Intel Releases 4004 Microprocessor Schematics

mcpublic writes, "Intel is celebrating the 35th anniversary of the Intel 4004, their very first microprocessor, by releasing the chip's schematics, maskworks, and users manual. This historic revelation was championed by Tim McNerney, who designed the Intel Museum's newest interactive exhibit. Opening on November 15th, the exhibit will feature a fully functional, 130x scale replica of the 4004 microprocessor running the very first software written for the 4004. To create a giant Busicom 141-PF calculator for the museum, 'digital archaeologists' first had to reverse-engineer the 4004 schematics and the Busicom software. Their re-drawn and verified schematics plus an animated 4004 simulator written in Java are available at the team's unofficial 4004 web site. Digital copies of the original Intel engineering documents are available by request from the Intel Corporate Archives. Intel first announced their 2,300-transistor 'micro-programmable computer on a chip' in Electronic News on November 15, 1971, proclaiming 'a new era of integrated electronics.' Who would have guessed how right they would prove to be?"

Heh

[Mitchell Mebane]

At first, I thought this was about Intel's new quad-core processors. How wrong I was. :P

Wouldn't it be cool, though, if Intel did name the quad-core chips the 4004 series?

Does it run Linux?

[BadAnalogyGuy]

With a better FPU and a faster front-side bus, that chip could possibly be useful.

As it is, I don't think it can even run a stripped down 1.0 Linux kernel.

Re:Does it run Linux?

[gadzook33]

No, no, it's fine. You just need to cross compile with ARCH=4004; OPTIMIZE_FOR_CPU=4004; STRIP_EVERYTHING_EXCEPT_RESET_INCLUDING_THE_KERNEL =true.

Re:Does it run Linux?

[mode13]

I can see it now:

From forums.gentoo.org / Architectures & Platforms / Gentoo on 4004 ...

Yea, I just did a stage 1 install, it took 12865 hours but the binaries are TOTALLY optimized!

Re:Does it run Linux?

[raynet]

Well, one could port the old Linux 8086 project to 4004, it doesn't need such silly things as MMU or 32bits CPUs, though I am not 100% if the project was ever finished. :)

Re:Does it run Linux?

[Anonymous Coward]

Try it out! The Java simulator on modern hardware should simulate it almost as fast as it ran 35 years ago in silicon.

Try Debian

[The_Abortionist]

Debian will probably catch up to it in a year or two.

Re:Does it run Linux?

[Anonymous Coward]

if you coupled it with a modern graphics card you should be able to use the 4004 to bootstrap linux into the graphics card and run it from there!

Re:Does it run Linux?

[springbox]

640 addressable bytes of memory should be enough for anyone....

Re:how about minix ?

[TheRaven64]

It couldn't run Minix, and it would be quite hard to port Minix to it. It already runs on 8086 CPUs, so it doesn't need an MMU (or an FPU). Originally it came with 40-bytes of RAM, which is certainly not enough for Minix. It supports 12-bit addressing though, so you can address 4K-words. Unfortunately, the word size is 4-bits, so that means you can only address 2KB of RAM, which is definitely not enough for Minix. For reference, Bash is about 284 times bigger than the entire address space of the 4004. If you tied it with a custom MMU chip, you could possibly extend this to 4096 segments of 4096 words, giving you 8MB of total address space. This would be enough for Minix, but you'd need to do a lot of paging, which would slow down the performance of the 4004 chip a lot. It would probably boot in under a week...

Re:how about minix ?

[Amadodd]

Operating systems are for sissies.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Re:how about minix ?

[drinkypoo]

I would think that most devices like that use an AVR or a PIC. Certainly that's been my experience.

Re:how about minix ?

[spaceyhackerlady]

I would think that most devices like that use an AVR or a PIC. Certainly that's been my experience.

They do now. But what did they use prior to that?

Intel really did start something new with the 4004. Anybody who minimizes the effect it had is just plain silly.

I had the 4004 manuals at the time, but never had the opportunity to play with the chips themselves. Of course, now it's easy to emulate one in software. I run Unix V5 and V7 on a simulated PDP-11, strictly for the hell of it.

...laura who wouldn't mind owning a real PDP-11, but who refuses to pay the electricity bills for a VAX

Re:how about minix ?

[drinkypoo]

Prior to that they mostly seem to have been high-volume products using purpose-designed ICs. Prior to THAT they may have been using the 4004, but it's been a while. And actually, the motorola processors have also been known to be immensely powerful for embedded applications, they tend to require less support hardware but the parts themselves are often more expensive which as far as I can tell is the only thing that stopped them from completely creaming intel. It was a long time after motorola got into the game that intel actually had anything superior again.

