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

  1. Small correction: Sodium Hydroxide is not an acid. Quite the contrary, it's a very strong base.
    584
  2. I hope you do a multipart series on AMDs history. It is truly fascinating. There are as many bold moves as there are instances of jeopardizing the company out of an insistence on treating employees like team members rather than consumables. That AMD survived from 2008 to 2014 is surreal in it's own right, and the comeback after that is even wilder.
    532
  3. YKKs classic all-metal "scoop" style zippers are the only ones that have never failed before the garment or footwear for me. Sure, most modern plastic zippers work great for low utilization stuff, but for jeans, jackets and boots I actively look for YKK. Fun fact: The Danish army insist on them.
    455
  4. As a programmer your videos about the never ending toils and tribunals of the lithography world reminds me not to wasteful with system resources. It's clear that at some point we're going to hit a wall called "physics don't care about your computing needs", so while I've seen stupendous performance gains from my Commodore 64 to my Ryzen 7950X3D that party is tapering off.
    332
  5. From the ultra pure sand used to make silicon wafers to the very-specific-size-grain sand used for concrete... thrown in with a Star Wars sand reference. Yup, Jon, you've become a Sand Nerd :P
    331
  6. Considering how much of an EV is battery cost then 10-30% cost reduction on the batteries could absolutely be worthwhile. Even if the energy density is a little less.
    311
  7.  @joefioti5698  No. Already back in the GeForce 3 days governments were freaking out because an off the self product was fast enough to do the job that had required application specific chips up until then. Missiles and drones can probably do perfectly fine with 90nm class shit. And I'd be surprised if there is any non-experimental fighter jets with anything better than 28nm chips in them.
    262
  8. As someone who's been following chip lithography for close to 3 decades: The next step is always "a major crisis". This started already in the early to mid 90s. And I think "we" learned from EUV to not put it off. That EUV was delayed by a decade just got the fabs to become insanely good at multi patterning. TSMCs N7 was an absolute bonkers smash-hit. It's only downside was that the hundreds of patterns made lead times enormously long. N6 uses DUV for almost everything, but has two critical layers moved to EUV. This, besides the small logic shrink, made it absurdly cost effective, and is the reason why AMD likes to do everything possible on N6. EUV vs High NA-EUV will probably be the same: They will figure out how to make sure they only use it for one or two layers in the beginning. If that's the action plan, then maybe a polarizing filter is financially viable.
    252
  9. I came for semi conductor tech but stayed for the well researched and delivered multitude of topics.... So the silicon bule analogy was welcome.
    246
  10. Prescott, also known as Preshot, was AMD's big opening. They'd realized that pushing for ever higher clockspeeds was a fool errands, and designed a much smarter pipeline that could execute many more instructions per clock cycle. It was the Athlon XP and it should have dominated the world. Why it didn't is a story best told by Jim from Adored TV.
    221
  11. The obvious flaw with comparing TSMC's RnD spending with Intel's is that Intel also designs a plethora of chips and software solutions. TMCS only has lithography, packaging and PDK development. Intel also has these on top of everything else. Alphabet and Meta seems to spend money on a lot of useless things "just in case it sticks".
    208
  12. Xilinx was known for their incredibly tight relationship with TSMC. AMD, having owned their own fabs previously, was already intimate with working directly with the lithography department, so they quickly ascended the same stairs with TSMC. Today AMD owns Xilinx, and it is speculated that their ultimate goal is to beat Apple to the leading edge and become the premier partner of TSMC. If that happens I think both Intel and Nvidia will be in long term trouble...
    199
  13. The semiconductor space offers little room for failure unless you're already one of the very biggest who can stomach a few years of faltering revenue.
    192
  14. This was largely correct, but I do remember three distinct times where ATI/AMD had the undisputed performance crown. The ATI 9700 PRO was the first of these that I owned.
    154
  15. Can't wait to see the "5 years from now" video in this saga. It's not that I learn anything new, I just like Jon's take on it.
    134
  16. TSMC N5 and N4 are not even fully EUV. Almost only the fins in the FinFETs are made with EUV. All the TSVs and most the routing layers are made with 193i. And the reason is simple: The wafer throughput is much much higher on 193i. So if a chip has 80 layers and only 5 of them are EUV and the rest are 193i you need, say, 1 EUV machine and 8 193i machines for the same throughput as 25 EUV machines would give you. Considering just how brutally more expensive the EUV machines are this is the difference of having a break-even after 3 years at full capacity or 30 years at full capacity.
