Comments by "Scott Franco" (@scottfranco1962) on "Asianometry" channel.

  1. Nvidia is a real success story. The only blemish is (as illustrated by Linus Torvald's famous giving the middle finger to them) is their completely proprietary stance on development. Imagine if Microsoft had arranged so that only their C/C# compilers could be used to develop programs for Windows. CUDA is a closed shop, as are the graphics drivers for Nvidia's cards.
    496
  2. My favorite application for mems is in aviation, since I am a recreational pilot. One of the first and most successful application of mems was accelerometers, which don't need openings in the package to work. Accelerometers can replace gyroscopes as well as enable inertial navigation, since they can be made to sense rotation as well as movement. With the advent of mems, avionics makers looked forward to replacing expensive and maintenance intensive mechanical gyroscopes with mems. A huge incentive was reliability: a gyroscope that fails can bring down an aircraft. The problem was accuracy. Mems accelerometers displayed drift that was worse than the best mechanical gyros. Previous inertial navigation systems used expensive laser gyros that worked by sending light pulses through a spool of fibre optical line and measuring the delay due to rotation. Mems accelerometers didn't get much better, but they are sweeping all of the old mechanical systems into the trash can. So how did this problem get solved? Well, the original technology for GPS satellite location was rather slow, taking up to a minute to form a "fix". But with more powerful CPUs it got much faster. But GPS cannot replace gyros, no matter how fast it can calculate. But the faster calculation enabled something incredible: the GPS calculation could be used to calibrate the mems accelerometers. By carefully calculating the math, a combined GPS/multiaxis accelerometer package can accurately and reliably find a real time position and orientation in space. You can think of it this way: GPS provides position over long periods of time,, but very accurately, and mems accelerometers provide position and orientation over short periods of time, but not so accurately. Together they achieve what neither technology can do on its own. The result has been a revolution in avionics. Now even small aircraft can have highly advanced "glass" panels, that give moving maps, a depiction of the aircraft attitude, and even a synthetic view of of the world outside the aircraft in conjunction with terrain data. It can even tell exactly which way the wind is blowing on the aircraft because this information falls out of the GPS/accelerometer calculation.
    197
  3. The FET, and especially the MOSFET, was a close equivalent to the way vacuum tubes work. This fact is often skipped in the history of transistors. In electronic design, the vacuum tube was an easily understood device. Put a voltage on the grid, stop current. Remove the grid voltage, current goes. On/off.
    142
  4. I think I told this story here before, but it bears repeating. In the early 80's windowed UV erasable proms were a thing. It was the time of Japan bashing, and accusals of their "dumping" on the market. We used a lot of EPROMs from different sources, and Toshiba was up and coming. We used mainly Intel EPROMs at the time. The state of the art back then was 4kb moving to 8kb (I know, quaint). Because of the window in the top of the EPROM you could see the chip. Most of us used this feature, when we had a bad EPROM, to get a little light show by plugging in the EPROM upside down and sealing the fate of the chip. Anyways, Intel and Toshiba were in a price war, so the chips from each vendor were about equivalent in price. But side by side in the UV eraser tray what you saw was shocking. The Toshiba chips were about 1/4 the size of the Intel chips. Yes, those "inferior" Japanese were kicking our a**es. Intel struggled along for a while, and exited the market for EPROMs. The "anti-dumping" thing had exactly one result. We could go to Japan, to the akihabara market (from street vendors!) and get chips with twice or four times the capacity of USA chips for cheap and bring them back in our luggage.
    130
  5. Good overview. I left verification just after the Verilog revolution came to pass. We used hardware simulators at that point, with accelerated simulation. Sea-of-fpga emulation was not yet popular. The bottom line was that the company I worked for (nameless, but huge trust me) was new to hardware chip design. They had lots of money to buy toys, but had the idea that the tool would solve all. Thus the expectation was that the new, more expensive tool would relive them from the need to write test cases. I enjoyed being the one employee who understood the tools, but rapidly realized that it was a falling rock I had volunteered to sit under. I left the company soon after.
    122
  6. Should mention here in the "dumping" era, yields (number of good chips on a wafer) in the USA were about %50, whereas yields in Japan were %90, and thus that was a good part of the cost advantage. The Japanese makers were also climbing the scaling curve. This was when Intel was still making EPROMs and they had transparent lids to allow UV erasure. To users of these parts, the difference between Intel and other US makers and Japan was obvious: you could clearly see the die sizes on US parts was twice as large as the Japanese parts for the same memory size. Thus the US electronics consumers took the whole "dumping" nonsense, led by Micron, with a large grain of salt. I recall in those days going to the Akihaba market in Tokyo and being able to buy memory chips from street vendors at 2-4 times the capacity of US available memories for half the price. And no, that is not a joke. Literally street vendors with one seller and a table.
    101
  7. As an aside, the book "Content Addressable Parallel Processors" by Caxton C. Foster (1976) discussed the ability to have a mass memory do computation. It is bit serial, but is parallel to all memory locations, meaning that you can do things like multiply all memory cells by the same number and similar operations. Its a good read.
    100
  8. MCMs faced two major issues during my time at Intel several years ago. The first was the most obvious: putting multiple chips in the same package meant the heat dissipation of the package now had to be designed for the two chips - ostensibly doubling the required heat dissipation. We handled that by throttling down the chip clocks, a process that has been largely automated by onboard temperature sensors and variable clock generators. IE., if you put two of today's advanced chips in a single package, they will self-throttle downwards. The second issue is that anytime you have a signal leave or enter a chip, it goes in or out via a pad driver, essentially an amplifier of that signal designed to drive the large capacitance of a PCB trace. This adds delay to the line. A lot of it. Internal trace drivers only have to worry about driving very short lengths of narrow and flat aluminium/copper. Not so with PCB traces. MCM pad drivers could be backed off a bit to compensate for minimal traces in the MCM, but typical MCMs unite chips that would normally be packaged separately, requiring a design change. PS., like the old movie "support your local gunslinger", keep in mind I am a stupid guy in the process world. The real process guys have to keep a fan on their heads to keep their brains from overheating :-)
    81
  9. A final vulnerability: Reverse engineering chips is an industry, and a highly automated one to boot. Shops take the chips, depackage them, then put them in etch baths and take progressive digital images as the layers come off, then use software to reverse the circuitry. And we'll call it "scotts law", you can create an RE shop for a fraction of the expense of creating a fab. There is a whole section of Silicon Valley devoted to just that.
    60
  10. As elsewhere said, properly "vi". A lot of programmers have professed their love to me of vi, but they are actually talking about vim. Vi was akin to using a chainsaw without a guard. We used it in the early days (1982 or so) for Unix on terminals in the 1200 baud range. Basically every key did something, which meant that if you accidently elbowed the keyboard, you were better off aborting the edit and starting over than to try to figure out what mode you had invoked. Worse, the crappy terminals and serial lines of the day would drop characters on the line, meaning that what you thought you were typing was not what you were editing anymore. For this, you had a "refresh key" that would cause the entire screen to be repainted, and you hit that anytime you had the strange feeling that your and the machine were no longer in sync. Loads o' fun.
    52
  11. Small nit: AMD started by making minicomputer components, that is, the chips that made CPUs before they were all single chips. Mini computers went from the 1960's into the 1980s and were still more powerful than single chip computers until the late 1980's/early 1990's. AMD also second sourced the 8080, then Zilog's Z800, and finally the 8086 series that started their long patent wars with Intel. AMD has come and gone as a competitor to Intel, more often than not a process generation behind that they "made up for" by being more efficient technologically. It was nice to see AMD finally pull ahead in both technology and process than Intel, but the game is not over. Intel is known for falling behind, then waking up and charging ahead again. We'll see. PS I designed hardware with AMD "bitslice" components in the early 1980's. IMHO AMD always was a step ahead of Intel in design, which Intel overcame by superior process and by sheer number of designers. I am an ex-Intel employee, but I bought Ryzen recently when I upgraded my home system.
    49
  12. To be specific, they are not a radiation hazard (they can be handled), but rather, like all heavy metals, they are poisonous and can contaminate things like ground water.