These days of course as we all know microcontrollers are dirt cheap.

If I were going to do any serious fiddling with them today I'd probably go with the AVR just based on the ease of programming and the overall functionality provided by the chips. The PIC is just about as capable it seems (although not as fast last I checked) but the AVR's programming langage is less funky, or so I'm led to believe. Certainly my one experience with programming my atMEGA was pleasant. I'm not much of a programmer though, I have to admit; I have a bunch of experience writing lame programs and one class in x86 assembler, and that's it.

Still, unless you want to solder together your own parallel port programmer, which is not difficult of course but is tedious, I'd just go with the AVR kit. You can get a development kit for like $100 that comes with a nice programming board that can program the full range of AVR chips, and comes with a couple samples so you have something to play with. The board has some buttons and LEDs so you can get started messing with it immediately. I also ordered a cute little board that functions as a name tag, and has a handful of sensors on it - heat, voltage, and light. You can download full schematics so it provides sort of a jump up on the typical sort of hobbyist activities.

The days of the Nibble...

[msobkow]

I can't say I miss the days of the nibble and CPUs measured in kilohertz.

Re:The days of the Nibble...

[Ant P.]

I do, because back then bloatware wasn't an option.

Re:The days of the Nibble...

[msobkow]

Sure there was bloatware -- so people upgraded to the 8080.

Re:The days of the Nibble...

[nurb432]

Well, some of us do and think todays processors are overkill and needed mainaly due to people having forgot how to code properly.

They still exist and are important

[EmbeddedJanitor]

4 bit cpus costing a few cents. Used to control rice cookers etc.

Moore's Law works both ways. Sure, you get faster CPUs, but you also get cheaper bottom end CPUs. People with the skill to design tight software to run on these can use micros in systems where they were not deasible in the past.

Zzzz

[KNicolson]

Get back to me once you've ported Linux to it.

And imagine OGG supporting a Beowolf cluster of them in Soviet Russia.

Re:Zzzz

[jpardey]

I can imagine that one would... profit!

Re:Zzzz

[Scarletdown]

Get back to me once you've ported Linux to it.

And imagine OGG supporting a Beowolf cluster of them in Soviet Russia.

Well, Belgium! You had to go and use up most of the old standbys yourself. But you missed at least one...

I, for one, welcome our 4 bit overlords.

Re:Zzzz

[tgd]

Now just dupe your post.

Fast-forward

[jmv]

Who would have guessed how right they would prove to be?

Who would have guessed chips produced 35 years later, would still inherit the brain-damaged ISA of the 4004. (OK, so the ISA probably didn't look too bad when it was for the 4004)

Re:Fast-forward

[the_humeister]

Actually, they're not the same. The 4004 has 46 instructions [pldos.pl]. The 8086 [wikipedia.org] has quite a bit more instructions and pretty much started us all on the x86 ISA, which weren't binary compatible with programs written for Intel's earlier processors.

Re:Fast-forward

[jmv]

While not binary compatible, the 8086 [wikipedia.org] was a 16-bit improvement of the 8-bit 8080 [wikipedia.org], which was compatible with the 8008 [wikipedia.org], which AFAIK wasn't too far from the 4-bit 4040 [wikipedia.org] and the 4004 [wikipedia.org]... and that's why the space shuttle's boosters are sized according to a horse's rear end [astrodigital.org] and a 64-bit quad core CPU architecture that is influenced by the first 4-bit microcontroller.

Railroad gauges

[Beryllium Sphere(tm)]

Snopes says not quite [snopes.com]. Though the lesson of the story is true and profound.

Re:Railroad gauges

[Cadallin]

I really rather disagree with their conclusion. Although it was not "inevitable" the fact of the matter is that the rail road gauge that became dominant in the USA and Europe CAN be traced to the one adapted for rail use from carriages designed to fit on roads built to a standard specified originally by the Roman Legions based on the width of the asses of two standard war horses. That this is merely coincidental doesn't make it any less true, or less telling about the nature of beaurocracy and resistance to change. And the fact of the matter is that the standard does continue to affect rail shipping to this day, as it most definately determines what an oversize rail car or load is. Whether or not this actually had a direct impact on the Space Shuttle's SSRB's is less clear, although certainly they had to be designed so that they could be shipped from the factory to Cape Canaveral.

The thrust of the point to me, is the very point that nobody sat around and actually considered what might be a good rail gauge to adopt for shipping lines, they just went ahead with a horribly odd standard that was already in existence.

