Having spoken to Germans who lived through WWII they put it well. "Back in 1939 when the war started, when they had fought in Poland. You knew history was being made. Of course we thought it was another history. But our daily life didn't change. Life went on at the homefront. You couldn't influence the front, so you kept going. We had kept going in 1940, and 41, and 42, and, and 43, and 44. Then in early 45 the Americans came. And life went on and we kept going."
It feels weird watching a history of some of the greatest human acheivements while living through some of its greatest failings. We can be so much better when we work together.
Oh please, there's always plenty of wars going on around the world even if it's not in the news. The same day that "what is going on right now" started happening at least three other countries where being bombarded, some others with ongoing genocide and man made mass starvation of civilians. Today is not worse than yesterday. Anyway continue with your videos. Didn't knew about Conway, what a legend!
Very good. I was a master's student in electrical engineering at the University of Melbourne, Victoria, Australia, in the early 1980s. The place was abuzz with VLSI, although outside my research area. I was surprised how students could design their own chips and have them made. The video explains what was happening then. A very exciting time. My own research was much more pedestrian but unique in the world, microcomputer control of controlled atmosphere apple stores. It is now used internationally as adaptive control of fruit and vegetable storage, where the controller monitors the response of the fruit itself, to dynamically and adaptively adjust storage conditions to give longer storage and reduce postharvest chemicals. Previously storage conditions were static.
Amazing, I've never heard of such a thing. Did you patent it? I did find on Google Patents a US patent (US7401469B2), originally assigned to GE, that besides the usual temperature and humidity levels monitors ethylene, a gas which ripe fruit produces.
@@raylopez99 No. I came up with the concept of adaptive control but then left the research area in the mid-1980s. It lay dormant for 20 years until it was picked up again and commercialised by others. However, I am cited as the original source of the concept. The controlled elements of controlled atmosphere storage are temperature, oxygen and carbon dioxide. Ethylene is usually removed rather than controlled during storage. The adaptive control is based on measuring the respiration rate, my concept, or the beginnings of skin damage (easy to detect). The controller takes temperature and oxygen levels down as low as possible without damaging the fruit; minimising the respiration rate. As the fruit begins to become distressed, the respiration rate begins to increase and/or the skin fluoresces. The controller then raises oxygen or temperature a little. It does this dynamically and adaptively over the storage time. It has been implemented for transport as well, to allow the sea freight of produce that would have otherwise been sent by air. The development and implementation of the concept of adaptive control have added new chapters to postharvest horticulture textbooks. See a history of the technology. I was involved with the work of Wolfe too. harvestwatch.net/history/
I'm forever fascinated by the 'hidden hand' of government in cutting edge technology as evidenced by the DARPA spending on integrated circuits. Maybe an intriguing subject for a future Asianometry video?
That's a fair point, one that ASML is intimately familiar with. Without the backing of the Dutch government, they likely wouldn't have survived their early days.
Great presentation! Some very minor observations: the term "VLSI" was already in common use before the Mead&Conway project. Mosis was an interface service for foundaries, but was not a foundary itself (you gave other examples of this, to which you can add Europractice). The key idea of Mead&Conway was to allow systems designers to create their own blocks instead of having to pass black boxes to specialized circuit designers. Some of the design would use standard cells, but for other parts you would connect transistors directly - for example doing a barrel shifter as a regular transistor lattice instead of a bunch of standard cell multiplexers.
i used to attend columbus state university in pursuit of a programming degree (BS in applied programming.) the 4 year degree was a product of the university and 2 world wide corporations (aflac & synovus) coming together to offer what these companies needed their new hires to know. In my 2nd year i took "circuit logic," and oh boy was it a pain to do so much math in binary. At the time i complained but afterwards was very thankful to have gotten a taste of why these things could be programmed the way they were. I saw nand and nor gates in my nightmares :D
My dad developed the standard block cells for GTE back in the day. The move to standard cells was a big deal at the time, since the company did low-volume runs for government use, but the chip designs were still very complicated. Later, GTE licensed a manufacturing process from Mitel, and my dad worked on the design rules for that, too. My favorite stories involved the ones where they built SOCs around the 6502. When they started doing that, they wanted to put a 6502 core in the center and built the rest of the logic around it, but didn't know how to test the finished cores for packaging. My dad suggested to just use two test probe masks: one to test the pins of the 6502, and another to test the rest of the logic. For some reason everyone else thought it was a crazy idea, but it worked. Testing designs was pretty hard back then. Before the introduction of standard cells, almost nothing worked on Rev 1!
@@ArneChristianRosenfeldt Basically, yes. A test mask is a plate with pins on it that lays on top of a wafer and connects to the pins of each chip on the die. They power up the chip and test it without having to attach bond wires or package the chip first. The trouble is, if your volume is really low, you can't justify building a unique test mask for each custom chip. Being able to use a single test mask for several different custom chip designs saves a lot of money. I'd link to a picture of a test mask, but RU-vid keeps deleting my comments if they contain links. Search for "wafer probe card" to see some modern examples.
(9:45) Back in the 1980s, the term "internet" was not commonly used. There were "ARPAnet", "CSnet", "BITnet", "DECnet", "UUCPnet" (UUnet), "JAnet", and in the UK they were using reversed domain names (e.g. "uk.ac.cam" form cambridge).
A great video that gave my minimal understanding a new way of thinking about VLSI. As upto now, i only thought it would refer to the amount of transistors used. Your video explained it as something completely different with it being a design language using standard parts. At the end you refer to it fading. I would like a video on that change and current methods used.
