Slashdot Mirror
Intel's 4004 Microprocessor Turns 45 (4004.com)
mcpublic writes: Tuesday marked the 45th anniversary of the 4004, Intel's first microprocessor chip, announced to the world in the November 15, 1971 issue of Electronic News . It seems that everyone (except Intel) loves to argue whether it was truly the "first microprocessor"... But what's indisputable is that the 4004 was the computer chip that started Intel's pivot from a tiny semiconductor memory company to the personal computing giant we know today. Federico Faggin, an Italian immigrant who invented the self-aligned, silicon gate MOS transistor and buried contacts technology, joined Intel in 1970. He needed both his inventions to squeeze the 4004's roughly 2,300 transistors into a single 3x4mm silicon die. He later went on to design the Intel 8080 and the Zilog Z80 with Masatoshi Shima, a Japanese engineer with a "steel trap mind," the once-unsung hero of the 4004 team [YouTube].
Long-time Slashdot reader darkharlequin also flags the " fascinating, if true" story of Wayne D. Pickette, who was hired by Intel in 1970, worked on the 4004 project, and according to ZDNet "claims that prior to that, during his job interview with Intel founder Bob Noyce, he showed the company a block diagram of a microprocessor he'd started to work on three years previously when he was 17."
Long-time Slashdot reader darkharlequin also flags the " fascinating, if true" story of Wayne D. Pickette, who was hired by Intel in 1970, worked on the 4004 project, and according to ZDNet "claims that prior to that, during his job interview with Intel founder Bob Noyce, he showed the company a block diagram of a microprocessor he'd started to work on three years previously when he was 17."
Does anyone know if any 4004s are still functioning in a regular everyday way (not museums / collections)?
FFFUCK. This thing is amazing! It's got a whopping sixteen 4 bit registers! Finally ... I have been waiting a while for this. And I can't believe they've actually got the accumulator and push down stack one ONE chip ... not to mention the 4 bit parallel adder. Can't wait to purchase,
a beeyowolf custard of them?
The first era of silicon valley was the commercialization the transistor between the 1950s and early 1970s by Shockley and his renegades from the east coast. This established the culture of quick startup companies and nomadic engineers you really hadnt seen elsewhere in the world. Then when the number of transistors on a single chip exceeded a thousand using cheap CMOS technology you could put a whole CPU on chip and complete computer in a box for a couple thousand dollars. This lead to the second era of the personal computer in the 1970s and 1980s.
The crazy thing about computers is the field is so young most of the big innovators are still alive. Federico Faggin is someone you can go out and frigging talk to. It's like speaking to Maxwell after taking an E&M course- frigging insanity.
I was born one the same date (... 4 digit slashdot id checks out...). I have been using microcomputers since I was 10. I have never worked at anything other than software and hardware development.
Our contemporary computing ecosystem has evolved from the microcomputers I was born with. They actually have some architectural details that can be traced to the 4004's successor, the 8008.
Our computers are not descendants of the mainframes that came before them. By now, they have acquired many of the advanced features of mainframes. Implemented badly, several decades later. It is fascinating to learn about the history of mainframes. It is also somewhat depressing.
Those who do not learn from history are doomed to repeat it. Those who do learn are doomed to watch everyone else repeat it.
Stop worrying about the risks of nuclear power and start worrying about the risks of not using nuclear power.
are largely responsible for the "American greatness" that Intel has become. It's fun reading how so many great "American" technology achievements are in the hands of Asian and European people, all while the Americans accuse everyone else of stealing ideas and inventions from them.
Federico Faggin ... later went on to design the Intel 8080 and the Zilog Z80 with Masatoshi Shima, a Japanese engineer with a "steel trap mind ...
Which leaves out the fact that the 8008 was not an in-house-conceived upgrade of the 4004. Instead it was a commission, from Datapoint corporation, to implement the instruction set of their Dreatapoint 2200 terminal as a microprocessor chip.
A failed commission at that: TI dropped out early, and Intel got theirs to work, but with a chip that came in late, and slower than Datapoint's 100-ish chip TTL design (even though the latter's ALU was serial rather than parallel). So Datapoint and Intel agreed to settle the contract, with Datapoint being refunded the costs and Intel getting to sell the chip as their own when they got it finished, and make derivatives.
Great deal for Intel. Not so hot for Datapoint, whose flagship terminal was now facing competition based on their own instruction set and designs.
When you cut a deal with a big semiconductor house, you have to watch out for this sort of thing. As I understand it, the TI calculators came from a similar situation where TI built a 4-bit processor as a commission for a calculator manufacturer, then built and sold their own products around it and its follow-ons.
Similarly with Ford and Motorola. Ford commissioned the processor for the EEC-III without including an option for a spin to include design upgrades identified as very-useful-to-necessary. They identified several things that would make the chip better. So they reported them to Motorola in the hopes they'd incorporate them in a follow-on despite no contractual obligation to do so. They did make a follow-on with the improvements, which they sold to GM. B-b
So, as with a Deveel, if you think you cut a good deal with a semiconductor company, be sure to count your fingers, then your toes, then your relatives...
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
In 1970 a reasonable CPU would need ~20k transistors, and often had some user visible registers. In that year a single chip RAM could have 256 bits
of static RAM (registers), the limit being chip area and power dissipation. Furthermore 16 pin chips were practical, but more pins were too costly.
