Comments by "Anders Juel Jensen" (@andersjjensen) on "Asianometry" channel.

  1. Laser in itself does not cause debris. Hitting things so hard with lasers that they turn into plasma causes debris. They're basically "boiling" the tin droplets so hard they turn hotter than "white hot". Think "super heated tin steam". But once the temperature drops some it turns back into tin... where ever it ended up after being vaporized.
    6
  2. Yeah, I laughed at the slip-up too :P
    6
  3.  @TheNefastor  That's just North/West European tradition pretty much... I have more paid vacation than ASML employees, and I'm just a lowly IT Tech who makes sure e-mail mail servers remember to check the stamps...
    6
  4. Funny that you pasta a gigantic blurb on a channel that just did a deep dive on Nano Imprint Lithography (and getting the acronym wrong in the frist line). While the technology is certainly interesting and many things can be speculated about it there is still the giant elephant in the room: Can it scale to the speeds needed to compete? Can it obtain per-wafer costs that can compete? Fail in either of those two and the technology will be relegated to countries that can't get an EUV permit... So China, Iran, Saudi Arabia, North Korea, Russia and the other usual suspects. And those countries will in practice be forced to only service their own markets as TSMC, UMC, Samsung and Intel Foundry Service (Global Foundries seems to have rage-quit on the EUV race) continue to make the latest high end chips at competitive prices and volume.
    6
  5.  @Syomiti  Yeah, that's about right. Longer "i" in Jutland and Fyn than in Copenhagen. But the tendency for ck to become gg is spot on :D
    6
  6.  @vulpo  Funnily enough Firefox stores its settings in a .sqlite file. SQLite is a runtime library that can give every application SQL functionality. It also happens to be the worlds most widespread SQL implementation, since it's pretty fast and only takes up 699KB of memory/disk space.
    6
  7.  @foobarf8766  The operative word was "direct-copper method" that Jon described in this episode. Yes, MCM/Chiplets have been used for a long time, but that was not what I was referencing. And no, the first wafer-wafer bond consumer product was the HBM on the Radeon Fury cards. But that was solder-bump bonding, not direct-copper.
    6
  8.  @doujinflip  Still doesn't change the fact that a union who doesn't try to strike a fair balance (like here where different factions within the union where having competitions on who could annoy management the most) between worker rights and worker responsibility will eventually self sabotage.
    6
  9. You're barking up the wrong tree. Stocks are just a given percentage of a company. If the company buys some back, the outstanding ones represent a corresponding larger percentage. This is nothing like the diamond industry which is sitting on literal mountains of diamonds that will never be sold because scarcity is the sole value of diamonds.
    6
  10. I think a lot has already gone into reducing the internal stresses in the boules the wafers a cut from. 300mm wafers are already only 400 micron before any thinning (so only 4-5x thicker than what you got down to).
    6
  11.  @reinerfranke5436  That is flat out wrong. Intel didn't think EUV was mature enough and were slow to adopt. ASML delivered several machines to TSMC before Intel even ordered their first one. Intel wanted their 10nm to be both very dense and very cost effective. That was Intel's first mistake. Then at the beginning of the pandemic Intel cancelled their orders because they were expecting a financial downturn. TSMC was the first to realize that the pandemic would have the opposite effect on the computing industry and simply ordered everything ASML could deliver for the foreseeable future. That was Intel's second mistake. In multi billion dollar industries there are no "lucky ones". There are the ones that makes the right move and there are the ones who makes the wrong move. CEOs don't get paid in excess of $100 million a year to "be lucky". Under Bob Swan's reign Intel managed to lose a perpetual node advancement over their competitors that they held for close to five decades in a row. Now they're half a node behind at the very least. And there is nobody else to blame than themselves. Swan wanted big fat profit margins AND the most advanced tech at the same time, but there is a saying in engineering that goes: "good, fast, cheap... pick two!" and his replacement by Pat Gelsinger (who is as hardcore an engineer as they come) is testament to the fact that the board of investors woke up and realized this before Swan did....
