The Microchip Welcome to the History of Computing Podcast, where we explore the history of information technology. Because understanding the past prepares us for the innovations of the future! Todays episode is on the history of the microchip, or microprocessor. This was a hard episode, because it was the culmination of so many technologies. You don’t know where to stop telling the story - and you find yourself writing a chronological story in reverse chronological order. But few advancements have impacted humanity the way the introduction of the microprocessor has. Given that most technological advances are a convergence of otherwise disparate technologies, we’ll start the story of the microchip with the obvious choice: the light bulb. Thomas Edison first demonstrated the carbon filament light bulb in 1879. William Joseph Hammer, an inventor working with Edison, then noted that if he added another electrode to a heated filament bulb that it would glow around the positive pole in the vacuum of the bulb and blacken the wire and the bulb around the negative pole. 25 years later, John Ambrose Fleming demonstrated that if that extra electrode is made more positive than the filament the current flows through the vacuum and that the current could only flow from the filament to the electrode and not the other direction. This converted AC signals to DC and represented a boolean gate. In the 1904 Fleming was granted Great Britain’s patent number 24850 for the vacuum tube, ushering in the era of electronics. Over the next few decades, researchers continued to work with these tubes. Eccles and Jordan invented the flip-flop circuit at London’s City and Guilds Technical College in 1918, receiving a patent for what they called the Eccles-Jordan Trigger Circuit in 1920. Now, English mathematician George Boole back in the earlier part of the 1800s had developed Boolean algebra. Here he created a system where logical statements could be made in mathematical terms. Those could then be performed using math on the symbols. Only a 0 or a 1 could be used. It took awhile, John Vincent Atanasoff and grad student Clifford Berry harnessed the circuits in the Atanasoff-Berry computer in 1938 at Iowa State University and using Boolean algebra, successfully solved linear equations but never finished the device due to World War II, when a number of other technological advancements happened, including the development of the ENIAC by John Mauchly and J Presper Eckert from the University of Pennsylvania, funded by the US Army Ordinance Corps, starting in 1943. By the time it was taken out of operation, the ENIAC had 20,000 of these tubes. Each digit in an algorithm required 36 tubes. Ten digit numbers could be multiplied at 357 per second, showing the first true use of a computer. John Von Neumann was the first to actually use the ENIAC when they used one million punch cards to run the computations that helped propel the development of the hydrogen bomb at Los Alamos National Laboratory. The creators would leave the University and found the Eckert-Mauchly Computer Corporation. Out of that later would come the Univac and the ancestor of todays Unisys Corporation. These early computers used vacuum tubes to replace gears that were in previous counting machines and represented the First Generation. But the tubes for the flip-flop circuits were expensive and had to be replaced way too often. The second generation of computers used transistors instead of vacuum tubes for logic circuits. The integrated circuit is basically a wire set into silicon or germanium that can be set to on or off based on the properties of the material. These replaced vacuum tubes in computers to provide the foundation of the boolean logic. You know, the zeros and ones that computers are famous for. As with most modern technologies the integrated circuit owes its origin to a number of different technologies that came before it was able to be useful in computers. This includes the three primary components of the circuit: the transistor, resistor, and capacitor. The silicon that chips are so famous for was actually discovered by Swedish chemist Jöns Jacob Berzelius in 1824. He heated potassium chips in a silica container and washed away the residue and viola - an element! The transistor is a semiconducting device that has three connections that amplify data. One is the source, which is connected to the negative terminal on a battery. The second is the drain, and is a positive terminal that, when touched to the gate (the third connection), the transistor allows electricity through. Transistors then acts as an on/off switch. The fact they can be on or off is the foundation for Boolean logic in modern computing. The resistor controls the flow of electricity and is used to control the levels and terminate lines. An integrated circuit is also built using silicon but you print the pattern into the circuit using lithography rather than painstakingly putting little wires where they need to go like radio operators did with the Cats Whisker all those years ago. The idea of the transistor goes back to the mid-30s when William Shockley took the idea of a cat’s wicker, or fine wire touching a galena crystal. The radio operator moved the wire to different parts of the crystal to