    123
  17. I'm happy that the Dresden project is taking the "known good" route. The whole Chips Act always looked to me as a "If Intel keeps eating their crayons and sniffing their glue"-emergency-exit kind of mechanism.
    94
  18. Oh yeah! Keep them coming John! I want to know everything about this industry, and you're a top dollar educator on the subject! :D
    87
  19. It would be interesting if PDKs were made to follow a certain standard. Like XML or what have you. That way the "big professional stuff" and the open source stuff would be able to work with the leading edge nodes provided you're in the good graces of the leading foundries.
    83
  20. Of CAUSE we want to hear about DSA!
    83
  21. Bob Swan insisted that 10nm should both be an aggressive shrink of 2.3x instead of the usual cadence of 1.7x AND it should be cheaper to produce (by omitting EUV) than 14nm. By the time 10nm actually started ramping TSMC was well ahead on their N5 node that was denser and cheaper using EUV. They did better in two steps than Intel did in one, and they did it faster. Intel's next generation Arrow Lake will "initially" be made on TSMC N3 "until Intel can ramp production" of their own version. This is what happens when you sent a bean counter to do an engineers job.....
    78
  22.  @christopherkelley2061  If you need a base stronger than a fully saturated NaOH solution at 95C you're doing some seriously questionable things! :P
    74
  23. Ugh... I thought I knew the GloFo <-> AMD split pretty well... but I learned a good handful of things today! :D
    66
  24. This is exactly what trade sanctions and embargoes do, when applied to countries that aren't rotten poor: It breeds domestic innovation out of necessity. Telling North Korea they can't have this or that is not going to make a big difference, but trying to boss India or China round is an exercise in futility..... Especially when higher education is now international business and open to everyone.
    65
  25. Ah Jon! Another episode that's not just a home run, but knocks it clean out of the park! The realization that an optical gate can't be smaller than the wavelength of the light it operates on puts everything into perspective. I'm hoping for an episode on the research on room temperature super conductors and their implications for chip making!
    60
  26. That's pretty strange... Hold the guy at gun-point, get the chop, stamp the deal, and it's official. Be unable to get the chop: Nope. Nada. Nothing. Even though you're the group who mandated that he got the chop in the first place you can't do shit without... the chop!
    57
  27. With the work you do I really hope TSMC will give you a tour of a legacy node fab. Hell is going to freeze over before they'll let you into a leading edge fab, let alone with a camera, but since the general process, from a high level perspective, is still the same, it would still be insanely interesting to see the whole thing, and I don't know anyone I'd rather have narrate it than you.
    56
  28. The sound is a bit harsh in this episode... But the content is top dollar as usual. You're like the EUV version of tech journalism! :P
    50
  29. If graphene can reach hundreds of GHz or even THz then it doesn't matter if power consumption goes up by 50x. The reason the transistor count is so absurdly high in today's CPUs and GPUs is because of the lack of switching speed. You roughly speaking need to increase transistor count by 70% to increase IPC (instructions per clock) by 15%. So a Pentium sized chip at 1THz would run rings around a modern core while still using significantly less power. Of cause, if the transistors can also be made as small (or thereabout) as today's leading silicon nodes, then the area of a Pentium would shrink to a pin-prick.
    49
  30. Your impression of "Internet Experts" was spot on! :D
    47
  31. On the other hand... those are exactly the kind of word-barfs you hear from people who do not get promoted due to personality issues. I happen to be "one of those people". But rather than bitch about it I've simply accepted that only certain kinds of people should be on my team, and that I should never be put in a position where I have to be "the considerate fatherly type" to make things go smoothly. I don't do smooth. I do effective. People on my team must naturally prefer extreme directness bordering on the outright blunt. "Yes" means "Yes", "No" means "No" and "This approach sucks" does NOT mean "You're a failure as a human being and I hate you now so please go and hang yourself". It means "This approach sucks. We must find another. Any ideas?". The upside is that my team is never late. We get frequent bonuses and extra days off, and our personel turnover is practically zero. People either ask to be transfered within the trial period of three months, or they stick around forever. And management has gotten increasingly good at figuring out who's a good candidate for my team. In return for never advancing I've gotten my own little kingdom. Upper management stays out of my hair, and I stay out of theirs. It's a win-win for everyone when not-at-all-people-persons acknowledge their shortcomings and just collect the other blunt/weird/awkward rejects and make a team out of them. And it can come as no surprise to anyone that in the tech fields there are quite a few who somehow traded social skills for technical prowess.