    6
  12. Correction: You're allowed to pronounce acronyms as you please. Though I tend to agree with you, as RAM was invented first, but when the Dynamic variant was later invented a distinction between SRAM and DRAM was needed, by which time the pronunciation of RAM was well established. But the thing is that Jon knows.. he just does it to mess with people :P
    6
  13. True. But also don't believe in magic.
    5
  14. Nobody knows anything for certain about it. There is good indication that the oxidation contamination at the Arizona fab only affected Sapphire Rapids server CPUs, and that the Raptor Lake and Raptor Lake Refresh failure are solely due to the ringbus getting cooked. But it is just that: good indication. Not outright proof.
    5
  15. ASML is mostly owned by it's own biggest costumers. With Intel, TSMC and Samsung being the most invested.
    5
  16. Huh!... I thought sports shoes were completely automated in production by now. I figured that maybe someone had to put the shoe laces in at the most...
    5
  17. These companies should come to Denmark. One of the most geologically stable regions in the world, stable infrastructure at every level and a water supply that cannot be emptied. Here artificial arrogation is all but unheard of because downpour is a given nearly 200 days a year. This also means that our ground water reserves replenish themselves as fast as we can pump it up. We basically just use the ground as a big filter.
    5
  18.  @timluo6120  Taxes are high on labour but there is plenty of skilled labour in every high tech field. Nearly 40% get a university degree here...
    5
  19. You should probably suggest some topic about Turkey, or some industry in Turkey, that he can talk about. He generally don't present "a country" but rather a development of some sort in that country, be it political, entrepreneurial or technological.
    5
  20. It may not be why they left, but recent attempts to bring leading edge node fabrication back to the EU and the US is running aground on many environmental factors.
    5
  21. Stellar work as always Jon!
    5
  22.  @heatengine9283  I left the comment for others to see. Trolls never really tries a comeback when you start throwing well researched facts at them :P
    5
  23.  @mytech6779  All node names are completely bogus at this point. In the beginning they were referencing gate size, but some clever buffins figured out how to increase the transistor density by more than the gate shrinkage. As for the 1.7x nomenclature, I have no idea why that became common. It means that a given amount of transistors that would take up 1.7mm2 on the old node takes up 1mm2 on the new node.
    5
  24.  @davidgunther8428  Heat density is a myth. We have been "hitting the heat density wall" for the last 30 decades. And you didn't read what I wrote: If each transistor outputs 50x more heat, but switches 500x faster, you still have a net reduction in heat output because the design will have 200x fewer transistors and still yield a performance boost.
    5
  25.  @E4tHam  Power does absolutely not scale linearly with frequency. The reason laptop CPUs can reach 70-80% the performance of desktop CPUs at 15-25% the power draw should clue you in on that.
    5
  26. I'm still enjoying my popcorn watching this train wreck.
    5
  27. Stellar episode as always! :D
    5
  28. Personally I'm not into bling at all, but if I were I'd much rather have a grown diamond, since I know it wasn't be mined at gunpoint in miserable conditions in one of Wagner Groups "protected" diamond mines somewhere in Africa.
    5
  29.  @gljames24  It's no use to speak science to people who talk about Adam and Eve.. and even misquotes the lore...
    5
  30.  @Phoenix56801  Intel's node have been expected to get far better for the last 8 years. They've consequently missed their target by multiple years. Their last normal rollout was 14nm. TSMC makes 5 times more money than Intel. Do you really think they can keep up with someone who has that kind of dole AND knows that they literally live and die by their nodes?
    5
  31. I'm watching this from a computer so old that I really wish the parts were from Fab 18 :P
    5
  32. They can. And some of the patents are starting to expire. But they lack the practical knowhow (a patent is practically useless other than telling you where to start looking), so yeah, 10 years sounds about right.
