A Short History of Microprocessors
The Early Days
The early history of the microprocessor is usually associated with the company Intel and the creative environment that existed in the San Francisco Bay Area -- or "Silicon Valley" as it is often called.   What is not commonly appreciated is that this was no accident, there had been a long and convoluted history of companies and people pioneering electronics firms in the San Francisco region for decades before Intel arose.

One could argue that the industry started back in the 1920's when three guys William Eitel, Jack McCoullough and Charles Litton decided to make radio tubes for their friends in the amateur radio arena having dabbled with radios themselves in the late 1910's. Back then, the Bay Area was a center of Ham amateur radio activity representing over 10% of all the Ham licenses in the country.  Indeed a couple of radio companies were actually located there.  Because amateur radio utilized higher frequencies than commercial radio at the time, Ham radio tubes had to be of a very high quality -- in particular hold very high vacuums.  Word spread about their tubes high quality. The business rapidly grew.

Litton decided to specialize in the manufacture of glass working laths that eliminate the need for tubes to be made by hand and founded the company Litton Engineering Laboratories. This increased tube reliability further. Eitel and McCollough joined forces to become Eitel-McCoullough.  The two groups produced their first major commercial tube called the 150T. It was an immediate success.  Their tubes had lifetimes of 20,000 hours against the norm back then of less than 1000 hours. The start of World War II saw an explosion in demand for tubes from the military. They made numerous high quality power grid tubes and tetrodes during those years.
 
While Eitel-McCoullough were concentrating on power grid tubes, Litton started working on the new British invention of the time, RADAR. This required the construction of a very unique and difficult tube, called a "Magnetron".   The microwave business really put Litton Industries on the map.  During the Korean war they were the first to manufacture pulse magnetrons, radar generating devices required to determine accurately target speed.  The company was very profitable, Litton retired from the company. It was sold it to an investment group and is a major corporation today.

Litton had a close relationship with a number of people at Stanford university.  In particular he was friends with two brothers there, Siguad and Russell Varian who went on to developed a better microwave device, the "Klystron".   The Klystron had a number of advantages over the Magneton, principally being it produced a more stable frequency controlled microwave. He enchoroughed them to set up their own business (Varian Associates) in the neighborhood.   Litton supplied the glass making machinery.  Varian went on to be a highly successful company, starting with Klystron manufacturing but later branching out into many industries.  Over time numerous support industries were founded around the above core of electronic professionals in the Bay Area.

The Transistor.
Case in point, shortly after the end of the war in 1945, Bell Labs in NJ formed a solid state physics group, led by William Shockley. While Shockley was born in Britain he grew up in the Bay Area.  His group at Bell Labs included John Bardeen and  Walter Brattain who along with others were responsible for inventing the transistor. Later internal personnel conflicts led Shockley to leave Bell Labs and come back to the Bay Area to head up a division of Beckman Instruments in Mountain View (called Shockley Semiconductor Laboratory) where he felt there was enough electronic talent to quickly pick up where he left off at Bell Labs. He recruited graduates form Stanford university and got off to a good start with the invention of the Thyristor.   However he became an extremely difficult person to work for -- particularly after he obtained the Nobel Prize in 1956 (along with Bardeen and Brattain).

A group of eight of his employees decided to leave and set up their own company focusing on the transistor.  They managed to obtain seed capital from an east coast company called Fairchild Camera owned by Sherman Fairchild in 1957.   This was arranged by Arthur Rock - back then, just a young security analyst. Fairchild camera made specialized aircraft surveillance cameras at the time and wanted to get into semiconductors.  The founders included Robert Noyce, Gordon Moore, and Jean Hoerni all of whom went on to be legendry in the future. They called the company Fairchild Semiconductor.  BTW, Shockley Semiconductor was purchased by Clevite in 1960, and officially closed shortly after being sold to ITT in 1968.

It's hard to appreciate now, but the early days of transistor technology were difficult. They got their brake when IBM decided to switch to silicon (instead of the then common germanium transistors) for core memories in their military applications for the B-70 aircraft.  Hoerni in particular perfected a large scale transistor manufacturing process that launched the company.  In time Fairchild Semiconductor dominated the production of planar silicon based transistors.  The company in fact became the major revenue source of revenue for the parent Fairchild Camera organization.

In the early 1960's the research group (and independently a group at Texas Instruments in Texas) realized that "microcircuits" could be laid down on silicon wafers.  However the parent Fairchild Camera company back east did not want to developed the technology.  Worse it was becoming clear that Fairchild Camera was not ploughing money back into Fairchild Semiconductor.   The founders concluded that Fairchild was not the best place for them to develop microcircuits.  They approached Arthur Rock about setting up another new company.  Hoerni and a few others left and founded a Teledyne subsidiary called Amelco.  This was the first of an avalanche of employees from Fairchild Semiconductor.  Many others defected and became seed stock for numerous Bay Area microelectronic companies in the early 1960's.  Perhaps most notable was Signetics which was founded by Kattner, Allison and Weissenstern of Fairchild Semiconductor. The financing ($1M), was arranged by Lehman Brothers in NY.

It was not long until it became clear that microelectronics was the future.   Signetics perfected DTL devices. These were easier to integrate into circuits than RTL devices.  "IC's" of increasing complexity and reliability started to appear.

Fairchild Semiconductor eventually realized they missed the boat.   They got financing from Corning Glass in NY along with Corning Glass management, scrapped a RTL logic approach they had in-house and competed directly with Signetics on a price basis.  Around this time the DIP IC's was developed by Fairchild Semiconductor. This was superior to the flat "T-packages" used earlier.  It gave Fairchild a boost and went on to be the dominant IC packaging format.