Re:Railroad gauges

[johnw]

The thrust of the point to me, is the very point that nobody sat around and actually considered what might be a good rail gauge to adopt for shipping lines

One man did. Isambard Kingdom Brunel did exactly that. He sat down and thought about what gauge to make his railway (The Great Western) and came up with 7 feet as a much more sensible value. He was entirely correct, but unfortunately his version was abandoned simply because far more people had used the existing default.

John

Re:Railroad gauges

[Teancum]

Keep in mind that the gague of the railroad has a strong influence on how tight you can have a turn radius when bending track. Larger gague tracks simply require much more room to turn. Over flat prarie and meadows this isn't that big of a deal, but when you get into mountains or along a sea coast it becomes a huge deal.

For this reason alone, many of the mining railroads actually use a standard "narrow gague" for their tracks (and even a "cog" railroad to overcome the steep slope of the tracks). There is less room to turn for things like a switchback track, so therefore they need a narrower gague of track.

The point here is that the "standard gague" was something that came about through years of experience of having to deal with a large number of factors, which met the compromise from all of the competing problems. If 7 feet really were that much better, it certainly would have been used more often.

Re:Railroad gauges

[rufty_tufty]

Not quite:
Such a wide guage had a number of problems; namly its ability to turn corners fast (not much use for the north of england which is reasonably hilly and used for much of the frieght at the time because of the industry around there) and the difficulty of operating points on such a system. Not that these problems weren't solvable, but like all things in enginerring it's a compromise to best fit your current problem.

Re:Railroad gauges

[johnw]

Good job Brunel only had to route his line through the flat prairies of Cornwall then.

Re:Railroad gauges

[nuzak]

> I really rather disagree with their conclusion.

You can keep on believing whatever you want. It doesn't change the fact that the wheel width of wagons wasn't set based on the size of two horses asses. Yokes attacking to carriages are designed to to have a lot of play, and could and did pull everything from two-horse imperial chariots to the rickety donkey-drawn haywagons that brought in the goods sold every day. Hell, those wagons probably wore far more ruts than the imperial legions did, because the legions generally either marched or rode horseback. Chariots were more for show. And having been around four-legged critters from shetland ponies to clydesdales, I can assure you that there is no standard gauge for a horse's ass.

Re:Railroad gauges

[johnw]

I can assure you that there is no standard gauge for a horse's ass.

...or even for an ass's arse

Re:Fast-forward

[Daniel Phillips]

While not binary compatible, the 8086 was a 16-bit improvement of the 8-bit 8080, which was compatible with the 8008, which AFAIK wasn't too far from the 4-bit 4040 and the 4004

Indeed. Does this instruction ring a bell? Decimal adjust accumulator DAA [pldos.pl]

Re:Fast-forward

[lenski]

The 8080's registers can be mapped directly from the 8008's registers: A, B, C, D, E, H, L. HL taken as a pair could be used to address anywhere in the 8008's 14-bit address space.

Intel's promise that the 8086 had upgrade compatibility with 8080 (or the 8085), came straight from the success of the 8008->8080 transition.

All this talk about ancient processors... I wrote a debugger for an 8008 system in 1976. A collection of geeks in college borrowed an 8086 development kit from the local Bell Labs office, and we did it with an ASR-33 Teletype. I forgot to return the 8008 manual, and still have it. The *fast version* ran at 200 KHz, and each instruction executed in a multiple of 5 clocks (single-byte instructions in 5 clock cycles, two-byte instructions in 10 clock cycles...)

I remember marveling a few years ago that one could trace the 8008's registers and capabilities right up through the latest Pentium4.

Re:Fast-forward

[Technician]

Who would have guessed chips produced 35 years later, would still inherit the brain-damaged ISA of the 4004

Didn't ISA come out with the IBM using the 8086? The 4004 was more suited to things like a calculator.

I did look it up.

http://en.wikipedia.org/wiki/Industry_Standard_Arc hitecture [wikipedia.org]

IBM PC XT ISA = Industry Standard Architecture released in 1981.

The Intel 4004 processor was first fabricated in 1971 a decade before the ISA buss.

http://www.intel4004.com/ [intel4004.com]

Please don't re-write history. Blame IBM for ISA, not Intel.

Re:Fast-forward

[0racle]

ISA has many meanings

ISA - Instruction Set Architecture

There are others of course, but I just don't see how the Irish Sailing Association is relevant here.

Re:Fast-forward

[Doctor Faustus]

"Industry Standard Architecture" (the old 16-bit expansion cards that coexisted with PCI on new motherboards until around 1999) could be relevant.

Re:Fast-forward

[0racle]

When the subject is Intel's 4004 it couldn't be.

Re:Fast-forward

[Mitchell Mebane]

ISA, as in "Instruction Set Architecture". Not the bus.