    5
  33.  @Neb_Raska  In terms of financial horsepower India, not your average Indian, can actually throw some weight round. In GDP ranking they're behind Germany and Japan (so 5th overall) but when adjusted to purchasing power parity they are up to 3rd place. The cost of even highly skilled labor in India is very cheap, and since they don't bother with much with environmental protection, carbon taxes and what have you, their production prices are pretty low.
    4
  34. As we approach the limits set by quantum mechanics, it's going to be interesting to see how they'll tackle the placement problem.
    4
  35. Their headphones are world class. And for some strange reason they're the only manufacturer of phones besides google that doesn't completely screw up the OS.
    4
  36. Uh... that was a highly entertaining, but also actually accurate, way of looking at it. Thanks for the chuckle!
    4
  37. As predicted... It's actually not nearly as bad as I thought it would be.
    4
  38.  @anarsosoroo2891  He, humorously, said: "Intel used to dunk on AMD by calling chiplets 'glued together cores' but now Intel recently published a specification for a, supposedly, industry wide approach to a cross-vendor way to mix and match different accelerator chipslets into a coherent package. It 'incidentally' happens to be perfectly compatible with AMDs Infinity Fabric". But @J Bali is also right, as AMD (and a few specific AMD employees) is actually mentioned in the white paper. So AMD was in on it right from the inception. Hence Intel didn't just "patent Ryzen".
    4
  39.  @jansix4287  I want you to break out a calculator and tell me how long time it takes a 100MHz processor to execute 1000 instructions (assuming one instruction per Hz). Ok, I'll give you the answer: 1/100000th of a second. Or 0.01 millisecond. But I wasn't talking about cars. I was referencing OPs remark about using lines of code as a measurement of software complexity. In security sensitive applications like fire alarm systems, collision avoidance systems (like in factories, etc) or similar you do NOT "optimize" the code. Everything is programmed in block/ladder format. This is a very primitive programming paradigm that has one big advantage: each block can be unit tested on its own, and you can only climb the ladder when certain conditions are met. This means you can do much tighter formal verification than normal. Or in layman's terms: you can be reasonably sure that your code does what you think it does. This is worth more than speed. If your code is correct, but not fast enough, you specify a faster processor as a requirement. Done deal.
    4
  40. It's not two weeks ago when he posted a video about Thailand's car manufacturing situation. However, quite a percentage of his viewership probably joined because of his coverage of TSMC, Samsung and ASML amidst the GPU shortage, so it's natural that he still want to cater to that audience. I, for one, will be gone if he exclusively starts covering less-than-high-tech production.
    4
  41. I think the highest grade Wagyu is waaay over the top on the fat. Middle grade tastes fantastic though.
    4
  42.  @glennac  He specifically joked about pronouncing DRAM as "dram" in an episode, so he clearly knows. And both wikipedia and google will happily play a sound clip for you of every word you're unsure about the pronunciation of....
    4
  43. Or UMC, Micron, SK Hynix, SMIC or the dozen others... Mentioning the three that runs leading edge was good enough to get the point across.
    4
  44. Go through his back-catalogue and find the episode on FinFETs. I don't recall the title, but I know it's there.
    4
  45.  @michelbruns  No. 10USD for a monthly plan (of unknown milage). You buy the first two batteries yourself with the scooter. Then you swap them for charged ones. Did you not hear a word the man said?
    4
  46. The TI-85 is as overpriced today as the day I was required to buy it.
    4
  47. I've heard that song since the micrometer days... But until we reach a gate pitch of 5nm we're still good. At 5nm quantum fluctuations start to become noticeable. (Observe that lithography nodes are not named after the gate pitch but rather what marketing thought sounded cool).
    4
  48. Since you can write that on the internet I'm going to postulate that you have no clue what you're talking about.
    4
  49. Sorry, but there are patreon responses from 2 months ago :P
    4
  50.  @JayDee-b5u  I work for a subcontractor of a major defence contractor, and we're located in Denmark, so yes, the pay is good to the point that I'm perfectly happy with an annually increase that matches inflation in exchange for not having to take on more management responsibilities.
    4