At that time Amelco took a different direction, concentrating on linear circuits.   These were technically more difficult.  Their 709 Op-Amp however put them on the map and became the standard operational amplifier of the era.  Jean Hoerni later left Amelco eventually setting up a the IC company Intersil focusing on MOS technology (watches, calculators etc).

Back at Fairchild Semiconductor in the late 1960's the research group there was in fact the first to start developing MOS integrated circuits.   The integrated circuits of the early 1960's were bipolar devices.  The simplicity and power reduction of MOS devices rapidly became apparent.  However the same old problem arose. Fairchild Camera/senior management did not want to rock the boat.  The key research team members became convinced they would get little financial rewards for their work.  At this time the financial community had organized in venture capital groups and were hunting down key people to do startup companies in the microelectronics area. One such company was National Semiconductor.   Charles Sporck and Pierre Lamond jumped ship from Fairchild Semiconductor to head up that company.  This devastated the Fairchild team and it was compounded by the fact their archrival Texas Instruments (TI) had just come out with a new and better type of digital logic, TTL. This was in essence the last straw for the company. Reorganizations came and went. Noyce had hoped to be CEO but was turned down. He, Gordon Moore and some of their closest friends there prepared to leave.

They approached Arthur Rock who immediately provided them with funding to set up yet another microelectronics company called Intel.  Another Fairchild employee Jerry Sanders left to start up his own company at that time also called Advanced Micro Devices, Inc. (AMD).

Noyce and Moore focused Intel initially on LSI IC's for memory circuits.  They also brought Federico Faggan with them from Fairchild who had perfected new techniques in MOS fabrication using polycrystalline silicon instead of metal gate techniques.  Their first memory chip was a MOS based 1101 256-bit static RAM chip.  It was TTL compatible and used in devices like printers.  They improved on its speed and a year later came out with the 1103 a DRAM equivalent chip.  Honeywell used their next generation DRAM chip a 512 bit chip in their early computers.  The company was airborne.

The Intel 4004.
In the summer of 1969, Intel was approached by a Japanese calculator manufacturer called Busicom to produce a set of custom chips designed by Busicom engineers for their new line of calculators. They came with their own circuit layout that had designed. It had a number of custom chips, each of which contained about 5000 transistors. A guy by the name of Ted Hoff at Intel was assigned to assist the team of Busicom engineers that had taken up residence at Intel to carry out the project. Hoff looked at the Busicom design and concluded it was way too complex to be cost-effective. Fortunately for them he had worked before with Digital Equipment Corporation's PDP-8 mini­computer, which had a very small instruction set. He reasoned that much of the calculator's complexity could be reduced if they used a small general-purpose processor.

Such a design, using software rather than electronics to do the calculating would be far more efficient and flexible.  It did not escape Intel's attention that it would greatly increase the memory requirements of the calculator, but then Intel was in the memory business at that time! Hoff also realized that this processor could be put to many other electronic applications as well. He pitched the idea to senior Intel management who to their credit, gave him the green light to go ahead. The Busicom engineers were still pursuing their original design when Hoff and his group (in particular Federico Faggin), started work on their new general purpose processor approach. And although the Busicom engineers had simplified their design, each chip still had over 2000 transistors.  It would take 12 such chips to make a functioning calculator. Hoff's team figured it would take only 1900 transistors to build his processor. Hoff's general-purpose processor design was chosen over the Busicom design, and Intel got a contract from Busicom to produce the chip that later became known as the famous 4004. Intel's big bang.

Producing the 4004 chip proved to be far more difficult than was earlier anticipated. In those days all lithographic layouts and designs were done by hand. Credit should really go to Federico Faggin for getting chip production going.  At that time Intel had only 150 employees most of them operators working in the wafer fabrication and chip assembly lines. The entire R&D organization was about 20 people!

Faggin brought with him two key inventions he made at Fairchild: the Buried Contact and the Bootstrap Load that were pivotal in making the 4004. He was at the time one of the few people experience in MOS logic and circuit design and had the crucial knowledge of the then new silicon gate process technology for the above "self-aligned gates". He took the 4004 chip from concept to silicon in just nine months.

At first Intel sold the 4004 exclusively to Busicom, but by the summer of 1971 they realized they had something special on their hands. They negotiated the right from Busicom to sell the chip set to other manufacturers. In November 1971 Intel advertised the 4004 as a four-bit processor that performed 60,000 operations per second. By February 1972 Intel had sold $85,000 worth of chip sets.  Management at the company started to realize they had a new focus.
  

	Intel 4004 CPU

   
The chip was primitive by today's design, it ran with a clock speed between 740 kHz to at the most 1 MHz.  It  could execute approximately 92,000 instructions per second and had separate program and data storage areas in RAM. It used a single multiplexed 4-bit bus for transferring: 12-bit addresses,  8-bit instructions and  4-bits of data.  It had a total of 46 instructions (of which 41 were 8 bits wide and 5 were 16 bits wide). The register set contained 16 registers of 4 bits each.   The datasheet for the 4004 can be obtained here. 


The Intel 8008
At the same time the 4004 was being developed, Another company called Computer Technology Corporation, (later renamed, Datapoint) placed bids with Intel and Texas Instruments to design a chip set for a new intelligent CRT terminal they were considering (Their popular Datapoint 3300 CRT terminal was plagued with overheating problems -- too many IC in a small space).