    41
  13.  @janlanik2660  I think you misread what I said. Microsoft (or any OS maker, Apple included) could have easily made it so that only their compilers could be used on their systems, no GCC, no independent developers. That is what Nvidia has done.
    36
  14. When I came to San Jose in 1987, there was a bit of a scandal that got printed up in the Mercury news. Seems a tester for the city had found a lot of toxic materials in the sewer, heavy metals and chemicals. That turns out to be easy to track. They just keep taking samples from the sewer upstream until they find where the materials no longer exist, then they have found the outlet of the chemicals. At that time, Intel was running a fab, and they were cited. However, there were others on the list. As the video says, it was a shock to many here who believed that the silicon fabs were "clean" industries.
    33
  15. Thailand, as far as having a native disk drive company, lost to Allan Shugart, my old boss at Seagate. They had lots of company. Seagate would sell their grandmother for a few cents of cost advantage. They are a shining example of how an American corporation stays on top, mainly by subsuming any advantages the competition has at any cost. Do you want a Thai native disk drive company? Seagate is probably as close as you can come. They do everything there.
    26
  16. Verilog and VHDL took most of our fun away. I liked the schematic method of design a lot. We used the same workstation for layout as well.
    24
  17. So start some. My brother in law from Colombia is high tech, but had to go to Texas to get a good job, even though he still lives in Medellin. Your country has the talent. Just stop exporting it.
    24
  18. Evolution in nature also followed this curve. Nerves started slow, but sped up using electrochemical reactions. It was a dead end. Nature didn't give up on speeding up neurons as much as discovered the advantages of parallelism. Your brain is a relatively slow communicating, but massively parallel machine.
    24
  19. Great video on one of my favorite subjects. I'd like to add a couple things. First of all (as the poster below said), this history skips a very important branch of IC history, the gate array, which FPGAs (which are a namesake, the Field Programmable Gate Array). Basically gate arrays were ICs that consisted of a matrix of transistors (often termed gates) without the interconnect layers. Since transistors then, and largely even today, are patterned into the silicon wafer itself, this divided the wafer processing into two separate divisions, the wafer patterning, and the deposition of aluminum (interconnect). In short, a customer could save quite a bit of money by just paying for the extra masks needed to deposit interconnects, and take stock wafers to make an intermediate type of chip between full custom and discrete electronics. It was far less expensive than full custom, but of course that was like saying that Kathmandu is not as high as Everest. Xilinx used to have ads showing a huge bundle of bills with the caption "does this remind you of gate array design? Perhaps if the bills were on fire". Altera came along and disrupted the PLA/PAL market and knocked over the king o' them all the 22V10, which could be said to be the 7400 of the PAL market. They owned the medium scale programmable market for a few years until Xilinx came along. Eventually Altera fought back, but by then it was too late. However, Altera got the last word. The EDA software for both Xilinx and Altera began to resemble those "bills o' fire" from the original Xilinx ads, and Altera completely reversed its previous stance to small developers (which could be described as "if you ain't big, go hump a pig") and started giving away their EDA software. Xilinx had no choice but to follow suit, and the market opened up with a bang. There have been many alternate technologies to the RAM cell tech used by Xilinx, each with an idea towards permanently or semipermanently programming the CLB cells so that an external loading prom was not required. Some are still around, but what was being replaced by all that work and new tech was serial EEPROM that was about 8 pins and approximately the cost of ant spit, so they never really knocked Xilinx off its tuffet. My favorite story about that was one maker here in the valley who was pushing "laser reprogrammability", where openings in the passivation of a sea of gates chip allowed a laser to burn interlinks and thus program the chip. It was liternally PGA, dropping the F for field. It came with lots of fanfare, and left with virtual silence. I later met a guy who worked there and asked him "what happened to the laser programmable IC tech?". He answered in one word: contamination. Vaporising aluminum and throwing the result outwards is not healthy for a chip. After the first couple of revs of FPGA technology, the things started to get big enough that you could "float" (my term) major cells onto them, culminating with an actual (gasp) CPU. This changed everything. Now you could put most or all of the required circuitry on a single FPGA and the CPU to run the thing as well. This meant that software hackers (like myself) could get into the FPGA game. The only difference now is that even a fairly large scale 32 bit processor can be tucked into the corner of one. In the olden days, when you wanted to simulate hardware for an upcoming ASIC, you employed a server farm running 24/7 hardware simulations, or even a special hardware simulation accellerator. Then somebody figured out that you could lash a "sea of FPGAs" together and load a big 'ole giant netlist into it and get the equivalent of a hardware simulation, but near the final speed of the ASIC. DINI and friends were born, large FPGA array boards that cost a couple of automobiles to buy. At this point Xilinx got wise to the game, I am sure. They were selling HUGE $1000 per chip FPGAs that could not have a real end consumer use.
    21
  20. Actually Microsoft forced that. Even besides , Itanium, Intel was ready to introduce an incompatible 64 bit upgrade to x86, and Microsoft didn't want to have to support two different instruction sets, so they forced the two to come to terms. Its a classic story that I assume will be covered here in part deux.
    18
  21. You already reached 100k???? Anyways, good video, well researched. Thank you.
    17
  22. The "dumping" wave was led by Micron for their own benefit. They were perhaps the first tech company that showed they could have a big effect by spreading money around Washington. The rest of us paid for that. DRAMs in the early days were a big cost item for computers. The 64k "wall" was a big step, forgotten now. It meant that the entire memory space of the 16 bit address processors could be filled with a single chip, and led to the implementation of transparent bootstrap roms, which, with a cpu hardware bit, could be read, but writes would go to main memory. This allowed the computer to copy it's bootstrap rom into dram, then turn the bit off and run in full 64kb with dram only.
    15
  23. ​ @marctemura2017  Sorry what had that to do with my reply?
    14
  24. Just a quick word. I the early days, what they called a "detector" or radio frequency detector, was just a (relatively) high frequency low current diode. For amplitude modulation, if you perform half wave rectification of an RF signal, you are left with high frequency DC waves with AM modulation. Those high frequency components of the DC signal would be rolled off by the primitive circuits of the day, and thus leave just the DC AM modulation. Thus even a point contact diode could receive AM broadcast on a strong signal, especially if listened to with headphones. Thus they called them "detectors", really just a rectifier in today's language.
    13
  25.  @bunyu6237  I think others would supply better info than I on this subject, since I haven't been in the IC industry for decades, back in the late 1980's. At that time, the industry (reverse engineering) was moving from hand reversing to fully automated reversing. However, if you don't mind speculation, I would say there is no concrete reason why the reversing industry would not have kept up with newer geometries. The only real change would have been that its basically not possible to manually reverse these chips anymore. I personally worked on reversing a chip at about 4 generations beyond the Z80, which was not that much. At that time, blowing up a chip to the size of a ping-pong table was enough to allow you to see and reverse engineer individual transistors and connections. Having said that, I have very mixed feelings about the entire process. I don't feel it is right to go about copying others designs. I was told at the time that the purpose was to ensure compatibility, but the company later changed their story. On the plus side, it was an amazing way for me to get onboard the IC industry. There is nothing like reverse engineering a chip to give you a deep understanding of it. However, I would say I think I would refuse to do it today, or at least try to steer towards another job. For anyone who cares about why I have a relationship to any of this, I used to try and stay with equal parts of software and hardware. This was always a difficult proposition, and it became easier and more rewarding financially to stay on the software side only, which is that I do today. However, my brush with the IC industry made a huge impression on me, and still shapes a lot of what I do. For example, a lot of my work deals with SOCs, and I am part of a subset of software developers who understand SOC software design.
    12
  26. Well good news and bad news, Joe. Good news is all the freaking money in the world is available for you complete this project. Bad news is everyone in the company is watching if you succeed, and nobody is even sure its possible. Good talk. Meeting at 11am.
    11
  27. I am guessing that GM is looking to transfer as much liability as they can to LG. Surely this has wide ranging consequences.
