light indicated that input was
expected from the terminal or a data-acquistion
port, and the SPECIAL RUN ON light warned users
that a high demand data acquisition task was in
progress, reducing the performance for all
others.
Courtesy of Gio Wiederhold
26
(No Transcript)
27

• The SAIL language was
• derived from Algol 60,
• expanded with
• direct access to PDP-10 I/Ofacilities,
• control over external interrupts
• macro-capabilities
• sets and lists
• data structures for associative search
• multi-processing
• The last three augmenta-tions were derived from
  LEAP, developed in 1969 by
• Jerry Feldman and Paul Rovner on the Lincoln Labs
  TX-2.

SAIL User Manual June 1973 Editor Kurt
VanLehn Stanford AI Laboratory The SAIL
language was, with LISP 1.5, the primary
programming language at the Stanford
AI Laboratory, and used, a.o., for its research
in robotics.
Courtesy of Gio Wiederhold
28
DataDisc Display System 1971 The DataDisc (DD)
used the disk you see here to store and
continuously generate 32 video channels that were
used as display screens on monitors around the
Stanford AI Lab. 1972 The DD video channels
were routed through a crossbar switch to any
combination of 56 DD display terminals in
the building. Users could view the same channel
from multiple monitors, or multiple channels on
one monitor. 1982 More and more DD channels
had become very streaky and annoying, so the DD
disk was replaced with RAM memory using the big
64Kbit chips in the newDD system designed at
SAIL. Here you see the DDs small read amplifier
cards mounted around a circle. On the other
side, arranged in a spiral, are the disk heads,
which you can see in the shiny mirror in the
back, which is the DD disk itself! (Note the
dark lines on the outer portion of the disk --
from head crashes which disabled only selected
channels.) One new DD memory board, holding
four video channels, is to the right.
29
(No Transcript)
30
(No Transcript)
31
Monroe Decimal Calculator. ca. 1930 Inventor
Frank Stephen Baldwin 1839-1925. This 10-key
calculator provided accurate manual
computation. Its operator was called a computor.
Each complete forward turn of the large crank on
the right will add the value set into the 8 x 10
keys into the bottom register of the carriage.
The top register counts the turns. Subtraction
is achieved by turning the crank in reverse. To
multiply the Repeat button is pressed and the
crank turned as often as needed for the low-order
digit. Then the carriage is moved to the right
with the handle in front, so the next digit of
the factor can be cranked in. The crank on the
carriage is for resetting result and counter
registers. Division is performed by subtracting
the divisor left to right.
Courtesy of Gio Wiederhold
32
Monroe Decimal Calculator,ca.1930 Inventor Frank
Stephen Baldwin 1839-1925. This 10-key
calculator provided accurate manual computation.
Its operator was called a computor. Each complete
forward turn of the large crank on the right will
add the value set into the 8 x 10 keys into the
bottom register of the carriage. The top register
counts the turns. Subtraction is achieved by
turning the crank in reverse. To multiply the
Repeat button is pressed and the crank turned as
often as needed for the low-order digit. Then the
carriage is moved to the right with the handle in
front, so the next digit of the factor can be
cranked in. The crank on the carriage is for
resetting result and counter registers. Division
is performed by subtracting the divisor left to
right
Courtesy of Gio Wiederhold
33
(No Transcript)
34
Mathematical Tables from the Handbook of
ChemistryPhysic, 1949 Chemical Rubber Publ.
Company, Cleveland OH.
35
(No Transcript)
36
(No Transcript)
37
The Stanford Arm Stanford Artificial Intelligence
Laboratory Hand-Eye Project, 1969 The arm
contains 6 joints, and was configured to
approximate human reach, but with a different
joint structure. A pair were mounted on a table
and operated in concert with a camera, which
scanned the table surface for objects, as blocks,
which then could be stacked. Specified tasks were
then accomplished without further camera
feedback. The claw provided force feedback.