Both companies proposed an 8-bit general-purpose processor. Eight bits, because unlike 4-bits for calculators which work with BCD (binary-coded decimal), ASCII characters require 8-bits. Interestingly as it later turned out, CTC chose neither processor (Intel could not get their chip going in time, TI second sourced the Intel version but it proved buggy). CTC went on in the end to build its terminal with standard logic ICs. Fortunately TI and Intel went ahead with their "8 bit processor projects" anyway.  TI focused more on establishing intellectual property in the area initially while Intel quickly blasted ahead upgrading their 4 bit 4004 to the now famous general purpose 8008 "CPU".

The 8008, introduced in April 1972, was the first 8-bit microprocessor on the market. It required at least 20 support chips, but it had 45 instructions that it executed at 300,000 instructions per second, and it addressed a then whopping 16K bytes of memory. That was a lot of memory back then, The 8008 was considerable an upgrade of the 4004.   To some extent is was.

	Intel 8008 CPU

The chip was implemented in 10 micron silicon-gate enhancement load PMOS. Initial versions of the 8008 could work at clock frequencies up to 0.5 MHz, this was later increased in the 8008-1 to a specified maximum of 0.8 MHz. Instructions took between 5 and 11 T-states where each T-state was 2 clock cycles. The 8008 had 3,500 transistors. The chip utilized a single 8-bit bus but required a significant amount of external support logic. It could access 8 input ports and 24 output ports.

Intel's primary goal with the 4004 and 8008 was to replace a large number of discrete IC's with a singe software driven CPU.  Few people thought that these chips were suitable for general-purpose computing. At best they were used for "demonstrating computer principles".  In 1973 Scelbi Computer Consulting Inc. announced the first general-purpose microcomputer based on the 8008 called the  SCELBI-8H.  The company was founded in 1973 by Nat Wadsworth and Bob Findley. The 8H came with 1K of RAM and was available either fully assembled or in a kit form, but it never really caught on.   SCELBI discovered that they could make more money selling microprocessor software books than hardware and the business switched to highly documented software published in book form.  There were a few other attempts; in particular the RGS-008 from RGS Systems and the Mark-8 by Jonathan Titus which was published in amateur-radio publications.  They too did not succeed.

That said, the 4004 and 8008 did find good traction in dedicated controller applications. The often mentioned "traffic lights" application is indeed true.  The datasheet for the 8008 can be obtained here. 


The Intel 8080
In April 1974 Intel announced the 8080, a significant upgrade to the 8008 that required only six support chips.  It had 75 instructions and a tenfold increase in throughput over the 8008.  It could addressed a whopping 64K bytes of memory.  The 8080 design was proposed by Faggin, but the design team was headed by Masatoshi Shima, a young engineer Intel had wooed away from Busicom (see above). Having learned from the limitations of the 4004 and 8008, the designers made improvements to make their new chip a truly useful computing engine. The 8080 really was the first microprocessor not aimed at simple discrete logic  IC replacement.  It looked much more like a computer than anything that had come before it, and it was also much easier to use from a hardware standpoint.
  

	Intel 8080 CPU

      
While it was an extended and enhanced variant of the earlier 8008 design, it did not maintain binary compatibility. The initial specified clock frequency limit was 2 MHz and with common instructions having execution times of 4,5,7,10 or 11 cycles.  This meant a few hundred thousand instructions per second. It was implemented using non-saturated enhancement-load NMOS, unfortunately demanding an extra +12 volt and a −5 volt supply.

An article in the January 1975 Popular Electronics magazine featured the first in a series of construction articles of a home computer utilizing the 8080 CPU.  It immediately caught on and sprung up a whole cottage industry. See below.


The Motorola 6800
Seeing the success of the 8080's success other major chip manufactures began to wake up. Motorola began working an equivalent chip which they called the 6800.  This chip was designed by a guy called Chuck Peddle. Motorola was the first company to introduce a complete line of peripheral chips designed specifically to go with its new microprocessor --thus providing a very integrated hardware solution. These chips included a parallel I/O port chip (the 6820), and serial port chip (the 6850).   Motorola also delivered with their chip set a data set manual that became the gold standard for the industry.  At $25 a pop everybody had to have one.  This probably more than anything else was that book that got many utilizing microprocessors in their hardware designs or as a hobby.
  

	Motorola 6800 CPU

       
This was also the beginnings of the two great microprocessor religions.   Intel and Motorola CPU's look at the outside world differently.   Intel treats all memory words with the low byte first in RAM then the high byte -- this is called "Little Endian" storage.  Motorola places the high byte first then the low byte -- this is called "Big Endian" storage.  Assemblers in the Intel world move data/registers right to left.  In the Motorola world it's the opposite.  Intel treats outputs to RAM and IO ports differently with separate hardware and software instructions. In the Motorola world, IO ports are simply special set aside RAM locations with no unique CPU hardware pins. 

BTW, the "microprocessor Endian camps" grew out of the mainframe and mini-computer camps which had grown up previously.  The first real "byte" oriented machine was the IBM System/360 mainframe (~1964).  This was a "Big Endian" machine.  Earlier IBM mainframes like the 7094 all the way back to the 709 (vacuum tubes) were 36-bit whole word only oriented machines.  The Data General NOVA series was 16-bit word addressed, and originally used Little Endian byte ordering in the software.  They later changed all of their software to big Endian.  This was probably because they introduced floating point hardware that used the IBM System/360 floating point representation, which was Big Endian. The DEC PDP-11 minicomputer (but not the earlier PDP-10) was Little Endian  as was the follow-on VAX-11.   The microprocessor world followed the mini/mainframe computer traditions, Intel went the Little Endian route following the DEC lead, and Motorola went the Big Endian route following the IBM lead.

