If you're interested in this subject definitely watch this 1978 documentary:
https://www.dailymotion.com/video/x64ng4h Robert Noyce appears about 7 minutes in, and I think it may be in that very office shown in this article.
I really enjoyed "Silicon" [1] by Federico Faggin. Besides telling his own version of the history of the microprocessor he also shares his other great passion: the study of consciousness
The official history of the Intel company claims that 4004 (in 1971) was the first "microprocessor".
The word "microprocessor" is typically used to mean "monolithic processor" a.k.a. "monolithic CPU", where "processor" a.k.a. "CPU" is a traditional term for a component of a general-purpose programmable computer, and "monolithic" means that the CPU is implemented in a single semiconductor chip.
Intel 4004 had nothing to do with general-purpose programmable computers. It was a chip used for making desktop calculators with fixed functions. Its architecture had nothing in common with the following microprocessors, from Intel and from other companies.
So I do not agree with Intel's claim. In my opinion, it is correct to consider that Intel 8008 from 1972 is the first microprocessor. Intel 8008, which implemented the architecture of the Datapoint 2200, was indeed a CPU for a general-purpose programmable computer.
The following mainline Intel processors, i.e. 8080, 8085, 8086 and so on, had ISA's derived from that of Intel 8008, and even the latest AMD Zen 4 and Intel Raptor Lake still include some derivatives of most of the 8008 instructions in their ISA's, for backward compatibility reasons.
The only reason why Intel 4004 has a place in the history of microprocessors is that this project has provided an opportunity for a team of Intel engineers to learn how to design complex digital integrated circuits in MOS technology.
Previously, Intel was a memory company. The logic design of memories was far simpler than the design of a CPU. Had Intel started directly with a general-purpose CPU, they would have needed much more time to discover how to design one.
After the experience with designing the Intel 4004, the Intel engineers have conceived a methodology of how to design complex logic circuits using their PMOS process, which they have then successfully used to design the Intel 8008.
If Intel 4004 would be considered as a processor, because it was also able to execute arithmetic operations, then it would not be clear where a line should be drawn. By that argument, why not also consider a monolithic ALU, like Texas Instruments 74181, as being a microprocessor.
The F-14 CADC thingy wasn’t a microprocessor though - it was a multi-chip solution.
I’m also not quite sure if it would qualify even if it were a proper microprocessor - because it was classified it didn’t contribute anything to the state of the art.
Finally - the marketing around CADC is weird and, frankly, quite dishonest. The website set up to promote it claims 4004 wasn’t really a single chip, and makes dubious claims about CADC performance capabilities, ignoring the fact that (if I’m reading https://firstmicroprocessor.com/wp-content/uploads/2020/02/a... correctly) it was serial, every instruction taking 20 clocks at ~300kHz.
EDIT: After spending some more time on that PDF... Wow, this was a curious beast.
First of all, looking from our current perspective, it was an asymmetric multiprocessor: there were three units, each consisting of a number of chips; one for multiply, one for divide, and one for logic. Each of them had separate RAM and ROM (128 instructions each); each had a different "instruction format"; finally each had a separate program counter. The instructions looked more like microcode than normal machine code, eg there were no registers, and instructions manually toggled specific logic signals to communicate with others. Same for jumps: you'd toggle the logic line to reset the counter that was part of the instruction ROM. There was no assembler either.
Thus, I'd argue not only was F-14 CADC not a microprocessor, it wasn't a CPU at all. It was a fixed-purpose computer, bit like Norden bombsight but implemented in silicon.
Being unknown to you doesn't disqualify something from being recognized as a first. Plenty of unknown inventors and scientists have been recognized as pioneers much later when their work was unearthed and made known.
It's an interesting challenge but the 4004 ended up in pinball machines and traffic lights I believe as well as Faggin's test equipment!
I'm not sure what is missing for it to fail to qualify as a microprocessor. It was limited certainly and doesn't have much in common with what follows, but it seems to pass the test for me.
It is reasonable to call Intel 4004 as the first microcontroller.
4004 had separate address spaces for data and for program. The program address space used 12-bit addresses and the program was stored in a read-only memory.
It was impossible to change the program of an Intel 4004, the program was written in the ROM when the memory was manufactured.
Like with any other MCU, it was possible to use an Intel 4004 to replace the logic that would have been previously implemented with multiple separate MSI and SSI TTL or CMOS logic integrated circuits.
As an aside the Intel Busicom contract calls it the NCM - ARU which I eventually worked out stands for Nippon Calculating Machine - Arithmetic and Register Unit.
> It was impossible to change the program of an Intel 4004, the program was written in the ROM when the memory was manufactured.
This is not accurate. If you used a 4001 ROM as program memory, then it can’t be changed, but this is a property of the system, not the processor, and intel made the 4008 and 4009 memory interface chips to enable using standard RAM chips as program memory for the 4004. The 4004 also had a WPM (write program memory) instruction that could be used in these situations.
I don’t understand why you are adamant the 4004 is not a microprocessor but the 8008 is. The 4004 is weird and clearly designed with some instructions specifically because it was targeted for a calculator, but that doesn’t mean it had “nothing to do with general-purpose programmable computers.” It was programmable and had a sufficient instruction set to be used for general purposes. Intel could have made a true fixed function chip just for the calculator, but they decided to made a cpu instead.
Hi everyone, author here. Didn't expect to be on the front page of HN!