Re:Fast-forward

[LWATCDR]

What people are complaining about is the "Instruction Set Architecture" not the old ISA buss that PCI replaced. Some of us miss the good old ISA buss because it was so easy to hack and interface to :(

You are in a way correct even if it is for the totally wrong reason. IBM used the 8088 for the PC not because they loved it but because it was cheap. If they had known that that it was going to be a run away standard that we would be living with to this day they would have never made it.
If IBM knew then what we know now they would have bought Microsoft and probably Digital research. At the time they could have bought both for less then they spent on coffee and pens each year. Or they would have written their own.
They would have never used the 8088. They would have developed their own CPU. Probably a 16 bit subset of the 360 or an early version of the 601.
The PC was a test system thrown together from a lot of spare parts. They used the 8088 because it was cheap an they already used it in their Display writer dedicated word processor. Since CP/M was already a sort of standard they wanted to use that and Microsoft's Basic. Digital blew the meeting and Microsoft jumped in with a clone of CP/M.
So yes we are living with Microsoft's and Intel's warts because IBM chose to build a quick test product to see how it would do in the market.

Of course it might have been worse if IBM had known. We would have an IBM that is as big and powerful in the market place as IBM, Microsoft, and Intel combined! Back in the day IBM was the evil empire. Frankly they made Microsoft look really good.

Re:Fast-forward

[mspohr]

If IBM had:

- developed their own CPU instead of the "cheap" 8088

- bought Microsoft and Digital Research or "written their own" OS

Then the "IBM PC standard" would be in the same boat as the DEC PDP-8 and would be ancient history.

Re:Fast-forward

[LWATCDR]

I really doubt it.
IBM could have sold both the OS and the CPU to clone makers and probably would have to keep the anti-trust off their backs. At the time IBM was in real danger of being broken up because of anti-trust litigation. That is one of the reasons that the System 36/38 was so different from the 360/370. IBM was getting ready to be split into different divisions if they lost.
Even before the clones the PC was selling like hot cakes. It over took every other system as the business system to have back in the day. Lotus 123 killed Visicalc as the the killer app. That alone could have kept IBM in the lead for a long time.
Even when cheaper and faster clones came on the market the PC sold well. I think you are way under estimating the power that IBM in the day when you used Lotus 123 and Microsoft Flight Simulator as the benchmarks for PC compatibility.

BTW the PDP-8 really predates the PC. The PDP-11 or even the VAX where closer to the PC. In fact DEC even made a micro version of the PDP-11 I think it was called called the LSI-11. We even have an old system around here that uses that CPU.

Re:Fast-forward

[mspohr]

My point was that an IBM proprietary CPU and OS would not have been widely adopted even if IBM choose to sell it to OEMs (not likely given IBMs business model at the time).

The reason the IBM PC became so popular was that IBM didn't control it and anyone could make a clone. If the clone makers all had to go to IBM for permission and parts, the market wouldn't have developed.

The DEC PDP-8 (and PDP-11) were arguably better architectures and had arguably better software and were implemented (somewhat belatedly) as microprocessors but they were ignored. I remember that when DEC did finally come out with their own PC clone, they were charging a premium for formatted floppy disks. They didn't provide a formatting program for floppies and charged you extra for their "special" formatted floppies. Needless to say, their attempt to control this small bit of the market failed and their PC clone failed and their company failed. They had great computers but they were proprietary and expensive and they failed to develop a broad market.

The IBM PC succeeded because it was not proprietary. IBM couldn't control it. The IBM PC was, at the time, a "rogue" project that didn't follow IBMs rules and it failed to follow the IBM business model (because IBM didn't feel that it would amount to anything).

Re:Fast-forward

[LWATCDR]

"The IBM PC succeeded because it was not proprietary. IBM couldn't control it. "
IBM did everything it could to control it.
They tried to sue people for copying their BIOS. Phoenix and Compaq went to great pains to legally reverse engineer the BIOS. The PC was wildly successful before the clones ever came out. It was much more propritary than the mass of faster and cheaper CP/M machines on the market at the time.

-8: no way; -11? Go with 68k...

[lenski]

After writing 10k+ lines of code on the PDP-8, I can tell you that there's no argument that could convince me that it was a better architecture than the 8088/8086.

One could almost sustain an argument that the PDP-11 was a better architecture than the 8088/8086 as far as it went, which at the time could be extended to a Mbyte of physical RAM (later instances of the -11 could get all the way to 4 Mbytes physical, as in the 11/73). Manipulating the APRs (active page registers), one could do some mild hackery to make a single process address more than 64kbytes.