    10
  28. Depends. On PCBs, when very high accuracy is required, they laser trim them. We used laser trimmers on resistors in hybrids as well, really nothing beats it for accuracy. There was an outfit selling laser configured ICs, like an FPGA but the fuses were laser burned. The company failed, and I ran into one of their engineers later and asked him why. "contamination" -- was the answer.
    10
  29. Maybe ranching is a better use of the land up there. Here in California they basically forced agriculture into the central valley, including stupid sh*t like rice farming, as if the place was a tropical rainforest. What we have now is water shortages and salt water incursion issues. Making the whole world into verdant farmland is a 1800's fantasy.
    9
  30. I think these go in two different directions. The old flow is graphical floorplanning, cell based design, etc. The new way is Verilog, tested on FPGAs and sent off to a "conversion" firm or to the end fab that takes the design and runs it though automated floorplanning and process prep. All of the SOCs I have been involved with for the last 15-20 years have been done this way. The nice thing is the Verilog route is going to be much more amenable to open source processing.
    9
  31. Just one small addition: When Intel pushed onboard graphics, where the graphics memory was part of the main memory of the CPU, it was thought that the video solution would actually be faster, since the CPU would have direct access to the frame buffer, as well as having all of the resources there to access it (cache, DMA, memory management, etc). The reason they lost that advantage in the long run was the dual advantages of VRAM or dual ported video ram, a ram that could both be read and written by the CPU at the same time as being serially read out to scan the video raster device, as well as the rise of the GPU, meaning that most of the low level video memory access was handled by a GPU on the video card that did the grunt work of drawing bits to the video ram. Thus Intel ran instead down the onboard video rabbit hole. Not only didn't they win the speed race with external video cards, but people began to notice that the onboard video solutions were sucking considerable CPU resources away from compute tasks. Thus the writing was on the wall. Later, gamers only knew the onboard video as that thing they had to flip a motherboard switch to disable when putting a graphics card in, and nowadays not even that. Its automatic.
    8
  32. Probably not a bad idea to also go into flat panel makers, which has become a "winner take all" industry much like ICs, and also rides the scaling curve.
    7
  33.  @harrickvharrick3957  The technique is to remove the passivation on the chip (top level of glass), then use a scanning electron microscope to read the values of memory cells and other elements on the chip. They use this to perform live debugging on chips as well. I have seen it in action. The operator can literally tell the machine what trace to watch, then it outputs a scope trace as if you had probed it. Its very impressive and very expensive technology. The simple answer is there is nothing that cannot be REed with effort.
    7
  34.  @janlanik2660  I see that is a recent addition (opencl on nvidia). However, that was not really what I was talking about. I don't think nvidia yet has opened their architecture so that anyone can write a driver or compiler for their machine, which is what Linus was talking about. If I am incorrect about that, I would be happy to be so.
    7
  35.  @psd993  ATI is known for opening their arch. Imagine if CPU makers had a closed architecture, and only allowed certain compiler makers to work with their products.
    7
  36. Who cares? I don't. The man does his research. Sit down and eat your jujubees.
    7
  37. I think of these as the baseball, then basketball wafers. I did fab work at the baseball generation, which was nice because you could easily carry around the wafers. By the time basketball wafers came out, I was out of the fab business. Now the next looks like "pizza" wafers. I plan to graduate to breathing dirt around that time.
    7
  38. Side note: IBM (back when they were a major fab power), bet big on X-Ray machines. The epilog to that chapter appeared in Spectrum magazine, who detailed that a visit to the IBM fab saw those machines sitting out back under tarps. End of dream.
    6
  39. In the category of alternative materials to silicon, unmentioned are diamond (carbon) and gallium arsenide. There was also a vogue for sapphire as a base material (coating of the wafer prior to patterning), which if this old man recalls was SOS or Silicon on Sapphire.
    6
  40. Just a comment about MOSFETs verses bipolar. MOSFETS are basically a capacitor. Its not rocket science to figure out that a capacitor is going be slower than a transistor that is not based on a capacitor. Moving stored charge around costs time. However, simple math would tell you that as ICs scale, the capacitor would become less and less of an issue, until a crossover point is reached. In the 1980s, I programmed an IC simulator that modeled transistors as an RC (resistor-capacitor) with a switch. IE, the function of the waveform out is that of a resistor discharging a capacitor. The idea comes from Bell Labs, and it allows analog simulation of transistor networks with a dramatically simpler model than SPICE, and thus much faster math than SPICE, which must carry out a relaxation because of reactive effects.
    6
  41. What people miss about the RISC revolution is that in the 1980s with Intel's 8086 and similar products, the increasingly complex CPUs of the day were using a technique called "microcoding", or a lower level instruction set inside the CPU to run instruction decoding, etc. It was assumed that the technique, inherited from mini and mainframe computers, would be the standard going forward, since companies like intel were increasing the number of instructions at a clip. RISC introduced the idea that if the instruction set were simplified, CPU designers could return to pure hardware designs, no microcode, and use that to retire most or all instructions in a single clock cycle. In short, what happened is the titanic turned on a dime: Intel dropped microcode like a hot rock and created pure hardware CPUs to show that any problem could be solved by throwing enough engineers at it. They did it by translating the CISC x86 instructions to an internal RISC form and deeply parallelizing the instruction flow, the so called "superscalar" revolution. In so doing they gave x86 new life for decades. I worked for SUN for a short time in the CPU division when they were going all in on multicore. The company was already almost in freefall. The Sparc design was flawed and the designers knew it. CEO Johnathan faced questioning at company meetings when he showed charts with Java "sales" presented as if it were a profit center (instead of given away for free). I came back to SUN again on contract after the Oracle merger. They had the same offices and the little Java mascots on their desks. It was probably telling that after my manager invited me to apply for a permanent position, I couldn't get it though their online hiring system, which was incredibly buggy, and then they went into a hiring freeze so it was irrelevant. I should also mention that not all companies did chip design in that era with SUN workstations. At Zilog we used racks full of MicroVaxes and Xwindow graphics terminals. I still have fond memories of laying out CPUs and chainsmoking in the late 1980s until midnight.
    5
  42.  @allentchang  Hummm.... back in 1987 it was LSI workstations, if I recall. I don't know the operating system, but I believe they were Tectronix graphics terminals (Zilog). They were not fast, but very high resolution for the day. In 1993 it was Apollo workstations (Seagate), which were running Mentor. It certainly ran a Unix variant, but it was an unusual one. Into the new century, its all been Verilog using Xilinx software (various startups), running on Windows (does Xilinx even run on Linux/Unix?). Our fabs also tell the story: last century it was custom fab (Zilog), then AT&T fab (Seagate), then after that probably TSMC, I don't recall. Afternote: Actually I do recall, at Zilog the Tek terminals were driven by racks and racks of LSI-11's, a PDP-11 that fit in a single or double RU. I remember because we had a big serial port mux that would allow you to get a connection to any of the machines. I used to write scripts that would start jobs on multiple machines overnight, which was the only way to get reasonable simulations of chips. I believe they were running Unix. Our chip simulations were done on a custom gate level simulator that I learned a lot from, since it would simulate things like domino logic. And yes, I am old.
    5
  43. I won't be holding my breath. The world's champion chess machine/program, Deep blue, used repetitive pathfinding to beat Gary Kasparov, not AI. Floorplanning and routing uses the same principles, and can use the same brute force attacks. Lets say I have my doubts. The most productive area for AI has, so far, been the obtaining of grants.
