Home

Oral History Panel on the Development and Promotion of the Intel

1. Oral History Panel on the Development and Promotion of the Intel 8008 Microprocessor Participants Federico Faggin Hal Feeney Ed Gelbach Ted Hoff Stan Mazor Hank Smith Moderated by Dave House Recorded September 21 2006 Mountain View California CHM Reference number X3706 2007 2006 Computer History Museum Oral History Panel on Intel 8008 Microprocessor Dave House Stan tell us how the 8008 began Stan Mazor Well in today s personal computers we have RAM and we have flat panel displays but in the middle of the 60s we had core memories on our minicomputers and we typically used either a teletype or a CRT terminal cathode ray tube We also nicknamed these glass teletypes because they played the role of the teletype but they were much faster The key ingredient to the CRT terminal in those days was the shift register there was no RAM memory in those days Shift register were being provided by General Microelectronics GME General Instrument and some other companies When Intel got started in 1969 in the memory business we were designing RAM components and we really had no customers for them So one of the things that we did was to do a custom order for Datapoint formerly Computer Terminals Corporation who was one of the makers of a cathode ray terminal and we made for them a 512 bit re circulating shift register So this gave us a ready customer and it was a type of memory product which was in use
2. Hoff Well know one of the discussions had with Bob Noyce was how would we support this And this was early 71 or early summer think said thought that my group could maybe support a dozen or so key customers Now that was a good number because that was the number of key customers we had for memory chips And then said that for rest we would provide literature but they d be on their own Now kind of figured in the back of my mind if enough of them came out we would have to do something else to support them But at least it was reassuring Feeney The thing that s very important with that as Hank and everybody were pointing out is that development systems carried the company for a while or carried the microprocessor business for a while We were certainly making greater revenue on development systems than on the individual devices that were going out And at that same time the number of manuals that we were printing was an incredible number and Stan and Ted were going out doing seminars for customers around the U S and then gradually it became around the world But backing up a little bit before that there was a pushback in early 71 where customers like Seiko were saying We want to learn more about what you re doing in microprocessors We have an interest in using those processor products There were customers from Europe and customers from Japan that were expressing interest very early on even before we had the devic
3. he started several projects One of them was a video game that was played on a television set using MCS 4 and we actually had built a primitive Space War game with objects moving on the screen And in an applications engineering group the idea is can you find new applications that you d kind of like to go into business with Well management not being familiar with that application thought it was pretty silly to play a game on a television set And similarly another application we talked about was building a scientific calculator I m talking about a handheld small scientific calculator And recall an upper management man had a slide rule on his belt and said Well why would you ever want one of those And so would say we Intel applications research had some visions and some of those visions were subsequently fulfilled by other companies and our customers Gelbach The microprocessor was a boon to Intel in non obvious ways Number one think we looked at it because the customers were very interested in it We would get more requests to talk about it Nobody wanted to talk much about memories They understood the memory sequence was going to go from one K to two K to eight K whatever it was but the microprocessor tickled the imagination of all of the engineers Every customer that we ever had for whatever product always wanted to talk microprocessors It was the entry door We used to give seminars that we even
4. that 4004 is the computed age of the universe if you look at the Bible Of course didn t know at that time but it was very interesting to find out Eventually wore down Vadasz and he gave in He said Okay okay Okay do whatever you want But my logic was that it was a family so it should have a sense of family number one Number two there was a ROM and a RAM and a shift register and a CPU so they would have to be called something different according to the system with names like 1202 for the CPU 1305 for the ROM and 1107 for the RAM or whatever It would have been a mishmash and logic finally won the day House When did the 1201 get converted to the 8008 Faggin think it was in marketing don t know who Probably you had the idea Smith Yeah and then also the MCS 4 and the MCS 8 That moniker was put on to try and tie all of them together and not give the impression they were individual devices but a family of devices that worked together As a matter of fact for the MCS 8 we kind of bastardized that because there wasn t really a family Feeney We created it Smith We created it and made that also a family of devices although there were standard memory products in there Feeney Following on that because of the low speed of our processor and the demands of it it didn t demand the highest performance RAM it didn t demand the highest performance ROM so it gave us the opportunity of designat
5. ve got a number of applications here in the advertising brochure and these weren t applications that were envisioned these were ones that customers actually did We put together this brochure of Intel microcomputer applications of the Seiko calculator a blood analyzer a business machine and think this was from Comstar a very large mechanical device or mechanical control device We had a point of sale machine from Staid a communication management device from Action Communications another point of sale device from Pitney Bowes Alpex Smith Point of sale was a major application Feeney Also an Inforex check processing machine So these were all early thoughts early concepts of how things could be done It was amazing how quickly the customers latched onto this The date on this brochure is April of 1973 so it s basically a year to year and a half after our announcement of the products that these devices the finished products were in customers hands Hoff We worked very closely with Regis McKenna on developing the advertising and support material And they came up with some pretty clever Feeney There was a traffic light ad right House remember talking about traffic lights and blood analyzers CHM Ref X3706 2007 2006 Computer History Museum Page 16 of 30 Oral History Panel on Intel 8008 Microprocessor Gelbach Traffic lights were a big application at the beginning because we needed something to talk abo
6. If you see the layout of the 8008 and the 4004 unless you know what you are looking at you cannot tell them apart Feeney Being able to piggyback on many of the tools was just incredibly helpful for getting design moving along And the other thing you point out is the fact that being limited on pins limited on power supplies whatever that the bootstrap load became very very critical Faggin Yeah it was essential We worked together for the first month or so Basically transferred my experience to Hal and then Hal pretty much went on his own Occasionally he needed some help in some issues but fundamentally Hal did the entire design by himself When it came out toward the end of the year toward the end of 71 everything was working except for a few errors And by that time Hank Smith was after me because he wanted Hal in marketing He was coming to me every week saying You know Hal has to come over Hal has to come over And said Well you know he s got to finish the project Then he can come over And eventually Hank Smith did win finally Faggin He won Hal went to marketing before the 8008 was totally completed and characterized The only hard time that got with that was that after Hal left found a problem with the memory That got me some sleepless nights basically the memory was losing its memory It was a dynamic memory and it was charge injection that had to figure out a way around Final
7. PL M was an operating system Smith The forerunner CHM Ref X3706 2007 2006 Computer History Museum Page 13 of 30 Oral History Panel on Intel 8008 Microprocessor Mazor PL M 8 was a programming language which was written as a cross compiler for the 8008 And the 8008 was a pretty primitive assembler language and PL M 8 was a pretty primitive high level language actually Gary Kildall worked alongside us at Intel and one of the things we did is provide him with a floppy disk as a form of payment and for the floppy disk he developed an operating system And think our development of ISIS inside Intel sort of paralleled the work that he was doing And that operating system of Gary Kildall eventually developed to become CP M at that time also had a PL M compiler that ran on it his operating system And it s quite possible that the operating system itself was written in PL M don t remember that myself We Intel had a tight relationship with Gary but of course he founded Digital Research and never did join Intel although we approached him on numerous occasions to become an Intel employee Smith What followed after that was the development of the Intellec systems the Intellec 4 and the Intellec 8 which were basically primitive personal computers if you want to call them that They were the forerunner of the MDS system and then of personal computers They were designed in our group and were marketed as a se
8. Panel on Intel 8008 Microprocessor use of a call to some routine that would be loaded instead of whatever was coming from memory when you serviced an interrupt And then we looked at the amount of logic that it would take to implement that outside the processor and it looked ferocious But we had this restart instruction which was one byte instead of three and we figured it would be a lot easier if that were a call And it turned out the way the processor was it looked like it didn t cost us anything in terms of utility if we made restart a call And so that became one of the reasons for the change Feeney Right And it became a natural fit into the instruction set and very very true no overhead cost at all because we were able to deal with it in exactly the same manner that we were dealing with other elements in the processor House So you basically faked an interrupt by forcing in a jump instruction Hoff Exactly Feeney Yeah but it was a one byte Hoff It s got to save the contents Feeney Yeah save the content rather than just moving But it was a one byte instruction And with three bits then directing the user or directing the program to one of eight different locations in memory where a legitimate call would then be inserted into the system But the one thing that s very important with the 1201 the 8008 in contrast to the 4000 series is that the 4000 series had its own special memories special ROM special RAM