38
(No Transcript)
39
Computer History Exhibits Installation in
Progress Watch this Space
size (44, 43.5, 43.5, 45) x 42.5
Computer History Exhibits Installation in
Progress Watch this Space
40
Computer History Exhibits Installation in
Progress Watch this Space
Computer History Exhibits Installation in
Progress Watch this Space
41
Computer History Exhibits Installation in
Progress Watch this Space
Computer History Exhibits Installation in
Progress Watch this Space
42
(No Transcript)
43
Computer History Exhibits Installation in
Progress Watch this Space
Computer History Exhibits Installation in
Progress Watch this Space
44
Computer History Exhibits Installation in
Progress Watch this Space
Computer History Exhibits Installation in
Progress Watch this Space
45

• The IBM/360 architecture was to cover the
  spectrum from
• modest to large machines, and data-processing as
  well as scientific computation. The principal
  designers were
• Gene Amdahl,
• Fred Brooks, and
• Gerrit Blaauw.
• The 8-bit byte, 32-bit word
• architecture is still used in
• todays IBM mainframes.
• It influenced greatly the later
• RCA Spectra, XDS S , Ryad,
• and Univac 9000 computers,
• and to lesser extent the DEC
• VAX and Intel architectures.

Console panel from an IBM/360-40
computer Announced April 1964, first
delivered 1965. Courtesy of
The Computer Museum The table held the console
printer of the ACME system, an IBM/360-50G with
1M. later 2Mb, auxiliary memory, performing
timeshared real-time data acquisition and
computation at the Stanford Medical School.
46
CORE Memory planes from IBM/360 series IBM
Corporation, ca. 1964 Ferrite-core memories were
first developed during the early 1950s for use
in the SAGE air-defense system. Each tiny
doughnout-shaped core stored a single bit of
information (1 or 0) by means of the clockwise or
counterclockwise direction (around the hole) of
the cores internal magnetization. Tiny electric
wires strung through the core holes were used to
write and read information. Ferrite-cores soon
replaced all other computer memory technologies
because of their superior reliability and speed.
The ferrite-core memory planes shown here were
used in IBM System/360 computer beginning in
1964. A memory consisted of many core planes
interconnected with electronic red-write
circuitry. System/360 memories provided
read-write cycles of 0.75 to 2.5 microseconds and
capacities of thousand bytes to 1 million bytes.
Manufacturing costs of ferrite cores were less
than 0.1 cents each, but a fully wired core
memory with all support circuitry cost 1 to 2
cents per bit. Semiconductor memories gradually
replaced ferrite-core memories after the first
all-semiconductor memory was introduced on the
IBM System/370-145 in 1970. Courtesy of
IBM Yorktown Heights
47
(No Transcript)
48
Notes from Pugh, Johnson, Palmer pp 338
-9215x -70 p 640 total range 2001 CACM vol
221.1 1978
A single operating system was planned as well.
However, it became soon obvious that the smaller
machines would drag down the larger ones, and
64K became the minimum size for IBM-OS, smaller
machines used a system called DOS. Stanford
developed new (ACME), or augmented IBMs
operating systems (Wylbur and Orvyl).
49
Computer History Exhibits Installation in
Progress Watch this Space
Computer History Exhibits Installation in
Progress Watch this Space
50
Apple Corporation Apple I I Designed
originally 1977 Magnavox 12 b-w TV, used as
computer display The early Apple computers used
TV sets to display about 20 lines of 40
characters each.

Computer courtesy of The Computer Museum, TV
c.o. Voy Gio Wiederhold
51
VisiCorp User Guide for VisiCalc Electronic
Worksheet program, 1981. Inventor Bob
Frankston at Software Arts, Inc, 1979. All
commands were single letter codes, combined
with arrow keys and functions.