Through it would have changed computing history fairly dramatically, I believe the MC68008, an 8-bit-bus variant of the MC68000 would have been a much better choice for IBM's new play-toy in the long run. The reason is the 68k began its design life with 32-bit registers. (IBM was still king of the mainframe hill with their mainframes back then, and they might be forgiven for not taking this new toy-sized computer seriously.)

But I will not second-guess IBM, since their inspiration was the Apple][ and they would have to be very forward-looking indeed to apply Moore's law 10+ years into the future.

4004 tic tac toe

[Salvance]

The 4004 tic tac toe hardware from their unofficial site looks wicked ... http://mywebpages.comcast.net/jsweinrich/ [comcast.net]. I never thought I'd be drooling over electronic tic tac toe!

Not to bust this guy's work, but...

[voidptr]

The 4004 has less than 2300 transistors, the lowest end Spartan FPGA (since I don't see the exact part # on there) is 40,000 gates (which is somewhere in the neighborhood of 160,000 transistors).

You could do TTT in the FPGA on that board with room to spare. You could probably re-implement the 4004 ISA itself and his glue logic inside that FPGA.

Re:4004 tic tac toe

[Tumbleweed]

Tic tac toe? A strange game. The only winning move is not to play.

Re:4004 tic tac toe

[ConceptJunkie]

No, you're thinking about Global Nuclear War and politics. The only winning move in Tic-Tac-Toe is to play a 6-year-old who's never seen the game before.

Re:4004 tic tac toe

[mgblst]

Extra points because it could detect my name. Brilliant!

John

640k

[Aehgts]

Ah, back in the good old days when 640K _was_ enough for anyone...

Re:640k

[Reality Master 101]

Ah, back in the good old days when 640K _was_ enough for anyone...

Dude, my first computer had 256 Bytes (not K -- *BYTES*) of memory (Built form the September 1976 issue of Popular Electronics -- Build Your Own Microcomputer, based on the COSMAC 1802 processor). 640K was beyond freaking imagination.

Re:640k

[Billly Gates]

Offtopic but I heard Weirld Al sing in New York a few years ago with the parody turkey on rye (Or pastrami). Now chicken pot pie. You may want to search for that song instead.

Re:640k

[DerekLyons]

Ah, back in the good old days when 640K _was_ enough for anyone...

Dude, my first computer had 256 Bytes (not K -- *BYTES*) of memory (Built form the September 1976 issue of Popular Electronics -- Build Your Own Microcomputer, based on the COSMAC 1802 processor). 640K was beyond freaking imagination.

Yep. And the computer that controlled the Apollo spacecraft (designed while Billy Boy was still in single digit years), wasn't much better than your homebrew (around 8K IIRC). The fire control system I worked on in the Navy (I was in over the eighties, the system was designed in the seventies) controlled sixteen missiles (each with eight warheads) using around 512K.
 
Fact is - most folks (including programmers) forget, or never knew in the first place, how much can be done with quite little. Cheap memory, fast processors, and cheap hard drives encourage sloppy habits.

Re:640k

[MichaelSmith]

Yep. And the computer that controlled the Apollo spacecraft (designed while Billy Boy was still in single digit years), wasn't much better than your homebrew (around 8K IIRC).

According to the Apollo 15 ALSJ the descent guidance system had a five vector model of the terrain around the landing site.

Re:640k

[hughk]

The 1802 was designed 'down' as a controller for unfriendly environments such as space. It was also completely CMOS so used very little power. Although the instruction set was 'quirky', the processor was quite important becuase of the applications. Of course, the 4004 and later the 4040 were 'made' by washing-machines. That is to say if you build something cheap enough to replace the elctromechanical mechanisms (drum-timers) in domestic equipment - such as washing machines then your market is huge.

Re:640k

[clickclickdrone]

>640K was beyond freaking imagination.
When I got my 48k Atari 400 back in early 1981, I couldn't get my head around how vast 48k was so I typed in 48k of rem statements then hit 'list' to watch it scroll by. Took quite a long time. I remember thinking 'Wow, so much space! I could do anything with that amount of data/program'
It also hurt my fingertips due to the 400's touch sensitive keyboard.

Re:640k

[Reality Master 101]

When I got my 48k Atari 400 back in early 1981, I couldn't get my head around how vast 48k was so I typed in 48k of rem statements then hit 'list' to watch it scroll by. Took quite a long time. I remember thinking 'Wow, so much space! I could do anything with that amount of data/program' It also hurt my fingertips due to the 400's touch sensitive keyboard.