9. about a serial machine you have to process all the addresses and data one bit at a time and the rational way to do that is low bit to high bit because that s the way that carry would propagate So it means that in the jump instruction itself the way the 14 bit address would be put in a serial machine is bit backwards as you look at it because that s the way you d want to process it Well we were gonna built a byte parallel machine not bit serial and our compromise in the spirit of the customer and just for him we put the bytes in backwards We put the low byte first and then the high byte This has since been dubbed Little Endian format and it s sort of contrary to what you d think would be natural Well we did it for Datapoint As you ll see they never did use the 8008 chip and so it was in some sense a mistake but that Little Endian format has lived on to the 8080 and 8086 and is one of the marks of this family Feeney The design hiatus that we had gave us an opportunity to go from that July timeframe to the following January and accomplish several things One the original design was intended to go into a 16 pin dual inline package or 16 pin DIP The reason for that was because Intel was in high volume production with memories at that time and production management wanted us to stay with the most popular package that was being used at the time Good fortune had it that over the next few months the 1103 was
10. at that time Well Datapoint having been in the display terminal business had an idea of going into the business terminal computer business in addition and they had a clever idea which was to use the shift registers that they were already using inside the display terminal as the main memory of their computer Now rotating memories had been used in earlier computers usually magnetic rotating memories and so they had come up with an idea of using the shift register that we were making for them inside a business computer as the main memory But inside that computer they needed some registers and they needed something called a pushdown stack and we had also made at Intel a bipolar memory 64 bits that was specified by Honeywell Honeywell said We ll buy these from anybody who can make them and almost everyone in the industry was trying to make this bipolar 64 bit memory Well Victor Poor from Computer Terminals looked at that chip and said Well could implement the registers of my 8 bit computer and that and could also implement a pushdown stack if you guys could add a counter to it So he met with Andy Grove in December of 69 to check on the status of his shift register order and in that meeting he mentioned to Grove Can you make a stack for me And Andy Grove really didn t know what a stack was and as an applications engineer was called in to be a consultant on that So met with Victor Poor around Christ
11. computer and we ll never be able to do that but actually when you look at it the instruction decoding logic doesn t need to know how big the data word is So our preliminary estimate was that if we simplified his instruction set a tiny little bit that we could probably build a 8 bit CPU on a chip to his specification and we made a proposal in around the March 1970 timeframe to do that They came back with the salesman and then as best as recollect the salesman and the buyer had an agreement that we would deliver a hundred thousand chips at 30 apiece and those would be the CPU chips Of course they were accustomed to doing high volume memory chip deals Probably computers were being sold from Digital Equipment Company five thousand or ten thousand a year so a hundred thousand was a pretty ambitious quantity So we embarked on a custom chip for Datapoint following the 4004 principles in general and that was how we got started in the 8008 which we designated internally as the 1201 House So what happened next Mazor Well we had to have a design engineer assigned to the project and Hal Feeney came on board and was my officemate for a while My job was liaison trying to explain to him what we were doing and what the chip had to do And he was a chip designer I was not so was there kind of as a helper and translator and we began to work on the general layout And think that probably Hal can tell the story best from here on Hous
12. designing the 4004 and had come up with some modules and methodologies in silicon technology that wound up being applied Federico could you tell us about how you moved from the 4004 and your involvement with the 8008 and your contributions to the 8008 Faggin Well was hired in April of 1970 just about a month after Hal In fact believe it was April ga And I was the leader of what was called the Busicom project in those days It had no other designation and was essentially four chips that consisted of the 4004 the CPU 4 bit CPU and ROM and RAM and a shift register Soon after joined also found out that there was another microprocessor project going which was the 1201 the 8008 And met Hal not as soon as got there because Hal was getting married and he was on a honeymoon and so when he came back from the honeymoon met him Feeney We met because we were sharing pullout trays on the same desk Faggin Yeah we had very very narrow accommodations But anyway found out that the 8008 1201 was also another project going on Actually felt a little jealous of Hal because Hal had one chip to design had four The 4004 was going to be the last one of the four And so thought that Hal was going to have the first microprocessor on the market But you know had enough to worry about in doing my own stuff and so soon forgot about it But basically with the design of the MCS 4 the Busicom project as it was later calle
13. engineering group that was knowledgeable in some of the computer areas Prior to that it was more discrete memory Then we changed the concept of the sales force but more predominantly the application group to could go in and talk computer language to all of our customers which most of us at that time could not CHM Ref X3706 2007 2006 Computer History Museum Page 14 of 30 Oral History Panel on Intel 8008 Microprocessor Feeney think our first application engineers were hired in the 72 or 1973 timeframe Gelbach thought it was earlier Hoff think it was earlier than that yeah Gelbach had kind of started the concept of application engineering at TI and had hired them there as component people And think I started it at Intel relatively soon after joined So would have guessed probably early 72 mid 72 maybe at the latest It was a necessity as no one in the company could really speak the language except the microprocessor group Smith But think as is true of a lot of successful inventions or processes when you re doing it initially you don t have any idea what the implications are going to be long term It was just fortuitous that we had the Busicom deal and the 1201 that this evolved It would be nice to say that we all knew this was something really significant and it was going to have a significant impact but don t think any of us really appreciated what was going to happen in the future
14. go back to what Stan was saying for a minute where he talked about the AMI chip and the size of the chip and the tables and tools When we did the 8008 it was think the last chip that was drawn at five hundred times actual size and the drawings were all done by hand It was also the last chip at Intel that didn t have any computer layout support with it We did it by hand and we actually had to do the drawings in two separate halves House When you say do it you mean the mask set design Feeney Yes doing the mask set design can t remember the exact width it was either four feet or five feet wide of the vellum style paper that we were using So we could only do one half of it at a time We CHM Ref X3706 2007 2006 Computer History Museum Page 21 of 30 Oral History Panel on Intel 8008 Microprocessor had to set it up in such a way that we could then later paste it together on a table that was the size of a ping pong table So everything after that was done at about two hundred times the actual size not five hundred times the actual size House Tell me about the use of computers in the design of the 8008 Mazor l Il make a comment about it and that was that Ted had written a logic simulator for the PDP 8 and we also had subscribed to a time sharing service a company called Applicon in Boston that had a time sharing logic simulator was Hal s officemate and when we were doing the logic design or he was doing the logic
15. our agreement with MIL Microsystems International Limited the company in Canada we had arranged for Ken Au a designer to come down and be my officemate The idea was that they MIL would develop that processor we d develop the ROMs and RAMs Well Ken was my officemate and was transferring the 4005 architectural ideas to him and he was working on the design turned to him one day and said And how are you gonna test it And he said Test it Thats not my problem That s somebody else s House It s another department Mazor Exactly another department And one of the things that Federico specifically did as he was designing the 4004 was putting in some features to make the chip more testable Because if you don t design a chip to be tested then you know possibly it can t be tested and what good is it Faggin Surely it can be tested Mazor Thank you So this reflects in this case of the Canadian company a larger company with more departments with a more narrow focus and you know it s much more problematic We of course had success with the 4004 so we eventually abandoned our efforts and did not join in with MIL in the 4005 CHM Ref X3706 2007 2006 Computer History Museum Page 29 of 30 Oral History Panel on Intel 8008 Microprocessor Feeney Well as part of this too we re talking about how all of us got started and were involved in making it happen but there were so many other peopl