Conventional programming, languages as BASIC and
PASCAL were made available for the Apple, but had
limited acceptance. The innovative interactive
VisiCalc spreadsheet program for the Apple II
and, later, the IBM PC, transformed personal
computers to useful business tools, and greatly
broadened their market. Visicalc was in turn
replaced by Lotus, due its intuitive
point-and-click interface.
Courtesy of Gio Wiederhold
52

UCSD Apple PASCAL 1.1 Developer Kenneth Bowles
1979 Graphic extensions by
Apple Corporation. Manual by Arthur Luehmann
and Herbert Peckham,
McGraw-Hill 1981
Pascal was defined in 1972 by Prof. Niklaus
Wirth and imple- mented in 1978 with Kathleen
Jensen at the ETH in Zürich, Switzerland for
the CDC 6000. The intent was to have a clear and
effective language for teaching. Its simple
type structure was in part a reaction to the
complexity introduced with Algol 68. Pascal
became rapidly very popular and was also widely
used in commercial practice. It was the language
used for teaching at Stanford CSD from 1979 to
1991.
Courtesy of Gio Wiederhold
53
Computer History Exhibits Installation in
Progress Watch this Space
Computer History Exhibits Installation in
Progress Watch this Space
54
The Galaxy Game Bill Pitts Co., 1971 The
Galaxy game was probably the first commercial
computer game built. It was installed in the
Tresidder Union Coffee House from 1971 to 1978. A
single PDP-11 minicomputer is used to drive two
separate game screens with two players each.
Galaxy is is a reprogrammed version of
Spacewar!, which was conceived in 1961 by Martin
Graetz, Stephen Russell, and Wayne Wiitanen and
first realized on the PDP-1 at M.I.T. in 1962 by
Stephen Russell, Peter Samson, Dan Edwards, and
Martin Graetz, together with Alan Kotok, Steve
Piner, and Robert A. Saunders using PdP-1
assmbley language. It very became popular at most
Artificial Intelligence research centers and is
now available in a simulated version on the
web http//lcs.www.media.mit.edu/groups/el/projec
ts/spacewar/. The original version used 4
keyboard keys to control each of the two the
spaceships spin one way, spin the other, thrust,
and fire. Solar gravity will cause the ships to
destruct if no action is taken. The Stanford
version added three types of space no gravity,
anti-gravity, and uncharted space. Courtesy
of Bill Pitts, with assistance by Ted Panofsky.
55
The Galaxy Game Bill Pitts Hugh Tuck, 1971 The
Galaxy Game was the first commercial video game.
Installed in Tresidder Union in September 1971,
the game was quickly and enthusiastically
embraced by the Stanford community, with players
often waiting for over an hour for their next
turn. Galaxy Game is a reprogrammed version of
Spacewar!, which was conceived in 1961 by Martin
Graetz, Stephen Russell, and Wayne Wiitanen and
first realized on the PDP-1 at M.I.T. in 1962 by
Stephen Russell, Peter Samson, Dan Edwards, and
Martin Graetz, together with Alan Kotok, Steve
Piner, and Robert A. Saunders using PDP-1 assmbly
language. It very became popular at most
Artificial Intelligence (AI) research centers and
is now available in a simulated version on the
web http//lcs.www.media.mit.edu/groups/el/pr
ojects/spacewar/. Spacewar was a magical game
that captivated everyone that played it. However,
since time on the mainframe computers required to
support Spacewar was billed to users at rates of
several hundred dollars per hour, Spacewar was
usually played only by system programmers when
the mainframe was idle times like 2am! In late
1970, Digital Equipment Corporation introduced
the PDP-11 minicomputer. Finally, there was an
affordable computer with the power to run
Spacewar!. So, Bill Pitts (a recent Stanford
grad and AI alumni) and his high school buddy
Hugh Tuck formed Computer Recreations, Inc. in
June of 1971 to build coin operated Spacewar
machines. Bill, a computer hacker, did the
programming and electrical stuff, and Hugh, a
mechanical engineer, designed the enclosures.
