Sometime back, I posted a note about the origins of the 4004 to
comp.arch -- Stan Mazor, one of the engineers on the project replied
with his take on how the credit should be split up.
This is his list, re-formatted in HTML:
Shima of Busicom specified ~8 custom chips, but there was some
micro-coding for floating point operations using 16-digit primitive
operations like shift left, add, subtract, etc.
Hoff believed there were too many complex special chips and that
the micro engine was too high a level to be practical for Intel
to design and build and proposed an alternative architecture.
Hoff proposed a simpler architecture using 4-bit data path and
simple instructions based upon the PDP-8 model we were using at
Intel to run 'Focal' programs.
Shima and Busicom opposed the Intel architecture at first.
When I joined in Sept 1969 I was a computer designer for
more than 3 years and just worked on Symbol at Fairchild, and
did logic design on decimal floating point unit (like calculator).
Hoff and I did program 'snipets' to prove how we could code the
'4004' to do calculator functions. I suggested enhancements to
ISA including; JIN, FIN, and FIM, and some other minor stuff.
The 4001- 4004 architecture was refined by Hoff and Mazor thru
Dec. 1969, and by Jan 1970 Busicom agreed to our architecture.
The highlights of the architecture are explained elsewhere but
include: on chip dynamic RAM for stack and registers, 4-bit data
path, time multiplexing 16-pins, and I/O ports integrated on
memory chips with local instruction decoding.
Faggin joined Intel after architecture was completed. Faggin did
all logic design, circuit design, layout, production supervision,
testing, etc. Faggin also 'invented' how to do capacitor with
silicon gate which had very low gate/diffusion capacitance for
bootstrapping amplifier for dynamic logic circuits.
Shima did all final calculator code, layout checking, logic checking
and greatly contributed to the success of project; he also changed
2 instructions BBL and ADDC.
Intel filed patents on unique features of 4004, and the holders
are Hoff, Faggin, Mazor.
In summary split the credit as follows;
Noyce supported Hoff's proposal.
Vasdasz was Hoff's boss and supported Hoff's choice.
Shima/Busicom proposed using ROM coded routines.
Hoff proposed simpler architecture and micro-coded vs special chips.
Mazor helped refine ISA and chip architecture and wrote test code.
Faggin did all chip design and took over 'ownership' of project.
The whole design of the 4004 is like a Rashomon story in real life --
everyone thinks they are the main contributor.
Four people are credited with designing the 4004: Ted Hoff, Stan
Mazor, Federico Faggin and Masatoshi Shima.
There are evidently bad feelings between Faggin and Hoff because
Faggin feels he did all of the real work, and Hoff got much of the
credit. Many accounts do not give Shima any credit, only giving
credit to the three Intel engineers (Shima was an engineer at Busicom,
a Japanese calculator company at the time, and later became an Intel
engineer).
Slashdot Mirror
This is his list, re-formatted in HTML:
- Shima of Busicom specified ~8 custom chips, but there was some
micro-coding for floating point operations using 16-digit primitive
operations like shift left, add, subtract, etc.
- Hoff believed there were too many complex special chips and that
the micro engine was too high a level to be practical for Intel
to design and build and proposed an alternative architecture.
- Hoff proposed a simpler architecture using 4-bit data path and
simple instructions based upon the PDP-8 model we were using at
Intel to run 'Focal' programs.
- Shima and Busicom opposed the Intel architecture at first.
When I joined in Sept 1969 I was a computer designer for
more than 3 years and just worked on Symbol at Fairchild, and
did logic design on decimal floating point unit (like calculator).
- Hoff and I did program 'snipets' to prove how we could code the
'4004' to do calculator functions. I suggested enhancements to
ISA including; JIN, FIN, and FIM, and some other minor stuff.
The 4001- 4004 architecture was refined by Hoff and Mazor thru
Dec. 1969, and by Jan 1970 Busicom agreed to our architecture.
- The highlights of the architecture are explained elsewhere but
include: on chip dynamic RAM for stack and registers, 4-bit data
path, time multiplexing 16-pins, and I/O ports integrated on
memory chips with local instruction decoding.
- Faggin joined Intel after architecture was completed. Faggin did
all logic design, circuit design, layout, production supervision,
testing, etc. Faggin also 'invented' how to do capacitor with
silicon gate which had very low gate/diffusion capacitance for
bootstrapping amplifier for dynamic logic circuits.
- Shima did all final calculator code, layout checking, logic checking
and greatly contributed to the success of project; he also changed
2 instructions BBL and ADDC.
- Intel filed patents on unique features of 4004, and the holders
are Hoff, Faggin, Mazor.
In summary split the credit as follows;Four people are credited with designing the 4004: Ted Hoff, Stan Mazor, Federico Faggin and Masatoshi Shima.
There are evidently bad feelings between Faggin and Hoff because Faggin feels he did all of the real work, and Hoff got much of the credit. Many accounts do not give Shima any credit, only giving credit to the three Intel engineers (Shima was an engineer at Busicom, a Japanese calculator company at the time, and later became an Intel engineer).
Interview with Shima (extremely interesting and detailed)
Another interview with Shima
Interview with Hoff
An e-mail from Mazor, and nice pictures of the 4004
A really nice picture of the 4004
A picture of three of the engineers (no Shima) years later
A picture of all four engineers
Federico Faggin's initials on the 4004 -- the only initials on the chip