16. released into the marketplace 1103 was a 1 000 bit dynamic random access memory The 1103 used an 18 pin package Faggin It the 8008 originally used a 16 pin package Feeney Right It originally used a 16 pin package and it became very clear and very urgent that another pin was needed As a result when the 18 pin package became quote a standard package within Intel that gave us an opportunity to take a look at potential limitations with the 16 pin design that was being done for the 1201 and with that the issues that Stan brought up in terms of the need for a few new instructions and a few more pins Two more pins gave us the ability to get more information to the outside world Most important the restart instruction was designed just to start the machine but it wasn t originally designed to push all of the information in the machine onto a stack The additional pins gave us the opportunity of doing that So by the time the January timeframe rolled around we had the additional pins we had additional instructions and enhancements and some additional timing things that came in from inputs from both Stan and Ted in the December January timeframe Ted Hoff might point out that one of the reasons for changing the restart instruction was that Stan and had figured out it might be nice to add interrupt to the processor And our first thought would be to make CHM Ref X3706 2007 2006 Computer History Museum Page 5 of 30 Oral History
17. the normal business channels of selling computers And we Intel might have been through the back door into the personal computer business as we know it today House So the development system had a keyboard it had a display it had DRAM it had floppy disk and later got a hard drive It had all of the things that were in a personal computer What stopped it from being the personal computer Mazor We even had customers who would tell us that if they wanted to buy a computer in their company it was a really big deal They had to have permission from their computer Czar but they could buy a Intel development system and they could quote use it as a computer That was an easier thing for them to do regardless of the price Feeney And a lot of customers bought it at that level and when it had an operating system it had WordStar on it and other tools think there was some version of Multiplan that was on it too And so they had tools available to them to use it as a personal computer House Later we made a portable version called the personal development system PDS which looked like a Compaq Later Compaq came out with their luggable and that was sold with a word processor and a spreadsheet on it And so it was about as close as we got In fact Jack Carsten tells the story that Rod Canion came and visited him and he showed him the PDS It wasn t too long after that that Compag was created with a product that was quite simila
18. two customers to a general purpose computer chip for the general marketplace Smith The MCS 4 family was introduced in November of 71 as a standard product As far as was concerned that s what we were selling It was not a custom product and the decision had already been made at that point Feeney But if could interrupt though at that time there was a restriction or a requirement that any customer using that product sign an agreement that they would not use that product in a calculator oriented environment Smith Yeah think that was part of the original agreement with Busicom Feeney think it was standard Smith Yeah correct Gelbach Later we renegotiated that out by the way Smith We also found that the traditional marketing organization wasn t correct for marketing a product like this and Ed set up the Microcomputer Systems Group And don t remember when you actually got there Ed Gelbach think in June 1971 Smith 71 We formed the group in early 72 and Microcomputer Systems which was given responsibility for had marketing which was Hal had software had engineering Phil Tai we had our own manufacturing It was a complete autonomous entity And we initially developed what we called CHM Ref X3706 2007 2006 Computer History Museum Page 12 of 30 Oral History Panel on Intel 8008 Microprocessor simulator cards the SIM 801 and the SIM 401 to make it easier for a designer to des
19. visited Digital Equipment Company DEC And you have to understand that if you have a component which is a central processor and you go to a CPU designer he s rather insulted You know How dare you Who do you think you are And also think Federico will comment that we had a lot of criticism Well why did you do that and why didn t you do it this way because obviously as CPU designers they knew and they cared about CPU design Faggin Yeah My first visit to a customer after the visit to Seiko with Ed was a few months later and we went in Europe And visited Nixdorf and Plessey remember they were incensed about the idea that a chip company could actually have microprocessors Really they did not like it at all Faggin Nixdorf was absolutely anal about it mean they were just you know How did you do that and You cannot do that That doesn t work And That s bad That s bad The contrast with other customers where you had customers that wanted to use it was incredible The wanted to make a blood analyzer in Denmark forgot the name of the company and they were elated because they finally got something that they could use to solve their problem And so you could see the difference in mindset But at the same time it was useful to get that feedback because in some sense they Nixdorf had some good comments Some of the things that they wanted to see in a computer that were not the
20. were ridiculous You just had to do it on the faith of the people that were involved And you re absolutely right the first applications that they talked about were doing your menu planning in the kitchen and these bizarre ridiculous things And actually Apple didn t really take off until there was a real application which was VisiCalc That really is what got that product going Mazor The so called killer applications were certainly probably text preparation and printing and then also VisiCalc Faggin VisiCalc was because there were word processors before Apple You did not have any way to do spreadsheets except with a large computer and even those were difficult to use Hoff We found a number of people in our marketing departments buying Apples and so on just for things like VisiCalc so that they could to their sales projections that way House Ed understand that when we first started selling microprocessors it was really coupled with EPROM Tell us that story CHM Ref X3706 2007 2006 Computer History Museum Page 24 of 30 Oral History Panel on Intel 8008 Microprocessor Gelbach One of the early products that Intel came out with was a programmable or erasable programmable ROM The intent was always that the customer would use it as a design aid Use a couple of em and then order in large quantities a fixed mask ROM As it turned out the microprocessor was the perfect tool to fit with an EPROM because customers alwa
21. We were not supposed to use computers because they were expensive in those days And so had my slide rule and graphs that had taken from actual transistors normalized and developed a methodology to do that Fortunately had invented the bootstrap load which nobody thought was possible That was actually what made possible the speed power required to make the CHM Ref X3706 2007 2006 Computer History Museum Page 8 of 30 Oral History Panel on Intel 8008 Microprocessor microprocessor work In those days without that you could not do it You would have had either too many gates or too much power And then of course there was the other thing that was necessary Another invention of mine was the buried contact which was the ability to connect directly the silicon to the junction so that you would have like two layers of interconnections And so got busy working and when the first run of the 4004 came out it was in January of 1971 it pretty much worked At that point it was Clear that we could do it so Vadasz asked me to supervise the 8008 project Hal Feeney of course did the detail design using the 4004 as a model because the layout of the 4004 was very similar to what was required for the 8008 Many of the circuits that were used in the 4004 were used in the 8008 so that became a much much more manageable project It was not a team project by any stretch of the imagination but it was much easier because there was a model that worked
22. algate technology in the early generations So they were way too expensive They were not as cost effective as silicon gate The 6800 it was the first device of Motorola that was silicon gate Also the F8 that was done by Fairchild was silicon gate So only with that generation about 1973 74 where the competition had a competitive technology to what we had Then you started seeing competing devices Before that they couldn t compete in speed never mind in size or cost Feeney Well the other side of that is because the EPROM erasable programmable read only memory came out within maybe a month or two before the 8008 came onto the marketplace It was initially very difficult to use It was think a program once only type of device Sometime after that they came out with the quartz lid so it could be erased by UV ultra violet light Even then it was very difficult to program So how do you make it easy for the EPROM to be programmed You use a microprocessor to do it With the ability to use a microprocessor the programming was done Also the programming algorithm could be changed as needed as we learned more about how to do it It was also less expensive So then customers started buying the microprocessors to do that There were other third party stand alone program manufacturers for EPROMs and those third party groups were selling in CHM Ref X3706 2007 2006 Computer History Museum Page 22 of 30 Oral History Panel on Intel 8008 M
23. and special serial memory device The 8008 was planned in such a way that it could be used with any standard type of memory because as Stan pointed out we had this exorbitant amount of memory 16K available that could be any mix of standard ROM for program standard RAM for storage and that could be partitioned in any way between all RAM or a mix of RAM and a mix of ROM And using the standard products that Intel had or perceived at the time and then as it turned out and we ll be talking about this think a little bit later when the EPROM became available that was really the item that provided the greatest jumpstart for the microprocessor But back to the design side of things we then went with our design starting or restarting in the January timeframe of 1971 and carrying it through to having our first devices in the late fall of 1971 So with that we implemented a relatively small chip and if recall it was something like a hundred and twenty four by a hundred and seventy one mils on a side It had to fit in the 18 pin packages that we had which was quite a constraint In contrast Ed mentioned that TI had a design and TI s design was basically something about two hundred and twenty mils on a side and they had the luxury in quotes of going ina 24 pin package And TI did announce its chip and had a photograph of its chip available in Electronics Magazine in approximately July of 1971 But from any information that we were able to obtain i
