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