    5
  44. @D R That's a great question. In fact we could just use the altitude from the GPS system. Right now you dial in the barometric "base pressure", or pressure at sea level. This is used to calibrate the altimeter so that it delivers accurate results, I believe it is within 100 feet of accuracy (other sources say the FAA allows 75 feet of accuracy). Its a big big deal. A few feet could mean if you hit a building or pass over it. Thus when you fly, you are always getting pressure updates from the controller, because you are going to need updates that are as close as possible to the pressure in your area. So why not use the GPS altitude, which is more accurate? 1. Not everyone has a GPS. 2. Even fewer have built in GPS (in the panel of the aircraft). 3. A large number of aircraft don't re calibrate their altimeters at all. 4. New vs. old. Aircraft have been around for a long time. GPS not so much. If you know a bit about aircraft, you also know that number 3 there lobbed a nuclear bomb into this conversation. Don't worry, we will get there. First, there is GPS and there is GPS, implied by 1 and 2. Most GPS units in use are portable (in light aircraft). Long ago the FAA mandated a system based on transponders called "mode-C" that couples a barometric altimeter into the transponder. OK, now we are going into the twistly road bits. That altimeter is NOT COMPENSATED FOR BASE PRESSURE. In general the pilot does not read it, the controller does (ok most modern transponders do read it out, mine does, but an uncompensated altitude is basically useless to the pilot). The controller (generally) knows where you are, and thus knows what the compensating pressure is (no, he/she does not do the math, the system does it for them). Note that GPS had nothing to do with that mode C discussion. So for the first part of this, for a GPS to be used for altitude, the pilot would have to go back to constantly reporting his/her altitude to the controller. UNLESS! You could have a mode S transponder, or a more modern UAT transceiver. Then, your onboard GPS automatically transmits the altitude, and the position, and the speed and direction of the aircraft. Now we are into equipage. Note "onboard GPS". That means built into the aircraft. Most GPS on light aircraft are handheld, which are a fraction of the cost of built in avionics. Please lets not get into why that is, its about approved combinations of equipment in aircraft, calibration, and other issues. The mere mention of it can cause fistfights in certain circles. Ok, now lets get into number 3. If you are flying over, say, 14,000 feet, its safe to say you are not in danger of hitting any mountains, or buildings, or towers. Just other aircraft. So you don't care about pressure compensation. So the rules provide that if you are over 18,000 feet, you reach down and dial the "standard pressure" of 29.92 inches of mercury, which the FAA has decreed is "standard pressure" (the FAA also has things like standard temperature, standard tree sizes, etc. fun outfit). So what does that mean? Say you are TWA flight 1, and TWA flight 2 is headed the opposite direction, same altitude. Both read 18,000 feet. Are they really at 18,000 feet? No, but it doesn't matter. If they are going to collide, they are in the same area, and thus the same pressure. Meaning that their errors cancel. It doesn't matter that they are really at 19,123 feet, they both read the same. Thus climbing to 19,000 (by the altimeter) means they will be separated by 1,000 feet. So the short answer is the final one. The barometric system is pretty much woven into the present way aircraft work. It may change, but it is going to take a long time. Like not in my lifetime.
    5
  45.  @DumbledoreMcCracken  I'd love to make videos, but I have so little free time. Allen was a true character. The company (Seagate) was quite big when I joined, but Allen still found the time to meet with small groups of us. There were a lot of stories circulating... that Allen had a meeting and fired anyone who showed up late because he was tired of nobody taking the meeting times seriously, stuff like that. He is famous for (true story) telling a reporter who asked him "how do you deal with difficult engineers".. his answer "I fire them!". My best story about him was our sailing club. I was invited to join the Seagate sailing club. They had something like a 35 foot Catalina sailboat for company use, totally free. We ended up sailing that every Wednesday in the regular race at Santa Cruz Harbor. It was owned by Allen. On one of those trips, after much beer, the story of the Segate sailboat come out. Allen didn't sail or even like sailboats. He was a power boater and had a large yacht out of Monterrey harbor. He rented a slip in Santa Cruz, literally on the day the harbor opened, and rented there since. The harbor was divided in two by a large automobile bridge that was low and didn't raise. The clearance was such that only power boats could get through, not sailboats (unless they had special gear to lower the mast). That divided the harbor into the front harbor and back harbor. As more and more boats wanted space in the harbor, and the waiting list grew to decades, the harbor office came up with a plan to manage the space, which was "all power boats to the back, sailboats to the front", of course with an exception for working (fishing) boats. They called Allen and told him to move. I can well imagine that his answer was unprintable. Time went on, and their attempts to move Allen ended up in court. Allen felt his position as a first renter exempted him. The harbor actually got a law passed in the city to require sailboats to move to the back, which (of course) Allen termed the "Allen shugart rule". Sooooo.... comes the day the law goes into effect. The harbormaster calls Allen: "will you move your boat". Allen replies: "look outside". Sure enough, Allen moved his yacht to Monterrey and bought a used sailboat which was now in the slip. Since he had no use for it, the "Seagate sailing club" was born. It was not the end of it. The harbor passed a rule that the owners of boats had to show they were using their boats at least once a month. Since Allen could not sail, he got one of us to take him out on the boat, then he would parade past the Harbormaster's office and honk a horn and wave. Of course Allen also did fun stuff like run his dog for president. In those days you either loved Allen or hated him, there was no in-between. I was in the former group, in case you could not tell. I was actually one of the lucky ones. I saw the writing on the wall, that Segate would move most of engineering out of the USA, and I went into networking for Cisco at the time they were still throwing money at engineers. It was a good move. I ran into many an old buddy from Seagate escaping the sinking ship later. Living in the valley is always entertaining.
    5
  46. So if the biggest fab goes to subsistence level by overproducing and oversupplying the market, and that causes the smaller fabs to go under, I am guessing the big fabs don't shed tears.
    5
  47. Spectre works from the fact that speculative execution can access pages that are normally forbidden, because if the access were to actually go through, it would terminate with a page fault. The "cracker" code then uses timing differences to read the fetched data. As an example (I'm probably messing this up, its been a long time) the speculative code can mask a single bit from the illegal data location, then use that as an address indirect to two pages, one with the bit 1 and another with the bit 0. The two pages have known access speeds, so the timing difference of execution tells you if the bit is set or not. Then you repeat for each bit. Thus you can know the contents of data at a location even though you cannot directly access it. This kind of thing can be easily prevented by proper CPU design, such as stopping speculative access immediately for addresses that do not have permissions. It can be stopped in software by randomizing the locations of sensitive data or even the ordering of pages. I recall a big push for Linux kernel designers to randomize the page placements of the OS to prevent reading of locations in the kernel.
    5
  48.  @HsingG32  Proprietary hardware is a pet peeve of mine. You are locking out independent developers, and there is no reason for it. If you open the specs to your hardware you will just increase the market for your hardware, and nobody (who didn't already have the capability) is going to be able to copy your hardware just by the spec.
    5
  49. Its a good question (further research into hard drives). They are still doing some amazing things, advanced magnetic materials, layered recording, etc. However, the basis of the industry is electromechanical, which means it is inherently slower and more complex than SSDs. You can only move a mass (head arm) so fast. The recent research in disk drives has gone mainly to increasing their density, and therefore reducing cost. Because this does nothing to help the speed disadvantage of HDDs, this trend will actually accelerate the demise of the HDD industry, because it accelerates the trend of HDDs towards being a backup medium only. HDDs cannot get any simpler. They have two moving parts: the head and the disks, and both probably spin on air now (certainly true of heads, not sure about spindles). Because HDDs are more complex and take more manufacturing effort than SDDs, the cost advantage of HDDs is an illusion. The fall is near.
    5
  50. IC masking and screen printing: Well, I think more accurate to say that these techniques were well known from the manufacture of printed circuit boards, which were in full swing at the time of the first ICs, and from there you get back to printing, both screen printing and lithography. Also, resists were in use before ICs, used to perform etching on metal, rock and other surfaces, which is very much a thing today. In fact, etching glass with acid, still done today, is almost a direct line to ICs, since silicon dioxide is basically glass.
    5
  51. Not sure Applied qualifies as unappreciated. The Wall Street Journal follows them closely, and has for decades. My favorite story about them was I briefly dated a woman who worked there. She was PHD physics, and standing to close to her brain could give you a sunburn. I asked her how she liked working at applied, she said "I like it, but don't like the fact they make me talk to CUSTOMERS". IE, applied employs brainiacs.