24. and the customers want to get on with their business of solving their problems and not going back and reengineering things And that s something that we didn t understand at the time House It wasn t socket compatible Feeney Correct Mazor ld like to reflect that some of the best years enjoyed were at Intel working with these fellows We were a small team and we had a lot of the correct ingredients Engineering was able to meet with the customer and hear their requirements interpret them and move pretty quickly into engineering products And working together for example myself in applications engineering with a chip designer in the same office This wasn t going through different departments and through e mails and so on but you know we had a lot of close synergy And so think part of our success was that we were young and small and quite able to move quickly as needed Hoff The other thing too is that all of us here were not specialists in one narrow area but had a broad range of interests And the idea that Federico had his prior experience with computers and calculators made the transfer over just so great You know with a different team it might have been much more difficult to accomplish that Mazor In particular Ted and early were worried about the 4004 being too aggressive and specified another chip which has never been heard of called the 4005 This was a very much simpler CPU than the 4004 And with
25. andard or full blown product The argument was that Intel was a memory company and if we go into the computer business our customers the ones using our memory chips to build mainframe computers are going see us as competition and they ll be looking to get their memory chips from somebody else So it could hurt the memory business That was one negative The other was that it was hard enough to train people in marketing to go out and deal with customers for the memory chips Computers were seen as an order of magnitude more difficult and the training and the support would be terrible And remember had a meeting with believe it was Bob and Gordon but it was Bob Noyce was talking to primarily sometime believe in the early summer of 1971 And Bob said to me something like Well we re not ready to make a decision to announce this product We have you know a lot of other And said basically Every time you delay announcing it what you re really doing is making a decision not to announce And said Eventually somebody else is gonna open up this business and we ll be left behind even though we re the first ones to have products So it really wasn t until Ed joined Intel that found a positive attitude in marketing toward offering these as a product Up until that time there was a lot of negative feeling House Hank you were there before Ed joined so tell us about your involvement with this product You were obviou
26. andicapped it was probably the number of pins in the package and had we been more aggressive in that regard we would have had a perhaps more successful and better product Interestingly enough TI when they did their chip design they used some automated tools and not all of the area of the die was active so the yield is not therefore so badly impacted if you get a defect in an inactive portion of the chip So perhaps in our general methodology we might have achieved faster turnaround times in subsequent designs if we hadn t had the mentality of a memory company But that just raises some questions Obviously we were successful in what we did CHM Ref X3706 2007 2006 Computer History Museum Page 25 of 30 Oral History Panel on Intel 8008 Microprocessor Feeney But Stan let me go back to a fact of life at that time The fact of life at that time is that we did not have the pizza sized wafers that we re dealing with today We were dealing with two inch wafers and we had to get as many devices on a two inch wafer as we possibly could And as a result to have a number of good die candidates and I m not sure how many were on there but we re talking about something in the range of one hundred two hundred devices on a wafer So from a production point of view getting the small chip was really critical mean we used every allowable square millimeter that we possibly could get inside of the package In fact we were right on the ed
27. child then went on to build a calculator chip which was bit serial and focused on CHM Ref X3706 2007 2006 Computer History Museum Page 20 of 30 Oral History Panel on Intel 8008 Microprocessor arithmetic and didn t focus on the peripheral control problem at all So think Fairchild did just what we at Intel thought was the wrong thing to do Interestingly enough think it was RCA that built a version of the PDP 8 and neglected the fact that the PDP 8 subroutine return mechanism wouldn t work in read only memory ROM And one of the big applications was control applications with read only memory in which case you have to have something like the stack for subroutines And remember people saying you know how ridiculous it was of us Intel to invent a new architecture Why didn t we you know go with an existing architecture not realizing that those architectures were not designed around program stored in read only memories and might fail for these reasons And another one that like very much is AMI which was a very successful custom house in Silicon Valley They thought the 8008 was a terrible design and they could make one CPU that was much better When they completed the masks for it there was no table large enough in the company to hold the artwork for it and so the chip was so big that they were never able to make that chip called AMI 7200 So you know competent people could say This is wha
28. d had to create the entire methodology and design with silicon gate technology for random logic There was nothing there Intel had never designed random logic chips They only had designed memories up until that point Memories are much simpler from a design point of view from a layout point of view and from a testing point of view and so on and there was not even an understanding of what it took to design random logic chips For example in the schedule that Vadasz had given Busicom the layout of the 4004 was supposed to take four weeks exactly just like the ROM or the RAM And in fact when got to Intel the customer expected to have samples of the entire set by September that year joined in April and found that the company was supposed to start the project in October of that prior year but nothing got started and so found that was six months late before even started it And the nine months that were required to do those four chips were really a tall order to start with So it was a very difficult time and had to do a lot of things with a very short time and essentially without any help was by myself didn t even have a draftsman and had to basically invent my way out Which is what did And so developed the entire methodology how to do it the circuit blocks the way you organized the entire layout and the way you do the simulation using graphics design so that you would quickly be able to size the transistors
29. d My frustration was lock in pin count packages When came up with the idea of the 8080 it took me nine months to get Les Vadasz to give me permission to do it sent him a memo back in early 72 to do the 8080 It wasn t until the end of 72 that could start And the fight that had to put up to get a 40 pin package you cannot believe it because 40 pin package was not acceptable Fortunately we had a custom project It was a single chip calculator that had to be in 40 pin So said Now Vadasz you tell me Why can we do a calculator on a 40 pin package and not a microprocessor on a 40 pin package nailed him with that one House He had to give up then Faggin Yeah he had to give up But it was always a fight For meat Intel it was always a fight to stay on the crest We had started the business and we were pulling back instead of going all the way For example to get programmable peripheral devices okay They were basically memories with a different name very small simple chips To have a coherent family of chips there was no interest It was too much of an investment and so they didn t want to go down that road That is why ended up starting my CHM Ref X3706 2007 2006 Computer History Museum Page 26 of 30 Oral History Panel on Intel 8008 Microprocessor company because of that frustration And the Z80 was conceived from the very beginning as a family as a family of components that would be seamless
30. design part of my job was to do logic simulation for Hal and verify that the logic simulated okay as a way of verifying it the design And other than that we did not have a lot of resources for computers What do you remember about it Hal Feeney Well we used a time sharing service for some resources At best we could do one or two or three transistors to calculate propagation delays and calculate the size of the gates size of some of the resistors and whatever and to generate speed for the devices We would go through and model small circuits as part of the design and then try to replicate or reuse those small circuits as much as possible throughout the design think ran up something like a fifteen thousand dollar bill at one point for time sharing services and was told that we shouldn t be spending that much money on design tools What a contrast now to the design teams and the millions and millions of dollars that are put into it It s just a totally different viewpoint We were bootstrapping our self into that business and from a management perspective had no concept of the business we were getting into Mazor think if you ask Federico what tools he had think he s had three colored pencils Faggin Yeah But to go back to the question of why was Intel all alone There was another aspect that s important to note and that is that Intel had silicon gate technology All the other micros that were done were all using met