    46
  32. I have no objections to the correctness of the statement, but I did not expect that particular delivery.
    41
  33. AMD's 3D V-Cache technology was the first product to market using the direct-copper method. It was, apparently, a rather deep collaboration with TSMC that stretched several years. The Zen 2 based chips (Ryzen 3000 series) were equipped with the receiving pads for the V-Cache, but the yields were absolutely atrocious so it was never marketed, but by Zen 3 TSMC and AMD had figured it out. By making the V-Cache on N6 and the chips on N7 they had a match that yielded right, and the rest is history.
    38
  34. We're really pushing the boundaries for what can be done. I wonder when technologies get so advanced that human ability to manage project complexity becomes the limiting factor.
    36
  35. Nice episode John :) I'm totally hooked on your deep dive into semi conducter manufacturing, so please, by all means, go on and explore every little nook and cranny. Methods, techniques, history of development, supply chain, global impact, etc, etc. It's all ace, as you're good at structuring an interesting narrative and delivering with impeccable pacing. 10/10.
    36
  36. "Everything that can be weaponized will be weaponized" -- Russia.
    36
  37. Heeey, mah man? I watched the cultured pearls video and enjoyed it! That said, I'm sorry for Abu's loss. They'll never make that back. But I'm happy that AMD got out. "Real men" or not they have certainly lit a fire under Intel, which was sorely needed.
    35
  38. It got pretty close to a "Hell naw!"....
    34
  39. But therein lies a problem: Something like Global Foundries 14nm process is so damn cheap per wafer today that mucking about with 42nm is pointless. 14nm is about four times as dense as 42nm, so you get four times as many chips per wafer. And when your chips are only a quarter the size the packaging gets cheaper, etc, etc. Micron, Samsung Foundry, SK Hynix, Global Foundries, UMC, TSMC and many many more are already 14nm capable. And the automotive market is already shifting towards 10/12nm for the infotainment systems.. Investing in 42nm today is utterly pointless, as there is already excess production capability in that area.
    34
  40. For those of you interested in learning more about MEMS I can highly recommend this video by New Mind: https://www.youtube.com/watch?v=iPGpoUN29zk It doesn't cover as much of the economic detail, but has some rather fascinating use cases.
    34
  41. They were twice as much in debt as they got for GloFo. There was no other option. They even sold their head quarters and rented them back. At the worst months, just before Zen went into mass production, engineers had to vacuum their own offices and cubicles because there was no money for anything but the most basic of sanitary musts. So no, it wasn't "sort of justified". It was that, or no AMD. It wasn't until 2021 they became debt free, which was what gave them the breathing room to purchase Xilinx, which required them to take up a loan of $2.5B along with converting Xilinx stocks into AMD stocks at favorable rates. They have brought the debt down to $1.7B since.
    33
  42. Standing ovation Sir! And it's a miracle that you didn't break your tongue at the Polish guys name! :P
    33
  43. Except they aren't simple components. They're the most technologically advanced thing humans have ever mass produced. And they're only financially viable because they need to be produced in such bonkers numbers.
    33
  44. The difference being that in Russia some actually believed in communism. The Chinese "communism" was always, at every level of design, doctrine and implementation, about creating an elite that had an excuse for living in luxury while telling the peasants that this is what social justice looks like.
    32
  45. Last I heard TSMC is introducing N3E which is "only" a 60% density shrink as opposed to the standard 70%, but it's already yield ramping right. This shouldn't be a big issue since N5 was already a bigger leap over N7 than normal. 84% in the best cases actually. So from N7 to N3, on average, follows the 70% density shrink every two years.
    30
  46. Crappy stuff from H&M always has the flimsiest plastic SBS zippers, and they sell A LOT of almost single-use clothes.
    29
  47. And this is one of the few times Microsoft actually did the right thing: they were ON it like a whiz. They had caught wind that Intel would partner with HP and their HPUX as the server operating system of choice for Itanium. Microsoft didn't like that one bit, and put all hands on deck to make a good x86_64 compiler and port Windows to the new instruction set extension.
    28
  48. Jesus we've been hanging in there with multipaterning for a long time. TSMC N7 is the last iteration that didn't use EUV. Even N6, which is only a 18% density increase over N7, uses EUV for the critical layers.
    28
  49. Thanks. I've never known much about Korea's political history. This was very interesting.
    26
  50. The 60mm Vulcan can take out a tank. And when it was designed it could take out mostly any tank with only 2-3 hits. But it's primary purpose today is to be "a remote general terrain fuckupifyer". It is especially effective against troops and vehicles under tree canopies. The relatively large spread combined with the stupid high rate of fire makes it ideal for "somewhere around there there is something we don't like - could you make it go away please?" type defending of ground troops. It is, obviously, equally effective at denying area access.
    25