I've long been fascinated by the early days of the microprocessor. A lot has been written about the 4004 but it's always seemed to me that Bob Noyce doesn't get enough credit. He's always there in the background making things happen.
I'm planning to write a series of short posts on related topics - some more technical in nature - others less so. The 4005, 8008 and competing architectures of the time will be first up.
Any ideas and feedback very welcome.
PS It's on Substack as it the only tool that has managed to cure my procrastination.
16 comments
[ 3.0 ms ] story [ 46.1 ms ] threadNoyce is looking at a picture of the 8080 I think.
[1] https://amzn.com/dp/1949003418
The word "microprocessor" is typically used to mean "monolithic processor" a.k.a. "monolithic CPU", where "processor" a.k.a. "CPU" is a traditional term for a component of a general-purpose programmable computer, and "monolithic" means that the CPU is implemented in a single semiconductor chip.
Intel 4004 had nothing to do with general-purpose programmable computers. It was a chip used for making desktop calculators with fixed functions. Its architecture had nothing in common with the following microprocessors, from Intel and from other companies.
So I do not agree with Intel's claim. In my opinion, it is correct to consider that Intel 8008 from 1972 is the first microprocessor. Intel 8008, which implemented the architecture of the Datapoint 2200, was indeed a CPU for a general-purpose programmable computer.
The following mainline Intel processors, i.e. 8080, 8085, 8086 and so on, had ISA's derived from that of Intel 8008, and even the latest AMD Zen 4 and Intel Raptor Lake still include some derivatives of most of the 8008 instructions in their ISA's, for backward compatibility reasons.
The only reason why Intel 4004 has a place in the history of microprocessors is that this project has provided an opportunity for a team of Intel engineers to learn how to design complex digital integrated circuits in MOS technology.
Previously, Intel was a memory company. The logic design of memories was far simpler than the design of a CPU. Had Intel started directly with a general-purpose CPU, they would have needed much more time to discover how to design one.
After the experience with designing the Intel 4004, the Intel engineers have conceived a methodology of how to design complex logic circuits using their PMOS process, which they have then successfully used to design the Intel 8008.
If Intel 4004 would be considered as a processor, because it was also able to execute arithmetic operations, then it would not be clear where a line should be drawn. By that argument, why not also consider a monolithic ALU, like Texas Instruments 74181, as being a microprocessor.
First microprocessor made by Intel.
https://www.wired.com/story/secret-history-of-the-first-micr... http://spingalhistory.blogspot.com/2018/04/microprocessor-hi...
I’m also not quite sure if it would qualify even if it were a proper microprocessor - because it was classified it didn’t contribute anything to the state of the art.
Finally - the marketing around CADC is weird and, frankly, quite dishonest. The website set up to promote it claims 4004 wasn’t really a single chip, and makes dubious claims about CADC performance capabilities, ignoring the fact that (if I’m reading https://firstmicroprocessor.com/wp-content/uploads/2020/02/a... correctly) it was serial, every instruction taking 20 clocks at ~300kHz.
EDIT: After spending some more time on that PDF... Wow, this was a curious beast.
First of all, looking from our current perspective, it was an asymmetric multiprocessor: there were three units, each consisting of a number of chips; one for multiply, one for divide, and one for logic. Each of them had separate RAM and ROM (128 instructions each); each had a different "instruction format"; finally each had a separate program counter. The instructions looked more like microcode than normal machine code, eg there were no registers, and instructions manually toggled specific logic signals to communicate with others. Same for jumps: you'd toggle the logic line to reset the counter that was part of the instruction ROM. There was no assembler either.
Thus, I'd argue not only was F-14 CADC not a microprocessor, it wasn't a CPU at all. It was a fixed-purpose computer, bit like Norden bombsight but implemented in silicon.
I'm not sure what is missing for it to fail to qualify as a microprocessor. It was limited certainly and doesn't have much in common with what follows, but it seems to pass the test for me.
4004 had separate address spaces for data and for program. The program address space used 12-bit addresses and the program was stored in a read-only memory.
It was impossible to change the program of an Intel 4004, the program was written in the ROM when the memory was manufactured.
Like with any other MCU, it was possible to use an Intel 4004 to replace the logic that would have been previously implemented with multiple separate MSI and SSI TTL or CMOS logic integrated circuits.
This is not accurate. If you used a 4001 ROM as program memory, then it can’t be changed, but this is a property of the system, not the processor, and intel made the 4008 and 4009 memory interface chips to enable using standard RAM chips as program memory for the 4004. The 4004 also had a WPM (write program memory) instruction that could be used in these situations.
I don’t understand why you are adamant the 4004 is not a microprocessor but the 8008 is. The 4004 is weird and clearly designed with some instructions specifically because it was targeted for a calculator, but that doesn’t mean it had “nothing to do with general-purpose programmable computers.” It was programmable and had a sufficient instruction set to be used for general purposes. Intel could have made a true fixed function chip just for the calculator, but they decided to made a cpu instead.
I've long been fascinated by the early days of the microprocessor. A lot has been written about the 4004 but it's always seemed to me that Bob Noyce doesn't get enough credit. He's always there in the background making things happen.
I'm planning to write a series of short posts on related topics - some more technical in nature - others less so. The 4005, 8008 and competing architectures of the time will be first up.
Any ideas and feedback very welcome.
PS It's on Substack as it the only tool that has managed to cure my procrastination.
The inventor of the diode deserves more credit for the transistor