31. e Ed did you know that this was happening Ed Gelbach Yeah knew it from a different aspect though at that point was think the Vice President of sales for TI Texas Instruments TI at that point was also involved And think he had a brother working there Gus was the brother s name and remember going in and negotiating it House Vic Poor had a brother Gelbach Yes He had a brother House Okay Gelbach And somehow got involved in negotiating what think the was the same product And was a little like Andy didn t know what it was but thirty dollars times a hundred thousand sounded like a pretty good product And you know we immediately said or TI immediately said they could build it and in fact did build it don t remember how it turned out but do remember they delivered working parts So was aware of that when left there House And you left Tl when CHM Ref X3706 2007 2006 Computer History Museum Page 3 of 30 Oral History Panel on Intel 8008 Microprocessor Gelbach 1971 mid 71 Mazor have a little bit about the part that we knew about and that is that our specification was written for Datapoint we didn t ask them to sign a nondisclosure agreement because it was our customer It s very likely that our datasheet ended up at Texas Instrument We believe their TI s original proposal was to do it in several chips three chips in fact in a way Intel set down the
32. e involved in this product all the way along through the whole production side of Intel the mask designers the people in production engineering and whatever to make it happen was going through some notes this morning and ran across the names of Rodney Sayer ph and Julie Hendrix the designers on the product and Bob Holmstrom ph who was the product engineer that actually got the thing through and go it into production So there s a whole handful of people And Hank pulled out a photograph earlier today of the early microprocessor group with about two dozen people in it all the people that it took really to get this thing into the market and make it really successful House Does anybody have anything else to add Well think we can wrap it up then Thank you everybody END OF INTERVIEW CHM Ref X3706 2007 2006 Computer History Museum Page 30 of 30
33. e laid out Faggin Exactly The decision to restart the 8008 was based on Seiko s interest As you remember Ed you and I visited Japan Gelbach That was the incident with the keys Faggin With the keys Yeah CHM Ref X3706 2007 2006 Computer History Museum Page 15 of 30 Oral History Panel on Intel 8008 Microprocessor Gelbach We re both leaving together to go to Japan Federico was waving out at the airplane and said What are you waving for And he says I have my car keys in my pocket and my wife is probably trying to figure out where am right now Faggin But anyway House You were going to see Seiko and they had an interest in the 8008 Faggin They really wanted the 8008 at that time They were very serious They already had pretty much decided to use it and the reason why the 8008 was restarted in January as opposed to being simply abandoned because Datapoint was not interested in it anymore was that wanted an 8 bit micro to do a scientific calculator That was what got that project restarted Feeney In a very early advertisement brochure that we put together we had a number of different applications for Intel s microprocessors One of the applications was the device that Seiko had It was very similar to the large calculators from Wang and also from Hewlett Packard in the 1970 71 timeframe House So what were the other applications that we had envisioned at the time Feeney Well we
34. gauntlet that it could be done on a single chip They TI were in somewhat disbelief but they sort of picked up the gauntlet and decided to run with it So think TI did build a chip to our specification It turns out my specification had a defect in it and it turns out they built that defect right into the chip and so the chip could never have actually worked And if you think about it with all the effort that TI put into that chip in announcing it ahead of Intel if it had worked they probably would have made a product with it To my knowledge they never did sell it Gelbach don t think they did either Federico Faggin Well Vic Poor told me that in fact they tried it and it never worked House So then Hal came on the scene understand you were hired to develop this chip Tell us your story Hal Hal Feeney got involved in what was called the 1201 at that time It was Intel s first 8 bit microprocessor and with the1201 internal name joined Intel on March 9 of 1970 remember the date for a variety of reasons but on March 9 1970 also have the handwritten spec that Stan put together that took the instruction set that Datapoint had defined translated that into a concept that we could use for developing the microprocessor So from that concept spec and from the concept timing diagrams that were there we looked at the pins that would be needed to communicate with the outside world and went to defining a f
35. ge of filling the cavity of the package with as much as we could That was the defining size for us But you re right If we could have gone to 24 pin packages or something larger it would have made a difference but it may also have made a big difference in the overall cost or the productability House The cost was never a limit though on this product There were not enough of them made where the yield was a significant impact This was the launch pad that got you to the 8080 Feeney Right We talked before about numbers and I m not sure what Ed recollects but 8008 certainly was produced well into or at least available to customers well into the late 70s early 80s And we re talking probably about hundreds of thousands of devices being produced not millions Gelbach We figured out how to price it cause we thought it would be eventually a replacement for the IBM 360 think we started the first pricing at three hundred and sixty dollars Feeney No it was the 8080 that went to three hundred and sixty dollars The 8008 started at a hundred and twenty and we had a speed enhanced version of it that was at one hundred and eighty And so for the 8080 since it was better it had to be higher in price and that s where the three sixty came from Faggin My frustration back in those days was basically the fact that Intel was a memory company Microprocessors were not really seen as something by itself that needed to be really pushe
36. gin using Intel devices in volume they were using them in their peripheral devices where they wanted compute power and they couldn t afford or it wasn t practical to get the compute power from the big iron computers that were available at the time Mazor think it s worth saying here that the real application of the microprocessor was in control applications inside instruments or systems of various types but not as a general purpose computer We had three or four companies that wanted to build something that you could more or less call a personal computer We tended in those days to call it that application a hobby computer But the limiting factor one has to recall is that printers were very expensive that magnetic storage was very expensive and so someone might build a personal computer but it really wouldn t have any capabilities without these expensive peripherals So what held back the personal computer as we know it today was the development of these low cost peripherals And we shouldn t underestimate the requirement need for those peripherals which we now use today and that are very inexpensive Hoff When we looked at the market we saw what is now called embedded control It seems like the media is not even aware of the existence of embedded control It s always the desktop or the notebook computer that gets the attention but it s still a pretty significant market but it does not get nearly the pub
37. icroprocessor competition with Intel but our devices were less expensive So you ve got a convergence of technologies that synergistically supported each other And then just the growing awareness that sales pushed from the development system perspective made it happen House Let me turn to Stan Mazor Well I m an engineer not a marketing guy but did reflect on Jim Lally who was leading the Development System Group for quite a while and quite successfully and he was quite good at analyzing the market He believed that the development system would be sold to engineers and that market was relatively small In that in that kind of marketplace which is not elastic with respect to price then the strategy is to raise the selling price by adding features which we did over time e g by adding in circuit emulators and other features I ve conjectured that had he made the opposite decision conjectured that this was an elastic market he would have found ways of reducing the selling price And then if that was the case the elastic market would have evolved much larger and Intel might have found itself in the personal computer business We certainly know that we were selling the MDS product not as a general purpose computer but as a microcomputer development system MDS will further add that even Digital Equipment Company didn t sell computers they sold Programmed Data Processors PDP because they also didn t want to compete in
38. ign in the product It then occurred to us also that we really needed to have a high level language to develop these products and brought in Gary Kildall House How did you find Gary Smith You know really don t remember what the sequence of events was Do you remember Gelbach dealt with Gary either with you or after you because we were gonna buy the operating system from him forget what happened to it Smith Well he developed PL M at Intel He had an office there where he developed PL M House Did he develop ISIS as well Feeney No PL M was the one that he developed House Where did ISIS come from Mazor We were using PDP 10s in the company Intel and it had quite a nice operating system so we developed internally the 8008 operating system which had the look and feel quite a bit like the PDP 10 I m talking about the commands the user would see like DIRectory DELete COPY file and so on Smith But the interesting part is is that PL M eventually evolved into CP M and Gary started Digital Research There was a real battle at that point as to whether CP M or MS DOS was going to be the operating system for the PC House But PL M was really a language and CP M was an operating system correct Smith PL M was based on PL I and then CP M was based on PL M And if things had worked one way or the other Gary could have had the operating system for IBM s personal computer House So
