Very Large Scale Integrated Circuits or VLSI meant that Moore's law of doubling the number of transistors in integrated circuits starting with the Intel 4004 with 2300 transistors (a technological breakthrough) was off to a strong start after that chip launched on November 15th, 1971. Intel 4004 was intended to be sold to Japanese BUSICOM for electronic calculator manufacturing, since an electronic calculator essentially a simplified general purpose math computer.
Intel opted to sell 4004 for integrated electronics to a broad global audience and was right on the dot, in terms of sales strategy. Widespread integration of 4004 into electronics was the start of the 4th Industrial Revolution or #4IR as it known on social media networking platforms today, like Twitter.
During the 1970's only a handful of geniuses possessed the intimate knowledge needed across the rubylith design cut-out to the actually silicon gate manufacturing processes, so only a handful of companies even possessed the ability to both design the gates & manufacture the chips. Intel turned into the 900 gorilla of vertically integrated IC manufacturing, largely because of Federico Faggin's personal genius.
It was realized in the early 1980's that within 20 years individual chips would have millions of transistors, making them too complicated for any one person to design. This meant the future of chip design & chip fab was a collaborate effort mixing thousands of people with billions worth of specialized custom made equipment in the most expensive buildings ever made, industrial cleanrooms with chips fabs inside.
Palo Alto California partnership between the university Caltech and business Xerox PARC realized the need for IC silicon systems chips for personal computers in the late 1970's. Professor Carver Mead had popularized Gordon Moore's Law from the 1965 article about transistor density increasing & LSI's Lynn Conway who was an expert at computer architecture with experience at IBM.
Together they created a standardized system design methodology with design rules, creating VLSI or very large scale integrated circuits, essential for increasingly complex silicon transistor system chips to be made on high volume production lithography processes at chip fabs to enable broad market diffusion of IC technologies that ended up revolutionizing most economies worldwide. Today inroads of such tech are still happening at city level governments who are often 20 years behind the times technologically speaking.
Going back to the DC vs AC power wars, DC or direct current systems were easy to account for mathematically, while AC with sign waveforms requires calculus, such that AC was harder for most people to understand and harder to measure and account for power transmission rates & other metrics important for profit generation or revenue capture by commercial power plants that make and sell electricity in nearly real time as the grid lacks any meaningful energy storage capacity, such that nearly all electricity produced used at almost the same time today in middle 2023.
It was Conways length rules or lambda design principles that enables top down chip set designs, starting with the chip architecture, then to segments, sections, and blocks, so that following the design principles would flow down the process stream to enable very complex chip systems made of many integrated circuit blocks connected together into a giant system of logic gates.
These principles took cues from software design, where abstraction in higher level language, what I like to call syntax specific boolean algebra math language, later integrated by the compiler, then converted to machine code so the firmware on the logic can accomplish switching control of the circuits to enable the features and functions of the software happening at the user interface layer on the screen, keyboard & mouse inputs respectively.
Complementary Mixed Oxide Semiconductors or CMOS are the dominant node production technology used on 200mm & 300mm wafer production at many chip fabs around the world today. Block shrinking methods meant the same chip designs can be made on smaller process nodes to make chips with better power efficiency using less wafer area at lower cost per unit, the way Intel did with Atom processors for example. Generic cell libraries enabled chip designers to create novel IC's of incredible complexity without worrying about the grainy transistor level details, as the EDA design tools automate the wiring interconnects & block networking or in-chip data routing.
That way the human chip designer can focus on the higher goals of using the resulting chips to energize information technologies to interactively improve economies with emerging software solutions on CMOS hardware electronics. Consider robotic traffic intersection traffic light controllers and street level inductive loop sensing of vehicles, the steel cabinets housing the all weather 24/7 Allen Bradley logic systems that control the intersection lamp switching automatically without any human inputs, other than initial programming by civil traffic engineers who input the latter logic diagrams into such controllers or rather microcontroller systems.