    4
  52. I feel like I should wrap this subject with an opine on reverse engineering (RE) in general. IMHO this tends to be a glint in the eye mainly of incompetent engineers and managers who like the idea of cutting engineers out of the process. Engineers are troublesome little critters who bitch a lot and eventually discover the joys of sex and the consequent burdens of children. I have always found that it is easier to create software/hardware designs from scratch than to copy the work of others, since REing others work makes it harder and also because the state of software design today is typically crappy, poorly documented code. I often do scratch designs even when given others' source code. REing a chip just gives you a fat binder of schematics (probably a lot of fat binders nowadays). It does not give you the design of the chip, and the design processes of the chip are farther than ever from the schematic thanks to the size of the designs and the fact most of it is done in Verilog. I'm willing to bet most RE today is done by silicon makers wanting to figure out how that new stacked DRAM cell works.
    4
  53.  @bunyu6237  A couple of reasons (better software than hardware). First of all, there is a larger group of people working on software than hardware, so the jobs are more plentiful and the demand greater. Second, hardware/software crossover people are considered odd birds, and when I used to do that I had people literally telling me to "pick a side", go one way or the other. I find it easier to get and do software projects, and the pay is better. I dabbled in Verilog long after I stopped being paid for hardware design, and I realized it would take a lot of work to get a foothold in good Verilog design with virtually no corresponding increase in salary, and more likely a decrease for a while during the time I gain credibility as a Verilog designer. The last time I was paid to design hardware it was still schematic entry (and yes, in case you haven't figured that out, I am indeed that old). Of course, a lot of this is my personal situation. I am not sure any of the above would serve as career advice. I definitely consider my hardware background to be a career asset, since specialize low level software design (drivers, embedded, etc). Having said that, I keep up with hardware advances and have often dreamed of uniting my Verilog experience with software experience. That dream is unrealized.
    4
  54. Good job. There was a bit of conflation there with microcode (its firmware?). It would have helped to underline that it is entirely internal to the chip and operates the internals of the CPU. In any case, microcode was discarded with the Pentium series, KINDA. It actually lives on today in so called "slow path" instructions like block moves in the later cpus, which use microcode because nobody cares if they run super fast or not, since they are generally only used for backwards compatibility and got deprecated in 64 bit mode. I await the second half of this! Things took off again with the AMD64 and the "multicore wars". Despite the mess, the entire outcome probably could have been predicted on sheer economic grounds, that is, the market settling into a #1 and #2 player with small also-rans. Today's desktop market, at least, remains in the hands of the x86 makers except for the odd story of Apple and the M series chips. Many have pronounced the end of the desktop, but it lives on. I have many or even most colleges who use Apple macs as their preferred development machines, but, as I write this, I am looking out at a sea of x86 desktop machines. Its rare to see a mac desktop, and in any case, at this moment even the ubiquitous Mac pro laptops the trendsetters love are still x86 based, although I assume that will change soon. Me? Sorry, x86 everything, desktop and laptop(s). At last count I have 5 machines running around my house and office and 4 laptops. I keep buying Mac laptops and desktops, cause, you know, gotta keep up with things, but they grow obsolete faster than a warm banana. Yes, I had power PC Macs, and yes they ended up in the trash. And yes, I will probably buy Mac M2s at some point.
    4
  55. Right out of Stienbeck's "the pearl", which, spoiler alert, is not the Pearl.
    4
  56. First class analysis.
    4
  57. I toured the NUMI plant before it closed, and I toured it again after Tesla started making cars there. I also went back their to pick up my wife's Tesla later.
    3
  58. Nothing has changed. Most true multicore/threaded designs are one-off. Think supercomputer clusters. True multithread/multicore software engineers are hard to find. The average software engineer is scared to death of multithreaded designs, and it would not be a stretch to imagine that most multithread apps have latent bugs.
    3
  59. Congratulations on hitting the big 100k!
    3
  60. You have tapped into a classic boondoggle here. When I was with Cisco around 2000, the big "new wave" was about all optical switching, which was supposed to be faster than converting optical to electronic and back again, often with DMMs (Digital Micro Mirrors). Howed that work out? The startup world was littered with smoking holes of failed companies. I think the bottom line is we know a lot about devices that operate on electrical signals, but not so much about devices that work on pure light. As in everyone knows what an electrical nand gate is, and optical nand gates are possible with optical signals, but good luck getting that to work, be integrated at high densities, and be efficient. Lets start with the basics. You can route signals easily in electronics, and the 10+ layers of interconnect on current ICs talk to this. What's a light conductor on an IC? Well, air, which should be free, but is far from it. You would have to couple in and out of the IC at many points, which is expensive in terms of real estate. You could conduct with glass, and that is a whole 'nuther level. I'm not saying never. I'm just saying that with any breathless new wave of technology you have to look at history and see if that wave has not broken previously, or like every 5 years or so (cough.... AI).
    3
  61. Thanks for the overview. I have a high end threadripper CPU, nice to know how it got manufactured. First class video as usual.
    3
  62. Another note: one of the drivers of the first transistorized computers was the realization that flip flops build with transistors could be built that were far more tolerant of variations in transistor manufactured parameters than analog applications such as radios. When attempting to make transistors with stable parameters for consumer applications, the method used was to mass manufacture them, test their parameters, and live with a high rejection rate. As a result, the computer makers were able to come in and buy up lots of "defective" transistors for use in computers that couldn't be used for consumer electronics.
    3
  63. I'll bet somehow the reactor also managed to leak oil...
    3
  64. ....and that was only the beginning. Now electron microscopes can be used instead of mask lithography, and allow the creation of ICs beyond the current reach of technology, giving researches the ability to see what lies at the next shrink level. There is a variation of the SEM that allows reading the voltage on a trace in a live IC with the passivation removed, giving a scope style trace of what is going on there. This is hooked up to the layout database allowing the user to select a wire in the design to display. And FIB or focused ion beams came from the same science, allowing companies to cut out and add material to ICs... literally editing an existing design. The possibilities are endless.
    3
  65.  @fleurdewin7958  Capacity means cost, and write burnthough is overrated. SSDs already meet the 10 year computer tech standard (anything that lasts for 10 years is ok, since it will be obsolete in 10 years). HDDs are a bad long term bet, and by the way HDDs don't work forever either.
    3
  66. Sure, what is the clock speed of the human brain? Probably less than a 1khz. Yet all of the computers in the world cannot equal one. Parallel processing in 3d solids baby! Oh, yea, and all that for the cost of a beer or so...
    3
  67. I used to be fairly blah on Asus Mobos, but the ROG I recently upgraded to is outstanding. Also, they answer the %*#$ phone, something uncommon with the Asiacos. Thanks for the vid. Your presentation is getting smoother and smoother.
    3
  68. Not sure where you were going with the "bigger chips have mask problems" the number of chips on a wafer is 100's to up to a 1000 or more. Bigger chips simply cut down the total number of chips on a wafer. The constraint on chip size is because of the net yield of the wafer in terms of number of working chips, which is a function of the defect density of the process. This in turn is the number of defects in the wafer, as spread across the wafer in essentially random locations (when I worked in semi we thought of this as like taking a shotgun to the wafer, and the shotgun pellets spread across the wafer). The bigger the chip, the more likely a chip will have one of the defects contained within it. It stops long before reaching one chip per wafer, because you reach a point where the likelyhood of any one chip escaping having a defect reaches zero.
    3
  69. We do it here in San Jose! come get your nice glass of recycled sewage! No takers?
    3
  70.  @SianaGearz  Microsoft method for dealing with limits: Ow! I cut myself on this broken glass bottle. I will throw it down the path a couple yards. Later.... Ow! I cut myself on this bottle. I will throw it....
    3
  71. ARM as tried to go into the desktop market before. I attended an ARM fest about 5 years ago when they were pushing the 64 bit ARMs for desktop and high graphics.
    3
  72.  @aymericdekerdanet9318  ARM tried it before. MIPs tried it before. Intel tried it before (Itanium). I think all this proves is people have short memories. Would I like a true desktop alternative? Oh yes. The real issue is that even if you give a IBM-PC motherboard based on a different CPU, but with the same peripherals and PCI-E slots, your driver issues and application issues are going to be terrific. Linux can handle it (easily) but then Linux is like %1 of the desktop market no? When you win over the gaming rig boys, we'll talk. I am not a gaming rig guy, but I follow them. I have to. If my builds don't burn holes in the rugs, they don't interest me.