Great presentation. I was at Westinghouse Advanced Tech Labs in Baltimore in the early 80’s developing processes to accommodate VLSI designs. We were producing 1 micron geometries by ‘81.
Very interesting video. I recently graduated from uni in EEE, and VLSI was my favourite course I ever took. It's strange that you seem to allude that VLSI is not used anymore... Could you elaborate more on that? What came after, and what is used nowadays? I'm very interested in this field but I never realized I may have been studying outdated methods...
There are still blocks, but the current situation is akin to moving from assembly to a higher level language like C. The "compiler" will take your code that specifies the chip and do its best to make it work and spit out elementary blocks. The fun difference is while you're probably not going to have any trouble if you run your stuff at 10MHz, if you go to higher frequencies you start getting time violations because the signal just can't travel through too many gates in one cycle. In some cases, a different elementary block will work (but probably taking more space or current), but quite often, you will have to adjust your requirements a bit and allow some delays. With current systems, the timings tend to be too tight for people to pick block themselves and do all the calculations.
Thank you for putting this together for us. You deserve more followers. Your mom was right, I feel asleep the other day at bedtime listening to your paper.
Daisy, VALID SCALD System & Mentor Grapgics ... the first EDA Workstations ... during the second half of the 1980s ... competing wirh HILO on DEC VAXs ... it was a wild ride ... my best part of my career ... ;-)
The only thing that I find weird about this is the choice of calling both Carver Mead and Lynn Conway "Computer Scientists". While EE and CS are 2 sides of the same coin with great overlapping between the fields (specially at their time when CS had only recently started to branch out of EE/math), they both seem to have studied EE.
Old days when minimum feature size of latest semiconductor process was 2 micrometer and semiconductor was designed by hand and color bishop tape on acetate transparency. This design revolution enabled advent of standard cell based ASIC design.
I pulled the Mead and Conway Intro to VLSI Systems on my bookshelf while watching. As a student in 1984, I didn't understand how important that text was. It's hard to imagine now, but we'd draw designs out out on paper and colored pencils.
Would it be possible for you to make a video on node efficiency of different Fabs? Like how tsmc 7nm is vastly different than 8nm samsung despite being very close in "numbers" ?
How about a video on beyond cmos? After all we are quickly reaching the limit of transistor minimization. So the only variable we are left with to improve is energy efficiency and more or less speed.
5:31 "Allowed semiconductor designer to smoothly transition" is probably not meant to be a pun,but I laughed a bit too hard when I hard it. Tragically,Lynn Conway's transition went......less than smoothly.
China has programs to reengineering the original VLSI circuits and have started doing this as an independent entity so can never be called out on it, this may very well produce a much better design for electronics circuits producing better chips.
thanks for sharing, learning the history of the modern worlds technology is something that should be taught early in life. ive learned more from your channel about chip design, transistor history and evolution, and so many other inspirational facts on your channel. thank you for all the hard work you put into your videos, and i only hope that some youngins find these videos and are inspired to carry on the evolution of our tech.
problem is that by the time the professors learn to teach it, the state of the art has moved on. Part due to the amazing way AI is changing the 3D architectures.
With the ukraine situation in mind; What else, other than TSMC, would make a war-torn Taiwan catastrophic for the rest of the world/the west, specifically with regards to tech? Thanks for your vids, incredible quality.
Finally. Someone, anyone mentions. *CHARLES PROTEUS STEINMETZ* This guy is the absolute GOD of electric engineering. This guy went blind and deaf and thus proceeded to have absolutely furious arguments with his friend, Einstein, by tapping morse code on each other's knees. Imagine witnessing that. Better yet: imagine being capable of doing that.
Recently subscribed, and have been binge watching all your videos. I want to design an asic chip. Despite the lithography not currently existing, at some point 1nm will be a thing. How can I best prepare, from an academia standpoint, to get into asic chip design at 1nm?
China has had to relearn some of the lessons that were available at the university training sessions when VLSI was first introduced. TSMC already had that training from the start of their companies. Hence the catching up that was needed to be done, but not so much now! Just the fabrication equipment has been withheld.
The concept of standard cells still live on, but today there is so much software aid on the hard design that the finished product is a far cry from the Lego like approach of VLSI. And today the concept has moved to IP blocks. That is basically the same approach done differently: keep complexity down by segmenting out functionality blocks, but keep things flexible with standard interfaces for those blocks.
@@andersjjensen Thanks. Perhaps Xilinx or Altera might disagree, depending on how you define a 'standard cell'. Maybe a "sea-of-gates" design is what's obsolete? Not my field.
Would you please mention the known regions which have and potentially will have the most pure silicon raw materials. *In Syrian crisis few myths spread: the largest resivoir in the sea, and the purist silicon in the world!!! Which i doubt. Thanks.
Given that it was translated to Russian in the '80s, it was likely due to being considered a security risk. Hell, Conway herself was fired from IBM for being transgender largely due to the Lavender Scare creating fears of queer employees being potential vectors for Soviet infiltration.
John, really love your channel, but right now my country is under full blown invasion of russian army. We would really appreciate if you in some way will raise the awareness about our war. There are actually a lot of topics you'll enjoy covering, like processors for high precision missiles (or their algorithms) that target all the military objects in my country right now. Or at least about modern methods of propaganda in todays digital world (Taiwan is, in a way, in a quite similar situation to ours right now). Thanks.
ok seen enuff on old chip history Move on to New tech more interesting Quantum - and who's making what in chip tech etc Microscope circuits videos 100,000k nm all that stuff We cant see by eye