The book "Introduction to VLSI Systems" sold well & created innovation positive loop feedback in the electronics industry worldwide. By 1983 there were more than 120 universities teaching VLSI classes. The guy who founded Silicon Graphics started by designing a geometry processor on the 6 micron process at Hewlett Packer, the precursor chips to all modern GPU chip designs. By enabling chip design to be modular, VLSI enabled the separation between chip design & chip manufacturing. The chip foundry model employed by TSMC today owes its origins to VLSI modularity principles designed by Lynn Conway. In this way people were able to specialize at specific parts of the chip process, that way each person involved did not need to be an expert at all parts.
This mirrors the RISC principle in ARM chips or the way high volume automobile production lines at Ford and other automakers that replaced hand made by craftsman luxury vehicles made slowly at high cost per unit in low volumes. This way people at the automakers can specialized, such that many tens of thousands of employees work together on parts of an automobile, on legal services, on manufacturing technology, on assembly, on integration, on testing, or in many other domains of marketing, planning, branding & similar. Thus modern industries allow people to specialized their knowledge, and experts emerge in specific domains.
DARPA funded hundreds of millions of university level VLSI research by 1982. Most of this was open public collaboration that fueled rapid innovation and very little was classified for military only use. Much of what was developed like powerful PC workstations, reduced instruction set computing, and computer aided EDA design software are still widely utilized today, even in bleeding edge upcoming smartphone SOC designs at AMD, Apple, Google, Intel, Nvidia, Samsung & Qualcomm.
PPW performance per watt a major design target to enable energy efficient compute, so that practical computers can enhance the engineering of CFD computational fluid dynamics so airplane designs & automobile designers can create enhanced aerodynamics so vehicles can get better fuel or energy economy with lower emissions by reducing drag created by atmospheric gas fluid effects interacting with vehicle surfaces when vehicles are traveling at higher speeds. With applied concepts like reducing turbulent gas flow by improving laminar flow that allows vehicles to slip through the air more easily.
You can see examples of this as underbody panels, upper rear window spoilers, aero channeling side view mirrors, wheel covers, anything that reduces vehicle drag improves fuel economy at highway speeds. Class 8 trucks upgrade with underbody aero kits & rear container aero kits realized lower fuel consumption that lower operating costs by enabling the trucks to achieve better fuel economy or to use less fuel to move the same containers filled with goods important to economic function in industrialized roadway enabled economies like America.
Multi-Chip Process at MOSIS means many chip designs from American universities could be made on the worlds first contract chip fab. It was this concept of a Chip Foundry servicing many smaller chip design teams that ultimately lead Morris Chang of Taiwan to create TSMC, today a $600 billion dollar giant in the chip fab sector.
In this way fabless chip companies like Apple can focus on M1, M2, M3,M4,M5 chip deigns while TSMC figures out how to use ASML EUV machines to make these chips in mass quantities for the next iPhone or iPad or MacBook Pro for example, electronic computer products that sell by the millions worldwide. Noting that many small children regularly use a smartphone today as of 2023, and newer iPhones with iOS are popular with small children due to the easy-to-use software touchscreen interface, fluid touchscreen UI experience, and ecosystem fluid exchange of data between many different Apple devices that are at home.
The internet back then in the early 1980's was limited to ARPANET, meaning that many universities were connected, but only a few big players had network access. Everyone knew everyone and it was a small number of key players sharing ideas.
193nm ArF systems made by ASML back in 2008 are still working daily 24/7 in legacy node chip fabs that make economically valuable automotive chips in the 65nm to 28nm node range. These expensive lithography machines and associated support equipment were machined & made using sophisticated expensive low volume manufacturing technique with exceptional QA:QC to minimize error & glitch at every stage from machine design, to early testing of prototypes, early pre-production alpha & beta units, the method for taking the machine apart into sections to ship in special containers to be later reassembled, tested, verified & then turned online for high volume continual production inside cleanrooms, in a process that takes a few technicians a few months working daily to complete.