    3
  73. Dang we are headed for Picometers? Wasn't that supposed to be impossible?
    3
  74.  @allentchang  So I have not been part of direct IC design for a lot of years. Last time was at Segate, but we didn't do the layouts or see the layouts (sad). I imagine that analog is still schematics. I think what happens now is that the Verilog tools deliver mostly automated layouts that use "fixer" layout people to correct automation problems and for special layouts, again, including analog. A company like Intel or AMD might typically use a mix of both, for example if the design calls for say a cache, that's going to be mostly laid out by hand for a single cell of it, then replicated many times. The author of this video series is not quite correct that everything is standard cell. Even in the old days, when we used cells they were usually from a library of custom layouts made for that particular design. I guarantee you the latest x86 processor does not use standard cells. The other issue is drive lines. One thing an IC designer does a lot is change the drive strength within the chip to match drive to length of line so that it does not drive it too little or too much, and matches the rise time requirements.
    2
  75. Test strips are an example off microfluidics? That's a stretch.
    2
  76.  @ttb1513  The last company I work at silicon design for, Seagate, was emblematic of the basic problems with ASICs. It was 1995 and they were new at ASIC design in house. And they weren't very good at it. It was random logic design before Verilog. An example of this was our timing verification. We had timing chains that were too long for the cycle time, and thus they simply alllowed for the fact that the signal would be sampled in the next clock cycle. Now if you were an ASIC designer back then, what I just said would have made you reach for the tums, if not a 911 call for cardiac arrest. Its an open invitation to metastability. And indeed, our AT&T fab guys were screaming at us to stop that. I got put in charge of hardware simulation for the design, and I have detailed this fiasco in these threads before, so won't go over it again. The bottom line was that ASIC process vendors were loosing trust in their customers to perform verification. The answer was that they included test chains in the designs that would automatically verify the designs at the silicon level. It mean that the silicon manufactured design would be verified, that is, defects on the chip would be verified regardless of what the design did. My boss, who with the freedom of time I can now certify was an idiot, was ecstatic over this new service. It was a gonna fixa all o' de problems don't ya know? I pointed out to him, pointlessly I might add, that our design could be total cow shit and still pass these tests with flying colors. It was like talking to a wall. In any case, the entire industry went that way. Designs are easier to verify now that the vast majority of designs are in Verilog. I moved on to software only, but I can happily say that there are some stunning software verification suites out there, and I am currently working on one, so here we are.
    2
  77.  @FlashMustache  Yea, but they drive on the wrong side of the road, too.
    2
  78. In China its "sheee links" :-)
    2
  79. Yes, I keep running into SGI vets in the industry. SGI died as a company, but lives on as a tech incubator like Fairchild from the early days. We occupied some of SGI's real estate in their final days, along Shoreline blvd. Eventually SGI occupied one of their own buildings closest to the bay. SGIs footprint from their heydays was large and now forms the basis of Google's headquarters. The final days of SGI remind me of an army making their last stand with their backs to the sea. From the roots of old trees grow the new ones. Google's campus is now like SGI on steroids. They not only bought up the old SGI buildings, but seem to be on a tear to buy up every last parcel of real estate on the shoreline, surrounding the amphitheatre like a giant octopus. You can still see the remnants of SGI in the giant sculptures they commissioned there. Its worth a drive around the shoreline area if you ever get to Mountain View.
    2
  80.  @johnhorner5711  When the Cypress 22v10 EE version came out, I built a programmer for it. Cypress was one of the few companies back then willing to disclose programming details. It didn't work! I could program it once, but then it stopped working. Spent a lot of time on the Cypress apps line trying to work it out, never did (maybe it was you I talked to!). I loved that device, I built it into a lot of things. After that I did chip design at Zilog and didn't get back to programmable logic for 10 years, which by that time was dominated by Verilog, and then got back into programmables with Xilinx.
    2
  81. PS tie the comments about water in Arizona to the record low in lake mead!
    2
  82. Sooooo close....100k.
    2
  83. This was a case where everyone won. Now do Huawei. I am getting tired of my laptop maker being bashed.
    2
  84. This will look wildly different as a the PLA marches up from the beaches. And that is certainly part of the plan.
    2
  85. You don't want too much of a paper trail after the Chinese invade... :-)
    2
  86. Went nuts over nuts eh? Sorry, I just had to. A guy got assaulted over a salted nut. Please, I can't stop.
    2
  87. TSMCometry?
    2
  88. Before China, India passed Hollywood, as well. Its just population numbers.
    2
  89. Concurring with the other poster, I would say China is a bigger threat than earthquakes. Earthquakes stop. China would not.
    2
  90. MCMs are old tech. NASA and the military used them to put into spacecraft because they don't like PCBs, which fail when heated and produce bigger circuits. We did mixed MCMs/hybrids (ICs, passives, transistors) at Hughes for satellites in the 1980's and before.
    2
  91. I feel the need to mention here, that I consider (along with others i'm sure) that the proper term is "stored program computer". Von Neumann had nothing to do with the invention of the stored program computer design, which was due to Ekert and Mauchly, the designers of ENIAC. Von Neumann simply took notes on the design and published it under his own name, which led to a misinterpretation that he (ahem) declined to correct.
    2
  92.  @housemana  You have not listed any FACTS about what I have or haven't done wrong, this is just a general attack response. Pretty rude at that. So get lost.
    2
  93.  @paranoidpanzerpenguin5262  mmmmm Sun, now Oracle, open sourced their SPARC cpus. For the second, you are muddying the issue. AMD/ATI released full details on their cards and their drivers, which was very important to the early days of Linux. Things have not changed much since then. AMD is suing someone for copying their chips? Stop the presses! Comon man.
    2
  94.  @Stef3m  1. Its the difference between no source and source. 2. I am just telling you what the Linux kernel group reported. 3. I suspect you are arguing to press an agenda here, thus this is my last answer in this thread.
    2
  95. I'm guessing TSMC would be the only one that could qualify for a bulk discount from ASML.
    2
  96. And AMD is not happy about Intel buying its second supplier.
    2
  97. I don't know. Most of the war and political turbulence of Asia passed the Thais by. Look at their neighbors.
    2
  98. It kinda happens already. AMD and Intel engineers can practically throw forks at each other during lunch...
    2
  99.  @AlexanderSylchuk  Oh, you are begging for my favorite story. I interviewed with a company that made precision flow valves. These were mechanical nightmares of high precision that accurately measured things like gas flow in chemical processes. This is like half the chemical industry (did you know a lot of chemical processes use natural gas as their feed stock?). Anyways, what has that got to do with this poor programmer? Well, like most industries they were computerizing. They had a new product that used a "bang bang" valve run by a microprocessor. A bang bang valve is a short piston that is driven by a solenoid that when not energized, is retracted by a spring and opens a intake port and lets a small amount of gas into a chamber. then the solenoid energizes, pushes the piston up and the gas out another port. Each time the solenoid activates, a small amount of gas is moved along. Hence the "bang bang" part. If you want to find one in your house, look at your refrigerator. Its how the Freon compressor in it works. Ok, well, that amount of gas is not very accurately measured no matter how carefully you machine the mechanism. But, it turns out to be "self accurate", that is, whatever the amount of gas IS that is moved, it is always the same. The company, which had got quite rich selling their precision valves, figured they could produce a much cheaper unit that used the bang bang valve. So they ginned it up, put a compensation table in it so the microprocessor could convert gas flows to bang bang counts, and voila! ici la produit! It worked. Time to present it to the CEO! The CEO asks the engineers "just how accurate is it?" Engineer says: well... actually it is more accurate than our precision valves. And for far cheaper. The story as told me didn't include just how many drinks the CEO needed that night. So the CEO, realizing that he had seen the future, immediately set into motion a plan to obsolete their old, expensive units and make the newer, more accurate and cheaper computerized gas flow valves. Ha ha, just kidding. He told the engineers to program the damm thing to be less accurate so that it wouldn't touch their existing business. Now they didn't hire me. Actually long story, they gave me a personality test that started with something like "did you love your mother", I told them exactly where, in what direction, and how much force they could use to put their test and walked out. I didn't follow up on what happened, mainly because I find gas flow mechanics to be slightly less interesting than processing tax returns. But I think if I went back there, I would have found a smoking hole where the company used to be. And that is the (very much overly long) answer to your well meaning response.