pick up different radio signals. Solid state physics was born when Shockley, who first studied at Cal Tech and then got his PhD in Physics, started working on a way to make these useable in every day electronics. After a decade in the trenches, Bell gave him John Bardeen and Walter Brattain who successfully finished the invention in 1947. Shockley went on to design a new and better transistor, known as a bipolar transistor and helped move us from vacuum tubes, which were bulky and needed a lot of power, to first gernanium, which they used initially and then to silicon. Shockley got a Nobel Prize in physics for his work and was able to recruit a team of extremely talented young PhDs to help work on new semiconductor devices. He became increasingly frustrated with Bell and took a leave of absence. Shockley moved back to his hometown of Palo Alto, California and started a new company called the Shockley Semiconductor Laboratory. He had some ideas that were way before his time and wasn’t exactly easy to work with. He pushed the chip industry forward but in the process spawned a mass exodus of employees that went to Fairchild in 1957. He called them the “Traitorous 8” to create what would be Fairchild Semiconductors. The alumni of Shockley Labs ended up spawning 65 companies over the next 20 years that laid foundation of the microchip industry to this day, including Intel. . If he were easier to work with, we might not have had the innovation that we’ve seen if not for Shockley’s abbrasiveness! All of these silicon chip makers being in a small area of California then led to that area getting the Silicon Valley moniker, given all the chip makers located there. At this point, people were starting to experiment with computers using transistors instead of vacuum tubes. The University of Manchester created the Transistor Computer in 1953. The first fully transistorized computer came in 1955 with the Harwell CADET, MIT started work on the TX-0 in 1956, and the THOR guidance computer for ICBMs came in 1957. But the IBM 608 was the first commercial all-transistor solid-state computer. The RCA 501, Philco Transac S-1000, and IBM 7070 took us through the age of transistors which continued to get smaller and more compact. At this point, we were really just replacing tubes with transistors. But the integrated circuit would bring us into the third generation of computers. The integrated circuit is an electronic device that has all of the functional blocks put on the same piece of silicon. So the transistor, or multiple transistors, is printed into one block. Jack Kilby of Texas Instruments patented the first miniaturized electronic circuit in 1959, which used germanium and external wires and was really more of a hybrid integrated Circuit. Later in 1959, Robert Noyce of Fairchild Semiconductor invented the first truly monolithic integrated circuit, which he received a patent for. While doing so independently, they are considered the creators of the integrated circuit. The third generation of computers was from 1964 to 1971, and saw the introduction of metal-oxide-silicon and printing circuits with photolithography. In 1965 Gordon Moore, also of Fairchild at the time, observed that the number of transistors, resistors, diodes, capacitors, and other components that could be shoved into a chip was doubling about every year and published an article with this observation in Electronics Magazine, forecasting what’s now known as Moore’s Law. The integrated circuit gave us the DEC PDP and later the IBM S/360 series of computers, making computers smaller, and brought us into a world where we could write code in COBOL and FORTRAN. A microprocessor is one type of integrated circuit. They’re also used in audio amplifiers, analog integrated circuits, clocks, interfaces, etc. But in the early 60s, the Minuteman missal program and the US Navy contracts were practically the only ones using these chips, at this point numbering in the hundreds, bringing us into the world of the MSI, or medium-scale integration chip. Moore and Noyce left Fairchild and founded NM Electronics in 1968, later renaming the company to Intel, short for Integrated Electronics. Federico Faggin came over in 1970 to lead the MCS-4 family of chips. These along with other chips that were economical to produce started to result in chips finding their way into various consumer products. In fact, the MCS-4 chips, which split RAM , ROM, CPU, and I/O, were designed for the Nippon Calculating Machine Corporation and Intel bought the rights back, announcing the chip in Electronic News with an article called “Announcing A New Era In Integrated Electronics.” Together, they built the Intel 4004, the first microprocessor that fit on a single chip. They buried the contacts in multiple layers and introduced 2-phase clocks. Silicon oxide was used to layer integrated circuits onto a single chip. Here, the microprocessor, or CPU, splits the arithmetic and logic unit, or ALU, the bus, the clock, the control unit, and registers up so each can do what they’re good at, but live on the same chip. The 1st generation of the microprocessor was from 1971, when these 4-bit chips were mostly used in guidance systems. This boosted the speed by five times. The forming of Intel and the introduction