In the early days (particularly for 16 bit CPU's, see below), Intel tended to designate specific registers to specific hardware/software tasks. Motorola tended to have more registers and treated them all the same.  Also in the early CPU's, Intel tended to utilize less silicon compared to the larger Motorola CPU's.  Each of the above have their followers and advantages.   Zilog followed the Intel model with its Z80, Z8001 etc. series, as did the strange Rockwell PPS-8 or Signetics 2650 CPU's.  On the other hand the MOS 6502 (but was Little Endian), Texas instrument TMS9900,  National Semiconductor PACE or NS16000 etc followed the Motorola camp.  This split exist even today with the IBM-PC/Clones/Windows world and the Apple family.


The MOS Technologies 6502 CPU
In early 1975 Chuck Peddle (and a few others), left Motorola and joined a company called MOS technologies. MOS back then was a small semiconductor design and fabrication company based in Pennsylvania.  They were setup to provide  a second source for Texas Instruments (TI) to provide electronic calculator chips.  Once TI started producing their own chips, MOS diversified and became a supplier to companies like  Atari, producing for example the single-chip Pong IC.

The Peddle team got the go ahead to build a new low cost 8 bit CPU. The result was the 6501.  While it was similar to the 6800,  by using several simplification tricks in the design, the 6501 could go be up to four times faster. Price was always a big selling advantage for the 6501. MOS simply outperformed others in early working chip yields.  (They discovered a way of fixing minor lithographic design layout bugs quickly and easily). 
   

	MOS 6502 CPU

       
The 6501, while code incompatible with the Motorola 6800, however it  was hardware pin for pin compatible. Motorola sued and forced MOS to make a slightly different chip -- the now famous 6502. A CPU price war started. Intel and Motorola CPUs were at that time selling for greater than $170 each. The MOS chip was one quarter that number.  The 6502 would quickly go on to be one of the most popular chips of its day. A number of companies licensed the 650x line from MOS, including Rockwell International, GTE,  Synertek, and others.  Further improved versions came later for example the 6507 was used in the Atari 2600 or the 6510 for the Commodore 64.  Of course the most famous of all utilizers of the 6502 was the Apple.


The Texas Instruments TMS9900
As far back as 1976 TI introduced the TMS9900. It was in fact the first general purpose 16 bit microprocessor and was way ahead of its time in many respects.  (There were earlier 16 bits CPU's, for example bit slice logic chips or National Semiconductor's  IMP-16 chip set in 1972 or its later one chip incarnation called the Pace. However these were not really general purpose CPUs). 

In their wisdom however, Ti never courted the early fledgling hobby PC community and almost went out of their way not to provide support or encouragement to an individual developing a hardware project.  Ti's microprocessor focus then was large corporations and industrial applications! Only the Marinchip TMS9900 CPU board was made for the S-100 bus. 
  

	Texas Instruments TMS9900  CPU

The TMS 9900 had a 15-bit address bus, a 16-bit data bus, and only three special internal 16-bit registers.  What was unusual about this microprocessor was all its other general purpose user registers (16 in all), were actually kept in external memory. A single workspace register (WP) pointed to the 16 register set in external RAM (each 16 bits wide).  While unheard of today for a general purpose 16 bit CPU, this was OK at the time because most static RAM access times back then were faster than CPU register times.  The CPU had a simple but powerful instruction set. Fifteen of its 16 registers could be indexed, although it should be remembered these registers were actually in RAM.  Interestingly, illegal opcodes were treated as NOP's.  Also unusual, Serial I/O was available through address lines. Common today, unheard of back then. 

The TMS9900 was however later used in the TI-99/4 home computers. Unfortunately, to reduce the production costs, TI limited the system to just 128 of 16-bit words of the fast RAM that the TMS9900 could access directly. The rest of the memory was 16 KB of 8 bit RAM that was accessible only through the video display controller, which crippled the performance of the TMS9900.


The Zilog Z80

In 1976, Federico Faggin (see above) left Intel and formed his own company which they called Zilog Inc. He took Masatoshi Shima with him. Their goal was to build a better 8080. Almost immediately Zilog announced the Z80, a significantly enhanced 8080 that ran all of the programs written for the 8080.  Not only that, the chip could run at 4 MHz, twice as fast as the 8080. In addition it had many additional more powerful instructions than the 8080 (total of 176 in all). The early Zilog "Engineering Samples" were in fact manufactured in Dallas, not in Silicon Valley by a company called  Mostek, a Z80 later second source of the chip.  Needless to say the chip caught on very quickly.   The very first S-100 CPU board with the chip came from TDL in 1977.  Eventually every S-100 board manufacture had to have one. 

	Zilog Z80 CPU

     
  
On a wider scale however Zilog ran into a problem.  Even though the Z80 was a much more powerful chip than the 8080 in terms of its instruction set, very few people were writing software to take advantage of the Z80's extra instructions because the majority of the machines installed at that time were 8080-based, and if you wrote code that only ran on a Z80, your market would be considerably smaller.  Not helping matters, Gerry Kildall of Digital Research had a close connections with Intel and insisted in producing all his CPM code and BIOS examples in 8080 format.  (The Zilog mnemonics were more logical and easier to learn but never really caught on).