Thus even legacy node chip fabs cost billions of dollars & took several years from breaking ground construction to high volume production of chips started up. When these node chips produced were state-of-the-art, back in 2009-2011, the resulting chips were sold at high prices that created enough revenue to pay for the chip fab, operating costs, all the energy, chemicals, equipment, labor costs, overhead & fixed costs. As new nodes came online at other newer chip fabs, the legacy nodes were fully depreciated, so that they could keep making chips at roughly the operating cost since everything else already paid for. This means that legacy node chips made in older fabs are very cheap to make so that they only sell for a few dollars each to automakers who use hundreds of them in modern automobiles for all the features & functions that come standards and in high trim forms known as limited models.
The increasing complexity of IC design caused chip designers to struggle in the 1970's, slowing IC technological progress because the ability to make photo masks of the chipset layer designs was initially constrained to hand drawn features on "rubylith" (sp?) dual layer polymer transparencies. Offset high volume printing etch technology used to make the thin metal image roller sheets applied to the rotating metal drums in off-set web printers was used to make photomasks, improving precision further while making replacement masks cheaper & easier to produce.
These photo-etched nickel metal alloy printer image plate photomasks were then used with high brightness light sources to layer etch photo resin or photoresist on silicon wafers to create the features that are then produced by chemical vapor deposition, physical chemical deposition, ion bombardment & other processing stages used to convert a high purity single crystal silicon boule wafer into dozens or hundreds of layers of transistors, capacitors, resistors in interconnects similar to in-chip wires to hook up all the transistors & other electronic components in the chip together into a complex very large scale integrated chipset, or high complexity integrated circuit like a CPU or GPU or memory chip.
1980 book by Mead & Carver sparked a revolution called VLSI enabling radical improvement in transistor density in later generations of IC or chips made in Chip Fabs that roughly speaking followed Moores law. Today a modern smartphone SOC can contain 20 billion transistors in many chips layered in the SOC such that the SOC or system on a chip contains the CPU, GPU, DRAM, and many hardware accelerators for video, voice, imaging, artificial intelligence, app specific functions, encoding, decoding, DSP, ADC, DAC, and other functions with an on-chip arbitrator chip communications system to control the moment of data within the SOC and into & out of the SOC to the screen, from the touch control layer, to the camera modules or from the camera modules, microphones, switches & similar used inputs.
Even voice by invocation of digital assistants like Siri on iPhone or Alexa on Echo devices or Ok Google on Android devices now, such that end users of these devices can use natural language to control the compute, apps, OS or other aspects of the device functionality, with the digital agents responding with natural language so that the user can listen and make sense of the communication output of such digital software agents. Yes, people are already & have been interacting with AI entities, they have birth dates & are increasing in both knowledge & skills progressively, in some ways already with super human intelligence by virtue of persistent internet access over high speed internet or 4G LTE or 5G NR network connections.
Going forward beyond the early 1990's VLSI block principles were abandoned as they wasted chip space or area on the wafer & newer EDA software tools enable tighter packing of chip features on the latest SOC designs for example.
Apple diverged away from Intel chips because their ARM style RISC Apple Silicon enables much better power efficiency in battery electric computers like the MacBook Pro M2 or upcoming M3 variants. This means much longer battery life on Apple Silicon electronics. It also means faster performance of iMovie, Final Cut Pro, or Safari the Internet browser in MacOS Ventura 13.3.1 today!
Apple Silicon leverages SOC principles to enable class leading energy efficient compute. This means playing Steam Games on an M2 MacBook Pro gives the same frame rates in CS:GO with less power consumed while player. It also means less waste heat for the fan & heatsink to remove when editing videos or exporting videos to upload YouTube content over WiFi 6 via Comcast Xfinity for example.
No comments:
Post a Comment