    2
  100.  @brodriguez11000  So is AI.
    2
  101. Amazing. It brings to mind that if there were a clear profit to be made on fusion reactors, there would probably be a portable plug-in model by now (Mr. Fusion). Money crosses all kinds of bridges.
    2
  102. even mechanical machines suffer damage.When the working babbage difference engine replica was shipped to california, the shippers also dropped it, resulting in many parts of it no longer functioning.
    2
  103. They have a new technology, you find a river, go upriver until you find a valley, which rivers tend to make, then place a dam there and let the river water fill it. Cheap and easy. Ok, maybe not that new...
    2
  104.  @bighands69  There was an article some years ago on it in the Wall Street Journal during the war with Serbia (yes, that old). The issue was the fired rounds, on hitting the target, sprayed outwards and even into the air.
    2
  105. My cut rate car from Japan was a Miata. My father had a Porsche 911, but I didn't have that kind of money. The Miata was my Porsche. I did that again with a Tesla M3.
    2
  106. The old near line storage was tape. I went from Seagate to working on tape drives, because as big as HDDs were getting, there was a need for more storage to archive data. But tape was continually in a race for capacity, with disk drives nipping at their heels. Eventually, arrays of cheap disk drives simply turned into the long term storage solution as well, so they wiped out the market for tape and optical disks as well. I recall the words my old boss at Micropolis said to me when, as a brash 19 year old, I read about optical disks and said to him "optical disks are going to replace our magnetic disks". He told me then that magnetic data could easily surpass optical disks since they were limited by the wavelength of light. That was the late 1970s, and it came true. Will disk drives come to a complete halt? It seems unlikely. HDDs have a use going on into the future for large, cheap storage.
    2
  107. Kodachrome They give us those nice bright colors Give us the greens of summers Makes you think all the world's a sunny day, oh yeah - Paul Simon
    1
  108. Feel free to refer to it as a "stored program computer". I don't believe in giving credit where none is due, and you previously issued a video acknowledging that Von Neumann did not, in fact, come up with the stored program concept.
    1
  109. So you are saying the idea sank? Sorry. Why not treat the iceberg water as... well, water. Oil tankers are some of the biggest ships on the seas. One could be converted to store water instead, then you go to the icebergs and lay two pipes to the middle of one, the first that pumps heated water to it, and the second that intakes the water, both the heated water and the melt water. You'd be creating a meltwater lake in the middle of large iceberg. You'd give it up after the iceberg went unstable. Basically it would be converting oil energy to water, much like desalination, but presumably cheaper, since the source was fresh water to begin with. The result would be a taker of fresh water, that would then sail home.
    1
  110. Gonna cross the big half million I see. I've been watching since less than 100k. Congratulations.
    1
  111. Congratulations on crossing 200k!
    1
  112. Yea, you have to go microwave before skin effect takes over. Like gigahertz.
    1
  113. What is the frequency of the sun, anyways?
    1
  114. How do you tell an artificial diamond? Its too perfect. That means the diamond mining industry is on its way out.
    1
  115. Intel did good, they sold us second rate architectures for decades. But with the lapse of "king PC" the X86 arch is slowly falling. Intel had their chance to get into the ARM business and they punted it. They could take over RISC-V by simply showing up, but they won't do that either. They have literally tied themselves to the X86, and they will go down with it.
    1
  116. And thus I offend the disciples of the religion of AI. And thus it is -- As the video says, the Google software is light on details. Much of AI, with some digging, is found to be conventional algorithms dressed up as AI. I have had a 50+ year career, and over and over again we have faced these "AI is around the corner" waves. Is it different this time? Certainly. The pay for AI experts has gone from nothing to a very high level. Formerly it was only academics. I'll apply "Franco's law" here: the "breakthroughs in AI could have been done by more conventional means, and if you look closely enough, probably are. Flame on!
    1
  117. "A/C systems are more economical than D/C".... welllll in old tech that was true. Transformers were easy to make, power came out of the generator in A/C, and the most common use of the power in the early 1900s was motors that ran on A/C. Now fast forward to today. Your power plug shrank. What happened? Well, copper is expensive, so anything that shrinks that is cool. Plus, power is lost in the transformer, which was what was in that heavy brick that used to power your laptop. In addition, generally you use D/C power in most of your house now. All the electronics, those LED lightbulbs, etc. Yep, all D/C. Electronics came to the rescue. Turns out if you use high power, high (er) frequency (than A/C at 60/50 hz) you can do the same power conversion with way way (way) less copper or even no copper at all. Plus, it is way easier to perform high power conversion now, even at high voltages. Thus things are changing, rapidly. There is a good chance that many or even most lighting systems will go DC as distributed on DC feeder lines. This is already true in some large offices and industrial concerns. This is because a lot of the power used in LED lighting is used in the conversion from A/C to D/C. Want to prove this to yourself? Go find a screw in LED lamp in your house. Feel the glass where the light comes out. Now feel the base (don't touch the metal). The base is hotter isn't it? That is where the A/C to D/C converter is.
    1
  118. Besides the white LED, blue LEDs enabled Bluray and the high definition revolution. It was a big, big, big (big) deal.
    1
  119. "inspired". Lovin' it.
    1
  120. It wasn't a law, it was an observation.
    1
  121.  @michaellau5329  Didn't follow that at all. Two chips in a package with X and Y heat generation somehow is not X+Y? New math?
    1
  122. Very well done, thanks.
    1
  123. Emperor Xi knows all about power. His first business venture was selling dried cow dung for cooking fuel...
    1
  124. 120,000 wafers at what, about 300-500 yield per wafer is about 60 million chips. These guys print money.
    1
  125. A tie in with IEEE Spectrum and the final approach to 1m subs. I watched your channel grow You truly made it. Congratulations.
    1
  126. I like your channel, its good information. However, the presentation comes off as a bit unprofessional. I think your working from a script vs. ad-libbing (as it appears you do) would be a big improvement.
    1
  127. Gee, AMD and Intel fighting, imagine that.
    1
  128.  @AlfaPro1337  I think you might find that AMD believes that to be the other way around.
    1
  129. "DRAM faces scaling limits" I would think this could mean that static memory would lap dynamic memory, since static memory cells only need transistors, not capacitors. This would be a good thing, since statics are faster and don't need refresh cycles. This already happened in random logic applications like CPUs, where they moved to static logic and dumped dynamic logic like domino logic, which had minimum clocking requirements. I like to think of IC tech like trains racing on multiple tracks. For example CMOS logic took a back seat to NMOS tech for many years, until NMOS hit the speed/power wall, then suddenly everyone switched to CMOS.
    1
  130. The drip drip drip of the puns here turned into a flood. We are drowning in them.
    1
  131. At the risk of being politically incorrect, the west's ability to make the rocks come alive and attack enemies in the middle east is like something right out of the Koran.
    1
  132. don't buy from people stupider than you...
    1
  133. God made millimeter waves to allow cellphone direct to satellite connections.
    1
  134. I never bought a computer. I made all of them. But I had to repair friends and relations Compaq and other computers. In those days, the makers went out of their way to make proprietary cases, motherboards and DRAMs. It was sad to see, and it, it my opinion, contained the seeds of their demise and the rise of the Asian computer dominance. I could get an off the shelf case, motherboard and other components and make a PC cheaper than Compaq and other makers pre-built machines. IBM gave us standard motherboard sizes. Companies like Compaq, despite creating the IBM clone market, choose to step on that.
    1
  135. The utopian idea is that everything will be free, and nobody would hoard because they knew they could always get more. But people are not stupid, and know full well that supplies will end soon, and so hoard as much as they can. Its obvious with food, but in the Soviet Union, it rapidly went to every possible item because people realized they could trade for other things. Thus when anything was offered "for free" the people knew to grab it. The result was obvious: go back to money so that the supply was self limiting. The genius of the soviet system after the reintroduction of money is that the state was the only approved buyer of any goods. Thus they were able to set all prices and control the economy.