However on the hardware side, designers stopped using the 8080 in new computers. The Z80 was just a far easier chip to use, requiring no support ICs and a single voltage power supply. It was much faster, even if you didn't use the extra instructions. On top of all that, it had built-in support for refreshing dynamic RAMs -- something no other CPU had at the time.  This allowed the production of a very cost effective the Radio Shack TRS-80 (designed by Steve Leininger) amongst many other lesser well known single board 8 bit computers at that time.


The Motorola 6809
The Motorola 6809 was an 8-bit (with some 16-bit features) microprocessor CPU from Motorola, designed by Terry Ritter and Joel Boney and introduced 1978. It was a major advance over both its predecessor, the Motorola 6800, and the related MOS Technology 6502. The 6809 had two 8-bit accumulators (A and B, which could be combined into a single 16-bit register, D), two 16-bit index registers (X, Y) and two 16-bit stack pointers. The index and stack registers allowed very advanced addressing modes. Program counter relative addressing allowed for the easy creation of position-independent code, while a user stack pointer (U) facilitated the creation of reentrant code. The 6809 was the first microprocessor able to use fully position-independent code without the use of programming tricks. The 6809 was source-compatible with the 6800, though the 6800 had 78 instructions to the 6809's 59.  The instruction set and judi bola register complement were highly orthogonal, making the 6809 easier to program than the 6800 or 6502. Other features were one of the first hardware-implementations of a multiplication instruction in an 8 bit CPU, full 16-bit arithmetic, and an especially fast interrupt system. 
 

	Motorola 6809 CPU

The original 6809 ran only to 1 MHz, 1.5 MHz (68A09) and 2 MHz (68B09) were later produced. Faster versions were produced later by Hitachi. While there were one or two S-100 boards utilizing the chip, the Ackerman Digital SBC being perhaps the best known, the chips main claim to fame was its use in the Radio Shack Color Computer. Except for its speed the 6809 probably represented the most advanced instruction set of its time for an 8 bit CPU.


Intel 8085
By the mid 1970's Intel had realized that its 8080 needed serious upgrading. In 1977 they introduced the 8085. It too had all the 8080 instructions, plus a few more. But Zilog had taken them by surprise. The enhancements that the 8085 had over the 8080 were nowhere near as extensive as those of the Z80. From a purely hardware standpoint, the 8085 was a much nicer chip than the Z80 in terms of clean simple signals lines, but the Z80 had mass appeal because it was faster. Intel had to come up with something fast!  When introduced, the 8085 ran at only 3 MHz compared to the 4-MHz chip that Zilog which was already shipping.  It was not a great success for desktop systems. However Compupro used it in their 8085/8088 S-100 board where for many it became their stepping stone from 8 bit hardware and software to 16 bit systems.  The chip did find some utility for standalone controller applications however.
  

	Intel 8085 CPU

Like the Z80 it required only a 5 volt power supply and did not require a separate 8224 type clock generator as its parent the 8080 did. It contained 6500 transistors.   In a pattern later found in its 8086 (see below) cousin its data bus was multiplexed with the lower 8 bits of its address bus.  This is something the Z80 did not do. This all but killed any chance of the chip generating its own onboard dynamic RAM refresh as the Z80 did.

The 8085 did have better support for hardware interrupts over the 8080.  While they were nowhere near as powerful as those of the Z80 with its own onboard interrupt register, there were three maskable interrupts (RST 7.5, RST 6.5 and RST 5.5), an 8080 style interrupt pin and an NMI interrupt.  These too were really only suitable for stand alone controller type situations.  Nobody wanted to support these "special interrupts" in software for a small player in the 8 bit desktop arena.
 
Software/register wise the 8085 behaved like a faster 8080.

Western Design Center W65C816
In 1982 Apple commissioned Western Design Center to develop a more powerful 6502 CPU for its next generation personal computers.  They wanted it to be software downward compatible with the 6502 but have 16 bit capability.  Atari also wanted a similar CPU.

The result (in March 1984) was the WD W65C816.  It was a fully static CMOS design for low power consumption (300 µA at 1 MHz).  In "Emulation mode" it was completely software compatibility with the 65C02. It had 24-bit memory addressing, providing access to 16 MB of memory space. It had a 16-bit  accumulator, stack pointer and index registers. There were separate program and data bank registers allowing program segmentation or 16 MB linear addressing (data only). There was even an option for Co-Processor support.
    

	Western Digital W65C816 CPU

Unfortunately this powerful chip came a little too late into the 16 bit arena.  Today up to 200MHz versions are available and are used in dedicate applications.


Intel 8088 and 8086
Intel realized that the 8080/8085 family needed serious upgrading. In 1976 they started to focus on a 16-bit processor.   They realized that earlier 16 bit designs never caught on because of the lack of software or incompatibility of software with the then current popular 8 bit systems.  They soon realized that direct opcode upward compatibility from the 8080 opcodes was not really practical.  Instead they settled for a situation where each 8080 register had its 8086 counterpart. This made 8080-to-8086 code translators possible and gave programmers a familiar starting point. This was a big factor in the success of the 8086, which was announced to the world in 1978. 
  