39. ing certain memory devices that fit perfectly with the 8008 processor And it gave CHM Ref X3706 2007 2006 Computer History Museum Page 10 of 30 Oral History Panel on Intel 8008 Microprocessor us the opportunity of selling memory devices that weren t quite as fast and EPROMs that weren t quite as fast House So yeah the 2102 and the Feeney 8102 House was the static RAM Actually the 8004 speed selection probably charged more Mazor When we were working on the 4004 went to Gordon Moore and said We re going to need an assembler for it Well of course an assembler is someone that you use on the manufacturing line to build stuff So ended up writing an assembler for the 4004 on the PDP 10 computer and I wrote it using the PDP 10 s macro assembler also But then for the 8008 software we ultimately went out to a contractor Sandy Goldstein of Gray Computer Co and wrote a contract to have an assembler program written in FORTRAN for the MCS 8 So we started in the early days to do a little bit of software for the microprocessor s support House Ted tell us how management and the salespeople looked at the microprocessor in those days Hoff When the product first became available it was still considered custom for these two customers Busicom and Datapoint although we had one or two others for the 8008 But there was a very negative feeling within the company toward making it a st
40. licity that the other machines get Gelbach Orders of magnitude higher volume in terms of number of units Feeney The earlier concept with selling the microprocessor in traveling with Bob Noyce Bob always tried to describe the microprocessor the same way you think of a fractional horsepower motor in your house You ve got your furnace motor you ve got your washing machine motor your dishwasher your garbage disposal you ve got motors in all of your clocks and whatever And at that point it seemed like Bob was certainly an optimist in terms of the way the microprocessor would become pervasive But on the other hand if you look at it in retrospect he was probably a pessimist because when you look at the BMW with fifty or sixty microprocessors in a high end version it s just mind boggling don t think any of us could have seen how far this would go CHM Ref X3706 2007 2006 Computer History Museum Page 19 of 30 Oral History Panel on Intel 8008 Microprocessor Hoff When the three of us were recognized for the 4004 at the Inventors Hall of Fame they told the audience You might have ridden here in an automobile that was designed with the aid of a notebook computer They totally missed the point Almost everybody probably drove there in an automobile that was running under computer control Mazor need to mention that Ted had some very innovative ideas when we were playing with the microprocessors in the early days and
41. ly fixed that and that was it The 8008 then was put into production so after that it was marketing s baby Feeney And with that just for the sake of this history the 4000 series was announced in November of 1971 and then the 1201 then was dubbed the 8008 to be consistent with and to show that it was greater and bigger etcetera than the 4004 That was announced in April of 1972 House Federico tell us about the naming These names didn t fit Intel s normal naming algorithms did they CHM Ref X3706 2007 2006 Computer History Museum Page 9 of 30 Oral History Panel on Intel 8008 Microprocessor Faggin Under Intel normal algorithm which consisted of the following the first digit referred to the technology so one was P channel two was N channel three was bipolar And then the second digit referred to the type of device one was RAM two was shift register believe Feeney No it was random logic Faggin Right right Two was random logic Three was ROM and so on Feeney Four was shift register Faggin Yeah four and five were shift registers So following that logic the Busicom set would have been probably called 1202 1203 1204 1205 which didn t like So said We should call it a 4000 family 4001 2 3 4 but Les Vadasz didn t want to hear about it He just didn t want to hear about it was breaking the system But I was insistent liked 4001 4002 4003 4004 By the way found out
42. made more money from some of the literature because we charged for the literature House So you sold that data manual and the program manual Mazor Datapoint did announce their computer They called it the 2200 and they ended up building it entirely of TTL and they ended up building a parallel version of it So by the time we had our 8008 chip ready they said our chip was slower than the computer that they were already producing And I was in constantly talking communication with one of their Datapoint architects Harry Pyle and they started to enhance their architecture So over time the Datapoint 2200 evolved and the instruction set got a little bit away from what we had done on the 8008 And also we didn t worry too much about software at the beginning of the 8008 project because it was for our customer and he had the system and we assumed that they d be writing all the programs But later on as we got into the project we had to start thinking about how are we gonna program this and what are we gonna do with it Feeney Right And that became very significant from two points of view One is looking at the customer but the other side of it is internally We had to test this thing We were testing a computer And so basically what we had to do was set up a system whereby we used a computer to test a computer And we didn t have a lot of tools We certainly didn t have any history on this because the 4004 and the 8008
43. mas time in December 69 and asked him what he needed the stack for and he said he was building an 8 bit computer asked him a little bit about it and had the advantage that I d been working with Ted Hoff on the 4004 since October for several months so the idea of a central processor on a chip was something that we were working on already And asked him about his computer and how smart it was and what it had to do wrote out in front of him three proposals One was an 8 bit register set with a stack and another was a register stack with an arithmetic unit And then went on to the third proposal and said Well it s possible we could do the entire 8 bit CPU on one chip Well he certainly wasn t about to believe that but he was interested said But we need to know more about your computer before we could proceed with that So in the January February timeframe he sent us a programming manual sort of an assembler language manual disclosing the instruction set of his computer Ted and had a look at that and one of the interesting things is that it had about the same amount of register memory as we had in the 4004 an 8 deep stack which is about the same amount of memory that we had in the 4004 At first blush you might think Oh an 8 bit computer s gonna require twice as much CHM Ref X3706 2007 2006 Computer History Museum Page 2 of 30 Oral History Panel on Intel 8008 Microprocessor logic as a 4 bit
44. n the chip specification but about a six month hiatus in terms of the design itself while we went and did some other work at Intel also put in some of my time working with Federico on the 4004 and the 4000 series products CHM Ref X3706 2007 2006 Computer History Museum Page 4 of 30 Oral History Panel on Intel 8008 Microprocessor Mazor Hal if could interject just a couple of things on the 8008 specification Ted also worked on the specification we worked on it together And just a couple of interesting things about that specification The Datapoint instruction set did have a branch on bit instruction which we decided was not needed so we removed it Basically we were building a computer for our customer so we didn t challenge their instruction set or test it out in any way But Ted realized that they were sorely missing an increment and decrement instruction so that was added based on Ted s idea So that was one of the enhancements we made Another minor point is that the jump address in the 8008 Jump instruction was 16 bits but because we were so short of pins we only used 14 bits of address and we thought that sixteen thousand bytes of main memory was a huge amount of memory Who would ever need more So although the instruction had room for it we only used 14 of the 16 address bits And lastly the original design for Datapoint what they wanted was a bit serial machine And if you think
45. other chips around it but could use general purpose memories and could run out of RAM in particular and so it was the more general purpose solution So here we were in the early days with two complementary product lines that were in no way compatible in terms of software or hardware features otherwise We were in a very interesting starting point with these two lines House Which one sold better Faggin think the 4004 sold better Smith think the actual volume the 4004 did Feeney It got into cash register type applications NCR was a large customer for a very long period of time The 8008 got into some fairly high running terminal applications But those certainly didn t run at the same rate or same volume as either a scale or a cash register application But want to go back to the comment that was made about having special things and needing special things and Intel s view This is a very early 8008 user s manual of fifty sixty pages on how to use it how to get started with an 8 bit CHM Ref X3706 2007 2006 Computer History Museum Page 27 of 30 Oral History Panel on Intel 8008 Microprocessor microprocessor We worked to meet the launch date and to get the user s manual out so we had all of the things that the customers would need to answer their questions At launch we didn t have a four page datasheet available in Intel blue And there was criticism because we didn t have the standard datasheet that everybod
46. ouse But you went anyway CHM Ref X3706 2007 2006 Computer History Museum Page 17 of 30 Oral History Panel on Intel 8008 Microprocessor Smith We did We modified some of our claims and we took off some of the more obvious references to computers but we went anyway and we introduced it at the show But have never seen Gordon either before or after as angry as he was then House And what was he angry about Smith Well it was the fact that we were going to be competing against our primary customers as far as Gordon was concerned We were not going to introduce a computer This is not a computer Hoff had an impression after we announced the products and ended up going out with marketing people to visit a number of customers particularly large computer companies they had been developing some concepts for controllers and their peripheral devices And when we came out with the microprocessor they were able to use the microprocessor and it ended up that they didn t have to do all of this elaborate development that was going to cost them a lot We were offering almost everything they needed So it seemed to me that we ended up complementing the big mainframe guys in some of these peripheral needs Smith The microprocessor became the peripheral controller And of course history bears that out that we never really were competing with those customers Mazor However we did have a chance to visit customers for example