Ted Hoff's trick was using dynamic RAM for the CPU's registers (4 x16, the PC, and stack), and a multiplexing schema for using only 16 pins for
each of the 3 chips: CPU, ROM, RAM. Note: the applications for the microcomputer were mostly ROM based programs and subroutines were
needed to save program memory, and a stack was a good way to hold return addresses. Most applications didn't need much RAM data memory,
and 320 bits of dynamic RAM were in the RAM chip, and 256x8 was the ROM program memory. Another trick was to put I/O on the ROM
and RAM chips, so each memory chip gave ports for getting data signals in and out. Hence don't discuss the single chip CPU as an idea
of invention, the question is validly HOW DO YOU DO IT with ~2.5k transistors, and 16 pins, and 1/2 watt/chip.
The intel patent claims 17 'tricks'--not all of these are 'new' inventions, but the combination allowed the first microcomputer system to be
cheap and usable in new and novel applications. Our extension to the 8008 CPU was based largely on the ideas of the 4004, but eliminating
special RAM and ROM chips. My choice for 14 bit address space was chosen to save a couple of pins on the (18 pin) package, and 16k seemed
very adequate in 1970. stan mazor retired Intel engineer us patent: 3821715
https://en.wikipedia.org/wiki/...
https://en.wikipedia.org/wiki/...
put I/O on the ROM and RAM chips
Like the 6502 which only had one bus, for ROM, RAM and IO. But memory mapped video RAM was obviously IO as well.
http://michaelsmith.id.au
Imagine a Beowulf Cluster of these!!!
General Relativity: Space-time tells matter where to go; Matter tells space-time what shape to be.
If bipolar transistors are faster than IC (MOS) transistors, then why don't they try to make bipolar chips now, being manufacturing either may have improved, or R&D goes further since chips are big biz? Or, do other factors matter more?
Table-ized A.I.
maybe no one posted this because the stupid "joke" is fucking pointless on a 4004...
what would this achieve? ignore the RAM restrictions - what could you get - the speed of a 486 DX?
useless cunt
Linus Torvalds turns 47 next month.
No, more like if you put an I/O port on a 2716 EPROM.
For the 8080/6800/6502-era there were a few "RIOT" type chips (RAM or sometimes ROM, I/O, timer) which combined the rest of the functions needed for a minimal system, but the memory was added to the I/O chips, which is the other way around.
I think in the case of the 4004 it was mostly due to which chips had enough pins. Until modern all-in-one microprocessor chips (probably starting in the 90s with PIC), it was rare to have a CPU with a low (16 or less) pin count because they needed a lot of pins for a bus. This also allowed a 4004/4040 system to start with a small CPU chip and scale up the I/O as needed for larger applications.
Don't forget the 4040 either. I think they might have expanded the stack, but I am pretty sure it had better clock support for external oscillators / there were a bunch of nice glue chips which had stuff like you mentioned for IO / RAM / expansion. I think a slightly larger address space / still done with the expansion style mentioned above.
The 4040 added some instructions to the 4004 and was a bit better. I seem to remember something about the stack, but it's been so long... Early BYTE magazine actually had articles with 8008 assembler I remember. The early flavors of micros were diverse. Feature rich machines were damn expensive in general. People would cobble together whatever they could, instruction sets and features of processors were actively debated. The major camps seemed to be:
-- RCA COSMAC 1802. You could build a minimal system cheaply. Speed wise this was fairly slow, but it was a real processor. Lots of registers. Architecturally it wasn't as bad as some people like to say. Yes, you could easily implement a CALL/JSR and RTS too.
-- Nat Semi PACE/IMP-16. Most of the folks who went this route seemed to want more 'mini' style features - e.g. 16 bit instruction sets. These were very slow performance wise. Minis based around them were cheaper than the big-name mini's but seemed super slow. Met people who said their early 8080 micro kicked the mini's butt in terms of performance.
-- Single & multi chip direct from mini architectures. Stuff like TI 990, DG NOVA DEC PDP 8 and PDP-11. The TMS9900 was a notable contender.
-- Motorola 6800 and Intel 8080 - Both were popular, and much faster than stuff like the PACE / IMP-16 / 1802. They were usually simpler to design against and often faster than the single chip mini's derived stuff too (excluding maybe the TMS 9900 ).
-- MOS Technology 6502 at the same clock speed was usually equal or faster than a 6800 / 8080 for most stuff. It was super cheap too.
-- Zilog Z-80 A notably enhanced 8080. Loads more instructions. Very non orthogonal. Easy to interface to DRAM. Very very popular. The Z-80 pretty much gobbled up the 8080 family. Even when the 8085 came out, people went for Z80's. CP/M & S100 bus stuff, but lots of other stuff used Z80's like home micros etc.
The 6502 also gobbled up a bunch of designs. Motorola responded with the 6809 which is a really nice 8 bit to program. But by then the world was moving on to 16/32 bit stuff.
One of the most interesting things (to me) is that the Mini folks totally screwed up and ended up being wiped out. One wonders what would have happened if DG or DEC had produced a decent microprocessor based on the NOVA or PDP-11 at a low cost. DEC sort of did this with the LSI-11, but there was always this tension of the high-margin mini stuff being eaten from the bottom up. What if they'd been smart enough to pragmatically say - our high margin business is going to get destroyed by microprocessors, do we want them to be our microprocessors or someone elses?
One thing people forget about Intel is they did a superb job of supporting their stuff / had great sample designs that made getting started easier. Motorola for the 68000 actually actively discouraged folks and banned their engineers from talking to people who wanted to use the 68000 in the early days. Read D-TACK GROUNDED for details. It seems absurd right?
The general thing seems to have been that everyone wanted the new micros to go into the nice old high-margin mini business. Except they didn't. Cheap powerful MPU's meant people didn't want to pay 10x the MPU cost for software licensing fees or other garbage. They figured out how to cheaply hook video up to the microprocessors and keyboards. They figure out how to cheaply hook up mass storage. They built what was needed, and what worked got used. Not always entirely fair - there are plenty of amazing designs that never got traction, but life ain't fair.