	Intel 8086 CPU

 
 
On top of that they added an option for hardware compatibility. Two 16 bit CPUs were designed,  the 8086 and an 8 bit 8088 CPU that behaved -- at least to some degree -- like an 8080/8085 externally, but behaved like a true 16 bit internally.  They also came up with the concept of a microprocessor  "bus-interface unit" (BIU). The BIU handled all communications with the outside world and was in charge of generating addresses and storing and retrieving data from the system. Inside the BIU was a queue. While the execution unit was busy crunching data, the BIU was out on the bus getting the next instruction and putting it in the queue. The 8086/8088 BIU can stay up to 8/6 bytes ahead of the execution unit by keeping bytes in its queue. Because of the queue, the 8088 performance only suffered an average of 20 per­cent compared to an 8086.

Intel did one other neat thing as well with the hardware.   They set aside one pin (33), called MIN/MAX, which when held high configured the chip as a simple 8085 style chip capable of directly addressing RAM, control lines and IO ports. This was useful for small dedicated microprocessor systems. When tied low however the microprocessor was completely reconfigured internally to talk externally to a bus controller (8288) and if required a "coprocessor".  For the latter there were two types. A math-coprocessor (8087) and an IO-coprocessor (8089).  This was before the days of sub-micron lithographic silicon layouts. Chip real estate was scarce. The concept of adding extra coprocessors on an as needed basis was smart. 

Finally also learned from Motorola the importance of good documentation/manuals.  Bookstores became flooded with 8086 hardware and software manuals.  Much to Intel's relief IBM chose the 8088 as for their PC.  The rest is history.


The clock frequency of the early 8086s  was limited to 5 MHz,  but the later versions in HMOS were able to reach 10 MHz and above. HMOS-III and CMOS versions were manufactured for well into the 1990s by many second source manufacturers (Fujitsu, Harris/Intersil, OKI, Siemens AG, Texas Instruments, NEC, Mitsubishi, and archrival AMD). There were a large number of successful S-100 8086 and 8088 boards produced (e.g.. Compupro, Lomas, Seattle Computer to mention a few).

The Motorola 68000
In 1976, designers at Motorola were working on a new 16-bit microprocessor as well. Their focus was software more than hardware. The wanted to have a CPU with internal 32 bit registers and an external 16 bit interface.  Unlike Intel, they wanted to eliminate any special-purpose instructions and allow the microprocessor to perform all operations, on all registers, on all data types, and in all addressing modes.  Assembly language programmers like this because it means they don't have to memorize a bunch of exceptions to the instruction set.  However this advantage is limited because the bulk of code programming even at that time was above this level of detail.  That said, the 68000 leant itself to UNIX style  systems. There were a number, but none really became successful.  If it were not for the Apple Macintosh the 68000 CPU would have probably have ended up like some of its relatives, the Zilog 8000 or National 16032 (see below).

 

	Motorola 68000 CPU

 

Not to be outdone in hardware Motorola also came out with an 68008, an 8-bit-bus version of the 68000, similar in concept to the 8088. However, the 68000 had no real internal queuing capability, and that meant that the 68008 ran half as fast as the 68000. This made the 8088 look even better!

The clock frequency of the early 68000s  was limited to 6 MHz,  but the later versions were able to reach 20 MHz

The 68000 could address 16 MB of physical memory with byte resolution. Address storage and computation used 32 bits, however, with the high-order byte was ignored due to the physical lack of pins. This allowed it to run software written for a flat 32-bit address space. Motorola's intent with the internal 32-bit address space was forward compatibility, making it feasible to write 68000 software that would take full advantage of later 32-bit implementations of the 68000 instruction set.  The CPU also had a System and User mode where certain instructions were not available in user mode.   There were a number of successful S-100 68000 boards produced (e.g.. Compupro, Dual Systems).  BTW, Alan Wilcox wrote an excellent book ("68000 Microcomputer Systems", Prentice-hall, Inc, 1987) about putting the 68000 on the S-100 bus. A very interesting read.

The Zilog Z8000
Flush with it's initial success with the Z80,  Zilog too started thinking about a 16 bit microprocessor.  Masatoshi Shima (see above)  begun working on a 16-bit processor, the Z8000, using "hand drawn" random logic.  Random logic is fast and worked fine for the 8080 and Z80 but for larger more complex CPUs things get too complicated the logic gates become a true can of worms. CPUs like the 68000 used internal micro-code ROMS to control the chip. To make changes you just change the microcode.  Not so for random logic. Bugs continuously cropped up is early Z8000 designs. To make matters worse Shima left Zilog to return to Intel delaying the final Z8000 even further.  The last straw for many potential customers was the Z8000's lack of a similar instruction set to the Z80 - as the 8086 had.   The chip never became successful and was the start of a slow lingering death for Zilog.  Zilog did go on to produce a 32 bit processor (the Z80000). It was a phenomenal chip internally, but sadly the damage was already done.
   
  

	Zilog Z8002 CPU

The register set consisted of sixteen 16-bit registers  that could use or combined to act as 8-bit, 16-bit, 32-bit, and 64-bit registers. The register set was completely orthogonal, with register 15 conventionally designated as stack pointer, and register 14 normally used for a stack segment. There was both a user mode and a supervisor mode and like the Z80, the Z8000 included built-in DRAM refresh circuitry.

It should be mentioned there was no actual Z8000.  The actual chips were the Z8001 and Z8002.  The Z8001 utilized a segmented memory approach like the 8086 (using a 16 bit offset and a 7 bit segment value).  However unlike the 8086 they were not added internally but required external hardware. The Z8002 was a simpler non-segmented version.  I know of no S-100 Z8000 CPU boards being made.