I had a TRS-80 with 48K around that same time (1980), and I remember thinking it was IMPOSSIBLE to fill it up with enough program and just DIMing "big ass" arrays of 24,000 short elements (Kazowee!). I also remember a few years later sitting in front of an Alpha Micro computer that had a 68K processor and a gargantuan 1 megabyte of RAM. I was shaking my head at the whole concept (cue Dr. Evil, "One MIIIIIILLION bytes!"). :)

Re:640k

[Tuna_Shooter]

I built the same system as you. (Still works) Also the RCA 1802 was the main processor used on the Voyager and Viking space probes.

Re:640k

[Reality Master 101]

I built the same system as you. (Still works) Also the RCA 1802 was the main processor used on the Voyager and Viking space probes.

In the interest of full disclosure (and it's kind of funny), I have to admit that my construction from that article was a failure. I was 13 years old, and it was a tad beyond my not-so-m@d skilz. I saved up my allowance and bought each part as I could afford it, then when I had all the parts, I made the bright move to solder wires directly onto the chips, rather than using sockets. :D

I recall being very shocked when I turned it on for the first time and it didn't work. I then really looked at my creation, which was a rat's nest of wires snaking across the back of the prototyping board, and realizing that maybe, just maybe, I might've made a mistake in the wiring (it hadn't occurred to me up to that point that there was any possibility that debugging would be required). Faced with an utter, hopeless mess, I gave up, and bought the kit that had come out since then with a nice motherboard I could solder to. That one worked a lot better. :)

Re:640k

[TheRaven64]

640KB would have been a luxury. The 4004 had a 12-bit address bus, and so it could address 4K-words. Each machine word was 4-bits long, so it could address 2KB of RAM. This wasn't a huge limitation, since it only shipped with 40 bytes of RAM, and no MMU.

Re:640k

[morcego]

Considering my first computer had 2K of RAM, I would consider 640K very nice.
And yes, it was based on Z80.

a bit of relevant info....

[frakir]

pasted from http://www.cpu-world.com/CPUs/4004/index.html [slashdot.org]> :

The first microprocessor in history, Intel 4004 was a 4-bit CPU designed for usage in calculators, or, as we say now, designed for "embedded applications". Clocked at 740 KHz, the 4004 executed up to 92,000 single word instructions per second, could access 4 KB of program memory and 640 bytes of RAM. Although the Intel 4004 was perfect fit for calculators and similar applications it was not very suitable for microcomputer use due to its somewhat limited architecture. The 4004 lacked interrupt support, had only 3-level deep stack, and used complicated method of accessing the RAM. Some of these shortcomings were fixed in the 4004 successor - Intel 4040.

More Relevant Info?

[octalman]

Somebody correct me if I'm wrong, but as I recall the 4004 wasn't a single-chip microprocessor. Depending on the chip set used, it took from two to four chips to put together a working microprocessor.
 
Intel's first shur-nuff single-chip microprocessor was the gosh-awful, horribly slow 8008. They took so long to get past the 8008 and the only marginally better 8080 that Zilog brought out a much-improved, instruction set compatible version, the Z80, which dominated the microprocessor market for a number of years.
 
The first true computer-on-a-chip was Motorola's 6800, but they muffed their opportunity by waiting too long to market it and priced it too high. Worse, some employees stole their chip masks and modified the design, which they sold (cheaply, compared to the 8008 and 6800) as the 6502, which was adopted for the Apple. Motorola sued and got the 6502, which they continued to sell, but lost years of opportunity and the chance to dominate the whole market.

Re:More Relevant Info?

[turly]

Dunno what you're smoking, fella, but Motorola never "got" the 6502. From this article [wikipedia.org]:

The 6502 was designed primarily by the same engineering team that had designed the Motorola 6800. After quitting Motorola en masse, they quickly designed the 6501, a completely new processor that was pin-compatible with the 6800 (that is, it could be plugged into motherboards designed for the Motorola processor, although its instruction set was different). Motorola sued immediately, and MOS agreed to stop producing the 6501 and went back to the drawing board.

The result was the "lawsuit-compatible" 6502, which was by design unusable in a 6800 motherboard; Motorola dropped their objection.
...
The 6502 was introduced at $25 in September 1975, when the 6800 and Intel 8080 were selling for $179. At first many people thought the new chip's price was a hoax or a mistake, but shortly both Motorola and Intel had dropped their chips to $79. Far from the intended result, these price reductions actually legitimized the 6502, which started selling by the hundreds.

Re:More Relevant Info?

[Alioth]

Indeed - you can still buy a plain, newly manufactured Z80 processor today. It's still popular in embedded applications. There's also a microcontroller version of the Z80, the eZ80.

Re:More Relevant Info?

[$RANDOMLUSER]

> Depending on the chip set used, it took from two to four chips to put together a working microprocessor.

That's the way I remember it too. I seem to recall (at least) something about a multi-phased clock, and an address decoder thingie. I still have Intel databooks from that era, but I'm too lazy to look it up.