of the 4004 chip can be seen as one of the primary events that propelled us into the evolution of the microprocessor and the fourth generation of computers, which lasted from 1972 to 2010. The Intel 4004 had 2,300 transistors. The Intel 4040 came in 1974, giving us 3,000 transistors. It was still a 4-bit data bus but jumped to 12-bit ROM. The architecture was also from Faggin but the design was carried out by Tom Innes. We were firmly in the era of LSI, or Large Scale Integration chips. These chips were also used in the Busicom calculator, and even in the first pinball game controlled by a microprocessor. But getting a true computer to fit on a chip, or a modern CPU, remained an elusive goal. Texas Instruments ran an ad in Electronics with a caption that the 8008 was a “CPU on a Chip” and attempted to patent the chip, but couldn’t make it work. Faggin went to Intel and they did actually make it work, giving us the first 8-bit microprocessor. It was then redesigned in 1972 as the 8080. A year later, the chip was fabricated and then put on the market in 1972. Intel made the R&D money back in 5 months and sparked the idea for Ed Roberts to build The Altair 8800. Motorola and Zilog brought competition in the 6900 and Z-80, which was used in the Tandy TRS-80, one of the first mass produced computers. N-MOSs transistors on chips allowed for new and faster paths and MOS Technology soon joined the fray with the 6501 and 6502 chips in 1975. The 6502 ended up being the chip used in the Apple I, Apple II, NES, Atari 2600, BBC Micro, Commodore PET and Commodore VIC-20. The MOS 6510 variant was then used in the Commodore 64. The 8086 was released in 1978 with 3,000 transistors and marked the transition to Intel’s x86 line of chips, setting what would become the standard in future chips. But the IBM wasn’t the only place you could find chips. The Motorola 68000 was used in the Sun-1 from Sun Microsystems, the HP 9000, the DEC VAXstation, the Comodore Amiga, the Apple Lisa, the Sinclair QL, the Sega Genesis, and the Mac. The chips were also used in the first HP LaserJet and the Apple LaserWriter and used in a number of embedded systems for years to come. As we rounded the corner into the 80s it was clear that the computer revolution was upon us. A number of computer companies were looking to do more than what they could do with he existing Intel, MOS, and Motorola chips. And ARPA was pushing the boundaries yet again. Carver Mead of Caltech and Lynn Conway of Xerox PARC saw the density of transistors in chips starting to plateau. So with DARPA funding they went out looking for ways to push the world into the VLSI era, or Very Large Scale Integration. The VLSI project resulted in the concept of fabless design houses, such as Broadcom, 32-bit graphics, BSD Unix, and RISC processors, or Reduced Instruction Set Computer Processor. Out of the RISC work done at UC Berkely came a number of new options for chips as well. One of these designers, Acorn Computers evaluated a number of chips and decided to develop their own, using VLSI Technology, a company founded by more Fairchild Semiconductor alumni) to manufacture the chip in their foundry. Sophie Wilson, then Roger, worked on an instruction set for the RISC. Out of this came the Acorn RISC Machine, or ARM chip. Over 100 billion ARM processors have been produced, well over 10 for every human on the planet. You know that fancy new A13 that Apple announced. It uses a licensed ARM core. Another chip that came out of the RISC family was the SUN Sparc. Sun being short for Stanford University Network, co-founder Andy Bchtolsheim, they were close to the action and released the SPARC in 1986. I still have a SPARC 20 I use for this and that at home. Not that SPARC has gone anywhere. They’re just made by Oracle now. The Intel 80386 chip was a 32 bit microprocessor released in 1985. The first chip had 275,000 transistors, taking plenty of pages from the lessons learned in the VLSI projects. Compaq built a machine on it, but really the IBM PC/AT made it an accepted standard, although this was the beginning of the end of IBMs hold on the burgeoning computer industry. And AMD, yet another company founded by Fairchild defectors, created the Am386 in 1991, ending Intel’s nearly 5 year monopoly on the PC clone industry and ending an era where AMD was a second source of Intel parts but instead was competing with Intel directly. We can thank AMD’s aggressive competition with Intel for helping to keep the CPU industry going along Moore’s law! At this point transistors were only 1.5 microns in size. Much, much smaller than a cats whisker. The Intel 80486 came in 1989 and again tracking against Moore’s Law we hit the first 1 million transistor chip. Remember how Compaq helped end IBM’s hold on the PC market? When the Intel 486 came along they went with AMD. This chip was also important because we got L1 caches, meaning that chips didn’t need to send instructions to other parts of the motherboard but could do caching internally. From then on, the L1 and later L2 caches would be listed on all chips. We’d finally broken 100MHz! Motorola released the 68050 in 1990, hitting 1.2 Million transistors, and giving Apple the chip that would define the Quadra and also that L1 cache. The DEC Alpha