The National Semiconductor 16032 CPU
National realized they missed the boat for early 8 bit microprocessors.  Their early SC/MP 8 bit CPU which with its limited addressing capability (4K), was really best suited as a simple micro-controller.  In 1981, National Semiconductor made a second attempt at the microprocessor market with the 16032. (It was later renamed 32016 to emphasize its 32-bit internals). The 16032 had a 32-bit (internal data bus) microprocessor with a 16-bit external bus.  The chip had a basic clock speed of 10 MHz.  It had eight 32 bit general purpose registers which could be used as bytes, words or double words. However unlike the Z8000 or 8086 bytes were restricted to the 8 least significant 8 bits.
    

	National Semiconductor 16032

   
Since Motorola had never been able to produce their promised math coprocessor at the time, and Intel's 8087 couldn't break the 5-MHz speed barrier, everyone was impressed when National announced its math coprocessor would run at 10 MHz,  twice the performance of Intel's 8087.  It did, the only problem was National could not produce its 16032 bug free in decent quantities.   Programmers liked its instruction set, which reminded them of a VAX. Because the VAX has been so successful for running UNIX, the 16032 seemed like a shoe-in for a UNIX computer. However the hardware delays as in the case of Zilog, prevented the microprocessor from getting traction in the early marketplace. It never recovered.  I know of only one S-100 16032 based CPU board being made (Compupro' s NS32016 CPU Board ).  BTW, there is a great web site outlining the history of the National Semiconductor CPUs at cpu-ns32k.net.


The Intel 80286
Having learned their lesson from being asleep at the wheel when the Z80 overtook the 8080, Intel did not want to make the same mistake with the 8086.  Realizing that the 8086 was limited to 1M of RAM and did not even have a User/System partition of opcodes (something the other 16 bit CPUs already had), Intel quickly got going on an improved 8086 which became the 80286.  (There was one side step in the form of the 80186 which was a 8086 with a DMA controller, Interrupt controller  IO ports etc onboard).   However this CPU really was meant to be for dedicated controller applications rather than a general computer.  There were one or two S-100 boards made that utilized the 80188 (e.g.. Lomas, IMS) , but the hard wired IO port addresses within the  80186 CPU caused conflicts with other popular S-100 boards in the bus).

The 80286 was introduced in 1982. It had back then a whopping 134,000 transistor count.  It was the first Intel CPU with memory management and memory protection abilities.  Initial versions had a clock speed of 6 Mhz.  This later was extended by Intel and AMD to 12MHz and above.
 
The 80286 operated in an 8086 "Real mode" where it behaved like an 8086 or in "Protected mode" where it could access up to 16MB of RAM.  However its onboard memory management allowed software to operate with up to 1 GB or virtual RAM.  The 8086 style segment registers in protect mode pointed to special tables in RAM that controlled all the CPUs RAM access. 

	Intel 80286  CPU

   
The chips real claim to fame was that is was used in the IBM-PC/AT computer.  However in that box it really functioned as a faster 8086.  The chip had a severe limitation in that it could not reverted back to real mode from protected mode without a hardware reset. This was just not practical on the PC/AT.  Consequently there was never a significant interest in true 80286 based operating systems. Digital Research tried with Concurrent DOS 286, but CPU chip bugs delayed the product until it was too late.  MS-DOS remained 8086 based on PC/AT's.  There were a number of successful S-100 80286 boards made (Compupro, S C Digital, Macrotech are good examples).


The National Semiconductor 32032
National took one more crack at the general compute/CPU market.  They were the first company to announce and ship a full-blown 32-bit microprocessor in 1984. The 32032, which used to be the 16032, was code-compatible with the 32032.   If life was fair it should have been a successful chip but it was becoming increasingly clear they the desktop computer world only had room for Intel and Motorola based chips.  Nevertheless the 32032 was an amazing microprocessor. 
  

	National Semiconductor 32032

	Motorola 68020 CPU

 

A Short History of the S-100 BUS.
  

The S-100 Bus started life in 1974 when Ed Roberts* set about designing a home "micro-computer" kit for electronics hobbyists. He set about designing the now famous Altair 8800 utilizing the Intel 8080 CPU. Apparently he could not get a proper bus motherboard  made in time to coincide with an article he was writing for the Magazine Popular Electronics then due out in Jan 1975.  He looked around for some suitable edge connectors that could be strung together. He came across a supply of cheap 100-pin edge connectors and decided to use these on a short board of 4 connectors at a time. Later hoping to link multiple boards together to form a bus. The micro-computer kit took off. He called the bus the "Altair Bus". It was an instant hit and soon sparked a whole cottage industry of Altair bus compatible machines. Almost immediately the company Micro Instrumentation and Telemetry Systems, Inc. (MITS) took off. However in May 1977 Ed  (and co-partner William Yates) sold MITS to Pertec Computer Corp., in Chatsworth, California based on economic necessities. MITS was totally overburdened with quality and delivery problems and Ed feared loosing his company with the threat of bankruptcy.  He couldn't otherewise continue the business without debt capital.

Ed Roberts,  is best known for designing the first successful home computer, the Altair 8800, and hiring Bill Gates and Paul Allen to write software for it.   (*Sadly, Ed passed away in early April 2010.)