47. parate product House Did they have any similarity to the SIM 8 and SIM 4 Did they use the same electronics inside Smith There was a lot of commonality Yeah absolutely Gelbach Actually it was the same boards and we just put a box around it Feeney Well no we made two different moves there We had the SIM 801 in its own box and then later on we translated it to believe boards that were similar in format to what the memory systems division had so that we could utilize some of their memories Those devices were critically important mainly because they were used both as PROM programmers or sometimes standalone The PROM was really the unsung hero for the development and the growth of the microprocessor business Customers were able to develop their programs ship their initial systems be able to use those systems and sometimes chose not even to move over to ROM but just continue to use the PROMs as their shipping device as they changed the software Smith We were really the first company to have the simulator boards to have the high level language to have the Intellec type systems None of this stuff existed and so we pioneered those products and actually had a pretty good business selling those products And it was critical think to the success of the microcomputer in being able to provide that kind of support to the engineers out there who were trying to design these products in Gelbach We also started the application
48. r At least Jack Carsten believes that was the genesis of the Compaq luggable CHM Ref X3706 2007 2006 Computer History Museum Page 23 of 30 Oral History Panel on Intel 8008 Microprocessor Mazor And that s not to say we would have succeeded in the business as we know it now but it still raises a question House understand management was afraid to go into the computer business again from fear of competing with the customers Mazor Yes Faggin But also believe that the potential with the personal computer was not seen at all definitely at Intel remember Gordon being quoted as saying Well what do you do with a personal computer Basically it was not understood it came out of a different milieu It came out of young people that were enthusiastic They saw the potential although it was not clearly articulated In fact the first applications of personal computers were really silly applications that eventually developed into the mainstream technology So think that there are other aspects to it other than just the lack of vision of Intel management think that it was House An idea before its time Faggin Yeah before its time and by different people Smith was at Venrock when Apple got started As a matter of fact we were investors in Apple We had no way to evaluate the initial business plan it because there was nothing even remotely close to it that had been done before and the projections
49. re became ideas for the next one CHM Ref X3706 2007 2006 Computer History Museum Page 18 of 30 Oral History Panel on Intel 8008 Microprocessor Smith A lot of these customers found that they could develop a family of products fairly easily by just changing the instruction set and they didn t have to redesign it Feeney The other part of it was that the customers at one point found that the only alternative that they had was going and putting a minicomputer in this blood analyzer or in this scale or in this point of sale device And it s probably in some respects lucky for us in the microprocessor industry that we weren t competing directly with the minicomputer manufacturers at that time because it allowed us to get in in somewhat of a stealth way The initial customers were all not on the tip of Intel s tongue as far as major customers go In fact recall a discussion had with Ed one day he was looking at our customer list and he was saying Hal where are all the big name customers Were dealing with UNIVAC we re dealing with Burroughs and we re dealing with IBM in the memory side of the business customers who rank in who s who look at your customers and it s always who s that But it gave us a baseline to sell lots of microprocessor development systems and then gave us the way of developing the base for all of these different computer applications On top of that when the CDCs the UNIVACs the IBMs did be
50. sly involved with both products at the same time Smith joined Intel in early 71 My initial responsibility was to market and support the memory chips the 1103 the 2102 the 3101 whatever the right nomenclature was for bipolar That was my initial CHM Ref X3706 2007 2006 Computer History Museum Page 11 of 30 Oral History Panel on Intel 8008 Microprocessor responsibility When the microcomputer was developed and we were talking about marketing the microcomputer got very much involved in developing all of the support documentation that was required naming the family putting together the packages that were gonna be required Very early on we understood that this was very different than selling a memory device Microcomputers were gonna require a lot of service a lot of support Engineers typically were not used to programming This was an entirely different way of designing In following on what Ted said being first was extremely important because you developed a very loyal customer Once they put their effort into developing the software and the instruction sets that were required it was very difficult then for them to move and try somebody else s device They were locked in So our initial marketing focus was to get this thing designed into as many applications as we possibly could This was when we were selling the MCS 4 family and the MCS 8 family House So how did the transition get made from this is strictly custom product for
51. t was never fully functional but it was a very good looking chip and it was a nice wall size to use And might mention and we can go into it in a few minutes but from the history point of view we had our photographs of the 8008 in the October November timeframe as we were setting out to debug it and get CHM Ref X3706 2007 2006 Computer History Museum Page 6 of 30 Oral History Panel on Intel 8008 Microprocessor the chip ready for production But with that we had a lot of experience at Intel of developing new chips and putting a two or four page datasheet around every chip that came out and equipping our salesmen with the whole stack of datasheets to hand out to the prospective users be it designers or buyers In the case of the 1201 the major challenge that we had was that internally our one page preliminary spec that we were restricted to having for every single product done within Intel had to be on a single sheet of paper and every place in this spec it effectively said See instruction manual See instruction manual See instruction manual And the real spec that went along with it was about thirty pages at the time with genesis from Stan s original document and it turned into a full blown book Intel got into the publishing business with the announcement of the product Ed probably never knew or never expected as he came to Intel that we d be selling books Sometimes think more books went out than chips Hoff In fact we
52. t you really need to do but if you weren t concerned about and didn t realize the limitations of the package and of the power and of the silicon if you didn t have the expertise like guys like Federico and Hal then they would do the wrong thing and they wouldn t succeed in the market or they couldn t build it at all Faggin On the other hand they were successful from the point of view of working microprocessors for example the PPS 4 of Rockwell It was announced probably a year after the 4004 maybe even less than that House And that was successful Faggin That was successful meaning it worked and it had some applications But what Rockwell did not do they did not do nearly as good a marketing job as Intel did They basically were selling it them as chips There were more like custom things so they would support you but they would not have tools they would not have the paraphernalia that we had developed at Intel And as such they had very few applications and that was it Feeney Many of the other companies also didn t have the potpourri of products to go around the microprocessor When we were first marketing the processors we d look at the ratio of memory to the processors and very often the value of the processor set going out was weighted toward other chips in the Intel portfolio and not to the microprocessor itself That was important as far as being able to take in funds for developments like this want to
53. this with it Can we use it in this mode What about these instructions that you have in it or what about some instructions that aren t being used What happens with the processor if we apply those instructions to the processor Can we use it this way or change things with memory or whatever else it may be There were a myriad of different customers looking at it in different ways Going back to Federico s comments about dealing with the customers in Europe and some of them saying You know it s got difficulties here and problems here and so forth you d go and talk to the same kind of customers in Japan and you would find that the Japanese would even be a little bit more critical and say There s a problem here there s a problem here but we can fix this problem we can fix that problem and they would race off into production with a product and not have any concern at all So that was more of an attitudinal thing Of racing ahead of the market getting products into the market using new innovative technology These customers the visionary ones were the ones that were driving all of this and that ultimately drove the success that led to the devices that out there today Mazor think upon reflection we Intel were a memory company and we were really paying a lot of attention to two things one was the small die size and the other was the number of pins If we look back on the 8008 something that h
54. tually charged for No one had ever charged for them before and we just came up with the idea Hey let s charge twenty five bucks apiece and at least it ll keep the riff raff out So it turned out to be and I m gonna say this a little simpler on the sell than it sounded because it was such a new product But everybody except computer engineers wanted to know what it was We used to push it that if you don t understand it you re gonna be left out It was really clear that this is the next generation So was it really lucky or was it skillful Faggin Or both Feeney Boy part of it was a matter of creating our own luck or creating the opportunity for it think Intel did it by having an environment where it could happen Certainly not intentionally getting into this business but by creating the environment where this could happen and ultimately change the direction of their business House If you look at microprocessor history there were a number of companies that came out with processors in the early 70s and mid 70s Few of them with the exception of one we ll talk about in a subsequent interview Zilog s Z80 ever actually became successful products mean RCA Intersil You ve all seen the microprocessor history poster There s a whole bunch of products that came out but they had no successors Mazor Well can speak about a few of them competitors because we Mazor Faggin had been at Fairchild Fair