Back in 1979/80 or so, I worked on readying for manufacture a prototype slot machine that Harrah's Casinos (there were only two then, Reno & Tahoe) had built using the 4004. Intel had quit making the 4004 by then, so the choice came down to buying the remaining supply, or do it over using a "real" (Z80) processor. The project never really got off the ground.

Re:More Relevant Info?

[julesh]

The first true computer-on-a-chip was Motorola's 6800, but they muffed their opportunity by waiting too long to market it and priced it too high. Worse, some employees stole their chip masks and modified the design, which they sold (cheaply, compared to the 8008 and 6800) as the 6502, which was adopted for the Apple.

Err... that doesn't sound right to me.

First: 6800 was released after the 8080, which really was a single-chip processor.
Second: the 6502 was actually an innovative design, being the first microprocessor to feature an instruction pipeline, even if it was only a 2-stage one; doesn't sound like a mask-level rip-off to me.

Re:a bit of relevant info....

[leob]

I've asked that question Monday night diring the anniversary meeting at the Computer History Museum. They said "none", then, jokinly, "NOP". I haven't verified, but I believe their first answer was right.

Digital archaeologists

[Timesprout]

wipe wipe
"early gang bang porn, log it"
wipe wipe
"early vivid movie, looks like Jemma was young and need the money, log it"
wipe wipe
"some girl on girl stuff, log it" wipe wipe
"holy crap I am taking this home"

Re:Digital archaeologists

[msobkow]

Before DivX pr0n there was MJPEG pr0n.

Before MJPEG pr0n there was JPEG pr0n.

Before JPEG pr0n there was bitmap pr0n.

Before bitmap pr0n there was ASCII art pr0n.

Before that, some weirdo was convinced that two LED's looked like nipples...

*g*

Re:Digital archaeologists

[earthman]

You forgot the GIF porn.

LED porn?

[Dion]

Like this?

http://www.gizmodo.com/archives/circuits-discover- each-other-024002.php [gizmodo.com]

Jon Katz could write about it

[Schraegstrichpunkt]

This will cause a social revolution in Afghanistan. People will now be able to build their own 4004-based, and use them to download movies and MP3s against the will of the Taliban...

Re:Jon Katz could write about it

[Jesus_666]

You're kidding, right? With these schematics the terrists could develop their own cruise missile control systems. Intel has doomed the Western world! DOOMED!

Re:Jon Katz could write about it

[o'reor]

Come to think about it, 16x16 pixels is probably the resolution with which you see the world at through the mesh of a burqa.

Somehow this thread is on its way to reach the Allahwin point...

Era of Intel's Ways

[Doc Ruby]

Intel patented the 4004, which they tried to use to enforce a patent on the "microprocessor" generally - though Gilbert Hyatt [thocp.net] eventually won it, 20 years later.

Does Intel still have a working patent protecting the 4004? And doesn't that patent include the schematics? What's the point of patenting an invention if other inventors can't tell whether they're reinventing what you've protected from "infringement"?

Re:Era of Intel's Ways

[stevesliva]

Well, you want a patent to be enforceable agaist others. So either you patent the entire microprocessor concept, or you patent one small invention that all microprocessors need to use. Intel probably patented the binary adder or something.

Re:Era of Intel's Ways

[KokorHekkus]

...Does Intel still have a working patent protecting the 4004?...

Hardly since patents are for a limited time only and at that time I think the time was 17 years from the patent being granted.

Re:Era of Intel's Ways

[Doc Ruby]

Do you really believe that patents are only for a limited time? You do know that the patent on the microprocessor, invented in 1969, to which I linked, was granted only in 1990, right? And that's just one guy. Intel has the patent army of Genghis Khan.

And even if they didn't, that's all the more reason that the 4004 schematics etc shouldn't be secret or private. They'd be public domain by now.

Re:Era of Intel's Ways

[Doc Ruby]

That's what I said.

Though your story about documenting/dating work prior to filing a patent is wrong. Only if that documentation is either published or entered in certified notebooks obtained in advance from the PTO can the work prior to filing be counted as prior art defending from a later filing (but earlier than one's own filing).

Trade secrets are unnecessary when that info is patented. That's the entire point of a patent.

Although an important starting point

[stox]

The 8008 was the real start of the micro-computing revolution. The Scelbi Mark 8H was the first system to really draw people's attention. By the time they figured out what might be done with it, the 8080's were released. The Altair was built, BillyG and friends wrote a basic interpreter in 4Kbytes, and the rest is history.