The Pertec purchase was motivated mainly by the ownership of the Microsoft BASIC sources and general license that Pertec thought to be included in the deal, but later it turned out to be false. Ed became head of a new department set up for him at Pertec Computer Corporation (and more than $2 million on his bank account). Unfortunately Pertec closed his department against all promises in November of the same year because of continuously decreasing sales. Ed left Pertec very disappointed and angry. He refused to sign another contract offered by Pertec, his only passion at this time being the Altair computer. He returned to Georgia to be a gentleman farmer and later took the opportunity to begin his medicine studies. He studied at Mercer University in Macon, Georgia and graduated with a M.D. in 1986. (9 years after the Pertec deal, at the age of 44!)
Early S-100 Systems A very typical setup in the mid 70's was an Altair connect to a Teletype printer. The Teletype printer could act as an input device, and you could use the paper tape to save and reload small amounts of code. One thousand bytes of code was a lot in those days! This setup was the basis of CPM's consol IO and data redirection and can be found in monitors like TDL's Zapple Z80 Monitor.    In 1975 another company IMSAI, came out with what became the gold standard S-100 motherboard and computer.  Other well respected early Altair bus complete computers at the time were  Cromemco and Processor Technology. Roger Mellon of Cromemco, referred to the bus as "The S-100 bus". The name stuck.  That same year Processor Technology sold a 4K static RAM S-100 board launching a cottage industry of S-100 add-on boards.   In 1976 Cromemco introduces the TV-Dazzler a dual S-100 video board.  The same year the disk operation system CP/M was advertised in BYTE magazine for the first time. However floppy disk based systems were still out of reach of most uses at that time.  Nevertheless the market expanded rapidly. Soon many other companies sprang up making individual S-100 boards.   In 1977 Tarbell introduced the widely popular Tarbell S-100 cassette controller board, which allowed users to load, save and exchange software with colleagues. It is now forgotten the impact this board had on knitting users together. It was the first general medium for example in which commercial software was sold.  Specialized S-100 boards started to appear. That same year DC Hayes introduced the first Modem S-100 board launched. TDL launches its Z80 S-100 board.  This board along with their famous "System Monitor Board "(SMB) made setting up a complete and functional micro-computer easy.  The bus had gotten critical mass and momentum. Other complete S-100 bus computer systems like Northstar and Victor Graphics S-100 Computers were introduced.  The boards themselves became more multifunctional and complex. S-100 board manufactures such as Morrow Designs, Godbout Electronics and SD Systems etc. were good examples of this.   1978 saw the widespread introduction of floppy disk systems. Initially the Northstar hard sectored 5" disk system was popular but this later gave way to a number of soft sectored format 8" and 5" CP/M systems. The  popular and reasonably priced  SD Systems Versafloppy S-100 board was introduced that year. The 5" Shugart floppy drive started to be used by hobbyists.  That year a company called Micro Pro Int. announced Word Master a word processor precursor to Word Star the first widely used word process or for PC's. That year also Epson announces the MX-80 dot matrix printer, which establish a new standard in high performance printing for a low price.   In 1979 Seattle Computer Products ships its 8086 S-100 board. In a little noticed event at the time, this also was that year also Xerox, DEC and Intel introduced the "Ethernet" standard for local networks.   Against this background slight differences in signal timing between some board manufactures became a problem. A standards committee was formed to define the S-100 bus. This helped. The standards group later went on to introduced a 16-bit data bus to the S-100 bus for 16 bit CPU's.  (Towards the end of 1983 it was standardized as the S-100 IEEE-696 standard).   By 1979 the initial S-100 companies IMSAI, MITS, TDL and Processor Technology had evolved into other companies. For example, Ed Roberts of MITS sold Altair to a company called Pertec because his personal goals were to make money to become a medical doctor (which he did).  IMSAI no longer was the dominant company it once was.  The whole field was evolving very rapidly. By 1979 there were probably over one hundred S-100 companies making boards or writing software for them.   That same year Intel introduces the 5 MHz 8088 microprocessor, a 16 bit CPU with an external 8 bit interface. Early prototype S-100 boards with 16 bit CPUs were being tested.      If 1978 was the year of the floppy, 1980 was the year many people switched to hard disk based CP/M systems.  A typical starter system consisted of one Seagate "Winchester" 5", five Mg of storage. At this stage S-100 computers were becoming quite sophisticated.  Primitive multi-user/network systems were also starting to appear in corporate environments.  The S-100 bus evolved into the standard for all "professional" personal computers most of which ran CP/M. In 1981 IBM introduced the PC with its own IBM-PC ISA bus.  The home computer landscape changed forever. In 1982  Digital Research started selling CP/M-86. While it found a home on some S-100 systems, most users utilized MS-DOS with the IBM-PC. 16 bit S-100 systems were starting to appear in numbers.    As microcomputers got smaller and faster, S-100 became obsolete.  The boards were becoming too large for the improved IC chip packaging of the time.  With the introduction of the IBM-PC ISA bus, the S-100 bus started to go out of favor.  However 10's of thousands of these boards still lie around. Some in functional systems, many in attics or basements.  In the past few years there has been a revival of these old systems and a new era of S-100 computers has appeared.  Hobbyist like those in the vintage car industry are actively getting together to bring new life to these wonderful machines.    Unfortunately because the bus fell out of active use just as it was to approved by the standard committee of the IEEE. It was withdrawn in the 1990's because there was really nobody around to push for it. A great disappointment to many later.   The history of individual S-100 companies is outlined further under the S-100 Boards section. A very good book about understanding and building circuits for the S100 bus is the one by Sol Libs and Mark Garetz Interfacing to S100/IEE 696 microcomputers (Osborn/McGraw Hill, 1976).  See here.  This book is really a must for any serious S100 bus hardware user.