After three and a half months of labor, Spacewar
was about to be delivered to the masses.
However, at this time (1971), the concept of
"war" was a very bad thing on campus. Astute
marketeers that they were, Bill and Hugh decided
to change the name to Galaxy Game. The first
version of Galaxy Game, packaged in a walnut
veneered enclosure, incorporated a PDP-11/20
computer, a simple point plotting display
interface, and a Hewlett Packard 1300A
Electrostatic Display. The PDP-11/20 (with 8K
bytes of core memory and an optional hardware
multiply/divide unit) cost 14,000 and the
display cost 3,000. Coin acceptors and
packaging brought the total cost to approximately
20,000. Galaxy Game was priced at 10 cents per
game or 25 cents for 3 games. If at the end of
the game your ship still survived and had some
fuel left, you got a free game. Perhaps Bill and
Hugh were not the most astute of businessmen .
56
. .
A second version of Galaxy Game, with a more
powerful display interface enabling the PDP-11 to
drive four to eight consoles, was developed to
amortize the cost of the computer over several
consoles. This version was installed in the
Coffee House at Tresidder Union in June 1972,
where it remained in operation until May 1979.
Throughout its tenure at Tressidder, Galaxy Game
was heavily used. Ten to twenty people gathered
around the machines most Friday and Saturday
nights when school was in session. After removing
Galaxy Game from Tressidder (because the display
processor had become very unreliable) the machine
was disassembled. The computer and displays were
stored in an office and the fiberglass cases were
stored outdoors for the next eighteen years.
Sometime in April 1997, Les Earnest (the former
Director of the Stanford AI Lab) received a phone
call from Bill Pitts. Bill was about to throw
away some old PDP-11 stuff, and he was wondering
if Les might know of a good home for old
computers. Les mentioned that the new Computer
History Exhibits might be interested. So, Bill
fired off a couple of emails in the direction of
Stanford and then finally, a reply! Yes, the
Computer History Exhibits would like Galaxy Game
as an operating exhibit. To get Galaxy Game
operating again would be no small feat. The call
for help went out. The biggest job would be to
build a new display processor using the original
design schematics. Ted Panofsky, who had
designed and built the display processor way back
when, soon received a call from Bill. Could Ted
please take complete responsibility for building
and delivering a fully functional display
processor in eight weeks? For free, of course.
Ted said he'd been waiting 25 years for just such
an opportunity! Yes, he would love to! So, with
Ted's generous contribution of time, energy, and
smarts, and help from Doug Brentlinger, Paul
Mancuso, and Victor Scheinman, the Galaxy Game is
back. By the way, the original display
processor's poor reliability resulted from using
early vintage Texas Instruments wire wrap IC
sockets. Ted was not the one that selected
them. Both versions of Galaxy Game were based on
the the Stanford AI Lab's PDP-10 version of
Spacewar. Galaxy Game is a faithful PDP-11
re-implementation of the AI Lab's PDP-10
Spacewar. Except, I don't seem to recall any
coin acceptors on the PDP-10 Bill Pitts,
October 29, 1997
57
(No Transcript)
58
Contributors Hector Garcia-Molina, Mark
Horowitz, Joe Oliger, Carlos Tomasi, Gio
Wiederhold We also acknowledge departmental
support for installation infrastructure
Special Thanks To Doug Brentlinger, Diane
Forsythe, John Goldschmidt, Ralph Gorin, Andrew
Kacsmar, Oussama Khatib, Jill Knuth, Verena
LaMar, Paul Mancuso, Robert Miller, Zae Ozaki,
Ted Panofsky, Bill Pitts, Victor Scheinman,
Eileen Schwappach, Marianne Siroker Organizing
Committee Zoe Allison, Gwen Bell, Les Earnest,
Martin Frost, Penny Nii, Bernard Peuto, Len
Shustek, Gio and Voy Wiederhold