Ah, Busicom

[igb]

I recall my father coming home from work (chemistry lecturer in higher education) and saying that he'd got access to calculators. A few weeks later I went over with him and played for a while on a Busicom, nixie tubes and all. This would be about 1972, I think, guessing from which building it was in.

Obligatory

[SeaFox]

Opening on November 15th, the exhibit will feature a fully functional, 130x scale replica of the 4004 microprocessor running the very first software written for the 4004. To create a giant Busicom 141-PF calculator for the museum, 'digital archaeologists' first had to reverse-engineer the 4004 schematics and the Busicom software.

Well I, for one, welcome our gigantic calculator overlords. And remind them that as an internet personality, I could be useful in rounding up citizen's to slave away in their underground button-pushing dungeons.

They have released docs for single core 4004...

[Brane2]

... only to encourage sales of dual core 8008... ;o)

.

No shit they were right...

[davidsyes]

They've got a few Ken Starlings heading departments and a reportedly missing Federation Timeship Aeon sequestered under the buildings.

Janeway will be BACK: for the timeship, the deep-fried alien jerky, AND the KFC chickens. And, she'll pick up a few humons from the White House to supply the Vidiians, cuz she's in NO mood to donate organs today. Fixing the timeline is a byatch!

That's a lot of processors!

[joetheappleguy]

Intel releases 4004 schematics? Man, that's a lot.

Oh, wait...

35 years

[kestasjk]

35 years ago this was the best personal computer you could get. Now the same company is bringing us processors which can simulate the entire thing in an interpreted language using a fraction of one percent of the available processing power.

Even though another company would have done the same if Intel hadn't, they deserve some kudos for getting in there first and staying on top. No-one would have thought they'd be able to push x86 to where it is today.

Replica is very cool, but

[ballpoint]

where are the blinkenlights ?

A custom DSBGA chip simulating a mosfet and including a driver for a tiny SMD LED could have shown the state of each individual gate.

Useless

[smcdow]

Doesn't run Java.

Who cares about the Intel 4004...

[Jesus_666]

...I want the MOS 6581!

hand-drawn circuits

[peter303]

Most chips and circuits boards manufactured before 1978-1980 were drawn or taped by hand. By taping, I mean you pasted a mockup from thin drafting tapes, then photo-reduced them to circuit boards and chips. I did a fair share of them myself. There was a certain artistic statisfaction in doing this, somewhat like rebuilding your own car engine or remodeling a room in your house. I would dream of taping circuits in my sleep as my busy subconsious worked out upcoming issues, much like my brain computes animation grphics now.

Computer-aided circuit design allowed getting past the 10,000 gate level with much less error. The results arent as artistic, but are a lot cheaper to implement and more accurate. CAD required interative computing and a minimum of vector graphics display, both possible by the late 1970s. Kind of incestuous: using computers to design bigger computers, with a dash of human intervention.

The Intel museum

[heroine]

With all the engineering in India, that Intel museum must be the only thing left in the Montegue Expy building. Has anyone actually gone to the Intel museum? Sitting on Montegue Expy for 30 minutes to get to Intel sounds like work.

4096 Processor Array of 4bit 4004 Chips?

[johnrpenner]

with only ~2000 transistors per processor element,
what could one do with a 4096 Processor Array of 4bit 4004 Chips?
hmm...

Re:Reverse Engineer?

[camperdave]

Yes. You see, the original schematics were stored as an EBCDIC art file on punch cards. The chip design itself is remarkable, as it was to be the first chip built with sub-miniature vacuum tubes.

Re:Reverse Engineer?

[Detritus]

It isn't that hard. You start by grouping things into functional units and move to higher levels of abstraction. What can be tricky are any weird and unique circuits. There is quite a bit of structure in the design. What's time consuming is examining the behavior of the instruction decoder for all possible instructions. This can be a problem if you don't have a list of the supported instruction set. On many old processors, there was no trap for illegal or unimplemented instructions. The processor would do something, although it might not be useful.

Re:Xilinx Spartan FPGA

[QuantumHack]

Well, another of Federico Faggin's designs, the Z80, has been implemented in VHDL (and Verilog). I implemented the T80, a VHDL variant, along with a VGA-grade video interface, and a triple-ported SDRAM interface into a Xilinx XC3S1000. The combination only used 3% of resources in the FPGA, but the processor itself was about 1%.

The 4004 had 3900 transistors, and the 8080 had 6000, and the Z80 had more than that (more instructions). So, let's say for argument's sake that the Z80 is about twice the size of the 4004.

If that's true, then you can stuff about 200 clones of the 4004 in a Xilinx $15 million-gate Spartan FPGA, and have block RAMs left over for program memory. Wow, I'm sure the Beowulf guys are scared now ;-)