came along in 1992, also a RISC chip, but really kicking off the 64-bit era. While the most technically advanced chip of the day, it never took off and after DEC was acquired by Compaq and Compaq by HP, the IP for the Alpha was sold to Intel in 2001, with the PC industry having just decided they could have all their money. But back to the 90s, ‘cause life was better back when grunge was new. At this point, hobbyists knew what the CPU was but most normal people didn’t. The concept that there was a whole Univac on one of these never occurred to most people. But then came the Pentium. Turns out that giving a chip a name and some marketing dollars not only made Intel a household name but solidified their hold on the chip market for decades to come. While the Intel Inside campaign started in 1991, after the Pentium was released in 1993, the case of most computers would have a sticker that said Intel Inside. Intel really one upped everyone. The first Pentium, the P5 or 586 or 80501 had 3.1 million transistors that were 16.7 micrometers. Computers kept getting smaller and cheaper and faster. Apple answered by moving to the PowerPC chip from IBM, which owed much of its design to the RISC. Exactly 10 years after the famous 1984 Super Bowl Commercial, Apple was using a CPU from IBM. Another advance came in 1996 when IBM developed the Power4 chip and gave the world multi-core processors, or a CPU that had multiple CPU cores inside the CPU. Once parallel processing caught up to being able to have processes that consumed the resources on all those cores, we saw Intel's Pentium D, and AMD's Athlon 64 x2 released in May 2005 bringing multi-core architecture to the consumer. This led to even more parallel processing and an explosion in the number of cores helped us continue on with Moore’s Law. There are now custom chips that reach into the thousands of cores today, although most laptops have maybe 4 cores in them. Setting multi-core architectures aside for a moment, back to Y2K when Justin Timberlake was still a part of NSYNC. Then came the Pentium Pro, Pentium II, Celeron, Pentium III, Xeon, Pentium M, Xeon LV, Pentium 4. On the IBM/Apple side, we got the G3 with 6.3 million transistors, G4 with 10.5 million transistors, and the G5 with 58 million transistors and 1,131 feet of copper interconnects, running at 3GHz in 2002 - so much copper that NSYNC broke up that year. The Pentium 4 that year ran at 2.4 GHz and sported 50 million transistors. This is about 1 transistor per dollar made off Star Trek: Nemesis in 2002. I guess Attack of the Clones was better because it grossed over 300 Million that year. Remember how we broke the million transistor mark in 1989? In 2005, Intel started testing Montecito with certain customers. The Titanium-2 64-bit CPU with 1.72 billion transistors, shattering the billion mark and hitting a billion two years earlier than projected. Apple CEO Steve Jobs announced Apple would be moving to the Intel processor that year. NeXTSTEP had been happy as a clam on Intel, SPARC or HP RISC so given the rapid advancements from Intel, this seemed like a safe bet and allowed Apple to tell directors in IT departments “see, we play nice now.” And the innovations kept flowing for the next decade and a half. We packed more transistors in, more cache, cleaner clean rooms, faster bus speeds, with Intel owning the computer CPU market and AMD slowly growing from the ashes of Acorn computer into the power-house that AMD cores are today, when embedded in other chips designs. I’d say not much interesting has happened, but it’s ALL interesting, except the numbers just sound stupid they’re so big. And we had more advances along the way of course, but it started to feel like we were just miniaturizing more and more, allowing us to do much more advanced computing in general. The fifth generation of computing is all about technologies that we today consider advanced. Artificial Intelligence, Parallel Computing, Very High Level Computer Languages, the migration away from desktops to laptops and even smaller devices like smartphones. ULSI, or Ultra Large Scale Integration chips not only tells us that chip designers really have no creativity outside of chip architecture, but also means millions up to tens of billions of transistors on silicon. At the time of this recording, the AMD Epic Rome is the single chip package with the most transistors, at 32 billion. Silicon is the seventh most abundant element in the universe and the second most in the crust of the planet earth. Given that there’s more chips than people by a huge percentage, we’re lucky we don’t have to worry about running out any time soon! We skipped RAM in this episode. But it kinda’ deserves its own, since RAM is still following Moore’s Law, while the CPU is kinda’ lagging again. Maybe it’s time for our friends at DARPA to get the kids from Berkley working at VERYUltra Large Scale chips or VULSIs! Or they could sign on to sponsor this podcast! And now I’m going to go take a VERYUltra Large Scale nap. Gentle listeners I hope you can do that as well. Unless you’re driving while listening to this. Don’t nap while driving. But do have a lovely day. Thank you for listening to yet another episode of the History of Computing Podcast. We’re so lucky to have you!