    1
  136. The story goes that the chairman of Sony saw little drink umbrellas that came from Japan, and Japanese goods were considered "junk", and went home to fix that. We are making the same mistake in the USA with respect to China. The days of China producing low end copies are over.
    1
  137. Gonna cross the big half million I see. I've been watching since less than 100k. Congratulations.
    1
  138. Intel likes to take over cities. Look at Hillsboro and Chandler.
    1
  139. Lithography, literally "stone printing", which is rather funny if you think about it.
    1
  140. Microsoft made 3d APIs about as well as Hitler made elderly care centers.
    1
  141. I literally went to war with people spreading misinformation on a subject for which I am one of the few remaining experts. I succeeded, but I would have had to continue the fight... forever. The last straw was I noticed that the same misinformation was being spread on the same pages in other languages. It just isn't worth it. Take all Whackipedia information with a grain of salt.
    1
  142. knock-on ef·fect noun a secondary, indirect, or cumulative effect. "a decline in butterflies would have a knock-on effect on other British species"
    1
  143.  @uploadJ  Sure. The first simulator I did was a standard switch level simulator. These are good, but don't consider many physical attributes such as the resistance of the switch other than being able to produce "weak" states, so that transistors with greater power can take over a node. The RC simulator gives a simulation between switch level and SPICE, and gives actual switching curves, a real picture of strong vs. weak switches, and yet gives reasonable speed.
    1
  144.  @uploadJ  Yea, the view of an IC as a series of capacitors discharging through resistors by switches is obviously limited. However, its nice to see curves coming out of such a simple simulation. I had hopes for implementing the algorithm in hardware, and still do. Like getting a large net that executes reasonably fast, but is actually analog based. The impetus for all that was that our regular switch based simulator at Zilog could not handle resets, so that was done by node sets. I was interested in seeing if I could make a simulator that didn't need that kludge, mainly because I had just solved a serious reset bug in one of our chips! I was able to do that, mainly by introducing weak states into the simulation (this was the digital only simulator). I think that is standard now. They just killed my old chip, the Z80. Ah well. Crappy instruction set really. I made a compiler for it, but I can't believe I did that now.
    1
  145. You are right. The probe was a real accomplishment. Do you think its possible we will see a mercury lander one day?
    1
  146. Going to be a party when you cross 100k!
    1
  147. Nice overview, well researched. Hoping you get to 200k soon.
    1
  148. "Stanfurd"??? Sorry I just had to.
    1
  149. Small nit: Jack kilby and Robert Noyce were determined to be co-inventors of the IC patent. Although it is true that Kilby was first to announce, the lack of interconnect was considered crucial to the invention.
    1
  150. The key would be to control your tooling down to the mask level (the layout). IE., separating the pure software/data phase from the physical implementation phase. Masks can be compared to the generated layout, and even a fabbed chip can be compared to its layout without reverse engineering it. It is of course true that many of the EDA phases get "thrown over the wall", that is, a Verilog design can be send to a vendor to be laid out. However, the security there is up to the client. If the chip needs high security, the client can bring the tooling in-house to be more secure.
    1
  151. Two short factoids: 1. IBM thought X ray lithography was the way to go and failed (ultrashort wavelengths). They may have just been to early. 2. How do they even know the next generation chips are even possible? Test chips can be created (slowly) by using scanning electron beams. Electrons have a really short wavelength. There is no mask used, each step on the chip is done by scanning it. Hence it is slow, but works for lab use.
    1
  152. Seems like this all ends with silicon cubes instead of silicon chips. There is already a lot of 3dness in tons of interconnect layers and 3d transistors. When they start bonding different circuit layers together then we start emulating the human brain.
    1
  153. Your channel is excellent. I would just ignore the critics. Arguing with people on the internet just justifies them.
    1
  154.  @makisekurisu4674  No.
    1
  155. Sure, but the problem is that anyone can get into the SSD business, and your SDD chip supplier is your god.
    1
  156. HDDs are an inferior form of storage compared to SSDs (how nice that the world can be described with TLAs). Thus HDDs are going to live or die based on being a form of backup. We saw this same dynamic happen before. Tape drives and optical drives, as a backup media, died out because arrays of HDDs were cheaper. HDDs certainly have a price advantage over SSDs, but that price advantage is eroding as SSDs become cheaper in relative terms. A quick dive in Amazon shows the price advantage at about 5 to 1, HDDs over SSDs, in the same format (SATA). M.2, the rising defacto standard for SSDs (M.2 modules have a significant speed advantage over SATA, which never accounted for the great difference in speed between HDDs and SSDs), have a price premium, but that is eroding rapidly, for the simple reason that there is no fundamental reason for such an advantage of M.2 over SATA. On the contrary, M.2 has less material than SATA and so holds the long term price advantage. SATA drives need a metal case. The upshot is that HDDs hold a clear advantage, and reason for existence, at the 10 to 1 price level. At 5 to 1 we see that the sales curve for HDDs is trending down. When the price advantage falls below 2 to 1, its time to get out of the pool. The HDD industry will die.
    1
  157.  @Chad.Commenter  Cite.
    1
  158.  @Intelwinsbigly  Yea we (OCZ) did that and sold drives like that. They didn't sell. Good for repetitive storage accesses, terrible for true random access. I worked on the second generation SSD drives for enterprise. We implemented PCIe connected drives, long before NVMe was a twinkle in Intel's eye. Just the fact that the SATA bottle neck (6g) was REMOVED caused our SSDs to blow every hard drive (enterprise level!) clean out of the water on all performance scores. Those 4 lane PCIe M.2 drives will clean any hard drive's clock, and that is half the bandwidth we used. Can you get TB for cheap with hard drives? Certainly. I can sell you several 6 packs of beer for the price of a good bottle of wine. Would I use hard drives for backup? Heck yea, and I do. A good NAS such as I have at home for cheap is very cool . However, again, I use M.2s (Samsung) with Samsung SATA 2.5's in a 4 x quickchanger/hotplug chassis to boot alternative OSes.
    1
  159. I have believed for a long time that RFID could solve recycling problems. An RFID tag molded into plastic containers could carry information about the chemical contents of the plastic. The biggest problem with plastics recycling now is that, even though plastics can be easily remelted and recycled, mixing different plastics makes that impossible. So plastics are ground up and mixed together and only go to degraded uses like plastic bags and plastic lumber. RFIDs could allow collection by type, and even return to the maker, as whole bottles or other packages, instead of being ground up. Additionally, the RFID would identify the product on retail shelves and boxes.
    1
  160. All it takes to beat them is an Arm... and a leg.... :0
    1
  161. Not to excuse the silicon manufacturers, but Silicon Valley, AKA San Jose, has had a number of contaminations over time. The first was mercury, which they used to mine in the hills above San Jose. It was used during the gold rush to process gold bearing ore. The hills are still quite contaminated with the stuff, and fishing is prohibited, leading to an odd abundance of fish there. Nothing saves your life like being a toxic fish. More recently, it was discovered that the company making rockets, UTC, in the hills, had left large waste ponds of the makings of rocket fuel, and those waste ponds were leaking into the ground water for many years. This leads to thyroid disease. I can't really complain. Starting in electronics at the tender age of 16, I have inhaled tons of triclor, solder fumes and rubbed up against a lot of lead. We used to carry it around in bars, and melt it in pots and soldering machines. But as you can see, it has had no effect on me. But then, as you can see, it has had no effect on me...
    1
  162. That was an awesome overview of the ASML system, thank you. On to 200k.
    1
  163. I suspect in the long run people will view mainframes and even minis the way we view engines. History books show stationary engines where the work was transmitted via belts, but no existing examples exist, and its just a series of odd photographs with no similarity to the modern world.
    1
  164.  @happygimp0  Not much danger of that. In the PC world, everything is open. Apple tried that with their hardware (maybe they will again eh), but its like waving a red cape in front of a bull, everyone rises to the challenge to break it. Linux can run on a MAC.
    1