55. unctional block diagram for the chip itself We also went into defining what the logic would be to accomplish and set up and decode all of the instructions that were part of the instruction set So as we went through that process it took about two months for us to develop the initial preliminary specification and develop the block diagram And going back to some comments that were made about TI it was at the same time that TI based on other knowledge that we ve had TI was designing a 3 chip set and were designing a serial processor Intel s was a single chip and was a parallel processor And to that point CTC Computer Terminal Corporation at the time before their name was changed to Datapoint was working with both companies and sharing information back and forth so that they could get the best product that they possibly could And as it turned out the Intel product was moving along up through the middle of approximately 1970 where we were finishing all of the functional aspects of the specification getting the specification approved both internally and then sharing it with CTC And that carried on up through about July of 1970 and at that point in time there were some problems wasn t really associated with it from the sales side but there were some problems with whether the customer really was going to follow on and use the chip or was not going to use the chip As a result there was about a six month hiatus A little bit of work was done o
56. ut and everybody understood it Smith Talk about anything yeah Feeney It was hard to talk about industrial control and some of those things but the things that would touch people s lives were helpful Feeney Toledo Scale was a big early customer Smith Computers in a teacup was something that one of Regis guys came up with The brochure was tiny also It was done just about the time of the National Computer Conference This was in 73 as well House Tell us the story about going to the National Computer Conference for the first time and the reaction you got from management This was a new era for Intel Smith Yeah Well Ed and both were in the room with Gordon Moore and Bob Noyce We were outlining what we were doing and what our plans were Gelbach And we had all the displays on the table remember And there were the standard blue boxes development system but there were also brochures and posters Smith Oh yeah Gelbach And we were really excited We were going after this business and we had worked on it six months maybe a year whatever it was And it was the future of Intel And we had no doubts this was it And as Hank s explaining it Gordon s face is getting longer and longer Smith And redder and redder Gelbach And at the end of it or somewhere must have stopped and said What s wrong And Gordon just made it very clear we weren t gonna do that you know and so Okay H
57. were the first two microprocessors done by Intel and we basically had to set up a system in such a way that we would test the individual devices and present the full range of programming to them and exercise every element within the computer using our own homebuilt test devices But as an example guess we can show the result of this effort House Turn it the die photograph towards the camera Feeney When each of these chips was built we took a photo micrograph of the die and then we were able to go through the design and set up and partition a block diagram on here In general we had a bus going around the chip we accepted all of the instructions and decoded them We had this whole mass of random logic on one side to go through and set up all of the operations that had to take place from the arithmetic unit and from any other random logic work that had to be done On the other side we had a CHM Ref X3706 2007 2006 Computer History Museum Page 7 of 30 Oral History Panel on Intel 8008 Microprocessor very regular section of memory You look at this in contrast to the chips that are developed today and you see how much more regular they are in their design with massive massive amounts of memory But this was the very first attempt and first embodiment of doing this kind of thing and trying to do it in what turned out to be a relatively small chip then Compared to today s chips it is a small chip House So Federico had been busily
58. would work together well It would make sense So need to register because it was that frustrating for me Smith Well think we got that Hoff think one of the problems was that so many people involved in Intel were out of the traditional semiconductor business and that bias carried through There was a few times where a new package caused problems At one point they were gonna save a few cents on lead frames and they went to a punched lead frame instead of an etched one and it caused reliability problems So there was a real negative view toward anything different in the way of processing or packaging even though the business of the microprocessor might have easily supported the research and everything necessary to make it work Mazor On the positive side with the EPROM we went to a quartz lid which was quite a revolutionary thing and we had ceramic packages which were more expensive as well albeit they carried a high average selling price Faggin Yeah but it was 24 pins Mazor ld like to say one other thing and that is in November 71 we announced the 4004 and in April 72 just four or five months later we introduced the 8008 So we really had a one two punch And would liken it later to having the motorbike the 4004 and the station wagon the 8008 Cause the 4004 was for control applications with a very small part count three or four chips and doing 4 bit arithmetic and the 8008 required some
59. y a bit so we met the goal In fact one thing that thought was kind of interesting when we looked at the sales of the 8008 and was once asked after the 8080 came along didn t the sales of the 8008 just evaporate And saw a plot of it about a year after the 8080 had come out What had happened is that the 8008 had ramped up in production and then just kind of leveled off with the 8080 and it just stayed constant and then the 8080 started to ramp up So apparently the sales continued on well after the 8080 came out Feeney One of the earlier customers was Kodak in a large photocopying machine Kodak was not an electronics company This was really one of the first electronic ventures that they got into Once they got it designed they didn t have the money nor did they have the inclination to go in and totally redesign it And there were many many applications like that So it s only when the applications totally turn over to something new that the design is changed Some of the first electronic cash registers in McDonald s utilized the 8008 And eventually it was replaced but the initial design was never upgraded to the 8080 Hoff That happened in spite of the idea that the 8080 was intended to be upward compatible from the 8008 CHM Ref X3706 2007 2006 Computer History Museum Page 28 of 30 Oral History Panel on Intel 8008 Microprocessor Feeney Certainly upward compatible but at the same time it required a reinvestment
60. y was accustomed to but there wasn t a recognition of the desperate need for documentation and more documentation And this little manual grew to be a much bigger manual as time went on And we all know the story today about the volume of books that are sold with every microprocessor that goes out Smith So we sold a bunch of those books too as recall Gelbach remember we were the first one in the industry to charge for literature which was unheard of But it worked Smith want to go back to Federico s frustration and Stan s comments You have to put it into perspective that Intel was founded solely as a memory company and its sole objective was to design and produce semiconductor memories We kind of stepped into this microcomputer business by chance It was never planned to be that way Everybody at Intel was really steeped in memory design and manufacturing so it s not hard to understand that there would be frustration and that there wasn t the commitment And of course nobody had any idea how large this was gonna get So it took a lot of pushing to get the microprocessor going Hoff When Intel was first formed Bob Graham and published an article predicting that we would get the cost of memory down to a penny a bit It was figured at that point we d compare with cores founda price list from think September of 1972 1103s were selling in like hundred quantities for around seven dollars so it was well under a penn
61. ys wanted to change the programs or somehow the way it worked They would start off and say Well we only need about a hundred of em We go into production we re gonna change it to a hardwired ROM And as it turned out none of the customers ever changed And the cost or at least the selling price of an EPROM was probably in the hundred dollar range and the cost of a ROM was in the three dollar range So we were more than enthusiastic about just letting the customer design and use all the EPROMs he wanted It was a significant profit margin for Intel in fact probably was the largest of any product Intel has ever made including microprocessors from a margin viewpoint So it turned out to be a really good strategy because EPROMs or programmable products they aren t EPROMs anymore have a large market House So you were more excited about selling the EPROMs than the microprocessors Gelbach In the beginning absolutely And you get smarter as the days go on Feeney Relative to the comment about getting smarter I d like to relate something that found as moved from engineering into marketing was kind of on the frontline as an application engineer working in the factory dealing with literally thousands of different customers calling in And instead of looking at the microprocessor through my own pair of eyes or Ted s or Stan s or Federico s you re looking at it through a thousand different pair of eyes Can we do

Oral History Panel on the Development and Promotion of the Intel

Contents

Download Pdf Manuals

Related Search

Related Contents