Comments by "kazedcat" (@kazedcat) on "Asianometry" channel.

  1.  @akatsuki6371  Overtaking DUV is a complicated question because foundries use the two machine alongside each other. DUV machines has higher throughput so EUV are only used on critical features that can only be done with EUV while using DUV for the rest. For example for the metal layers of a 7nm process. EUV is only use for the first and second layer. For 5nm the 3rd and 4th metal layer is also in EUV and smaller nodes will use them on even more layers. But Processor chips can have up to 12 to 14 metal layers. So DUV machines will not get abandoned. As process nodes that requires EUV grows they will still need DUV to complete the fabrication.
    24
  2.  @akatsuki6371  They are economically viable. They are using this monster machines for several years now. The foundries raise their wafer prices but the chipmakers have enough margin to absorb the increase in price. It is very economically viable in fact that intel already ordered the version 2 an even larger monster machine.
    23
  3. If you think EUV optics is crazy then X-ray optics is a lot more crazy. The resist is even more problematic with X-rays. You have backscattering and the photons are energetic enough to alter and cook your substrate.
    15
  4. They are doing weird things to the light source to get the highest contrast possible and this puts a lot of constraints in the optics geometry. There is just no other way. The main constraint is the intensity of the source itself. Brighter source allows faster exposure. But the intensity is already fix because of how the EUV is generated so they have to resort to optics jujitsu.
    14
  5. It is already 3d printed at a much faster rate than macro scale 3d printing. The whole fabrication process is printing the 3d transistor structure on top of the silicon wafer then after that printing the 3d labyrinth of wiring on top of the 3d transistor so that they function as a logic circuit.
    10
  6. The future is a lot harder because you don't know which technology direction will win in the end.
    10
  7. You need a wavelength that silicon is transparent. This limits your light to infrared
    10
  8. Why would it be downloadable do you have spare silicon fab in your kitchen?
    9
  9. AMD will just buy whatever process nodes TSMC delivers. Remember that Intel was overtaken by TSMC because they put too much technology at once that turned into multi generation delay when they could not deliver. This is the same song and dance again Intel chooses multi technology jump while TSMC chooses to do one step at a time.
    8
  10. TSV is not the critical step in BSPN it is the grinding into precise nanoscopic thickness.
    7
  11. The parts are not off the shelf component. They are custom parts tailored to ASML's design. The design itself is the outcome of a multi year research and development.
    6
  12. Intel recently rename their nodes so that they are comparable to TSMC.
    6
  13. No EUV use gravity to drop a droplet of molten tin which is then vaporized with high intensity laser. Doing it in 0g makes the process of hitting a droplet of molten tin with lasers a lot harder because the droplet will float into random direction.
    6
  14. No. Just because you have already melted the salt does not mean they are charged. This are rechargeable battery you need to charge them. Getting melted salt is just preconditioning so that the battery can actually be charged.
    5
  15. Smaller transistor= smaller impedance= faster switching.
    5
  16. Removing all logic removes a lot of noise these allows higher density. The thing with SRAM is that it is partially analog. The two bitline does not only carry data but they also serves a function when reading and writing. When you have logic that is very noisy the bitlines are giant antenna that might flip the state of the SRAM cell.
    5
  17. Are you willing to pay the man-hours needed for your optimized calculator?
    5
  18. linkzable I just read the article on soft x-ray it is not even lithography just x-ray imaging. X-ray imaging is the device use in hospitals to see your skeleton. It is a completely different technology. Yes now you can see your bones at a higher resolution but you cannot use it to manufacture microchips. Even the optics use for imaging is completely different to the optics use for lithography. And then where is the technology for x-ray mask. Where is the technology for x-ray photoresists. Where is the technology for 2nm features overlay. Where is the technology for x-ray aberration correction. Lithography is not just firing lasers in fact firing lasers is the most trivial component in lithography.
    5
  19. linkzable Your contacts are idiots. Again where is the x-ray mask technology. Where is the x- ray photoresists technology. Where is the overlay technology. Where is the aberration correction technology. If they think lasers = lithography then they are dumber than a glue sniffing monkey. And since you believe their rambling induced by their glue sniffing high your intelligence must be even lower than theirs.
    5
  20. E beam lithography is to slow. It is good for laboratory prototypes but not for mass production.
    5
  21. Process nanometer is marketing. 1nm process is actually 20nm pitch. We are nowhere near atomic resolution except for layer thickness.
    5
  22. Tooling is highly reusable while the mask is consumable. Every new chip design needs several hundred different masks but you can reuse the EUV scanner.
    4
  23. It is very good technology but expensive. Gamers could not afford them but AI companies could.
    4
  24. This are work by hundreds of engineers from all over the world. Improving fabrication technology is a global effort.
    4
  25. 13.5nm needs vacuum chamber because air is opaque to the wavelength so immersion lithography where they use ultra pure water as focusing lens is not possible with EUV. But multi patterning is possible with EUV and it is part of the roadmap for future improvement.
    4
  26. Transistor are tiny on off switches. If you put voltage on the gate it turns on the source to drain line. Imagine you have a redstone line with a cut off device in between. When you put a redstone signal to the side of the device it cuts off the redstone line. I believe this is how locking relay works. If you put a redstone relay in between a redstone line then put a strong signal to the side it locks the relay and cut the redstone line.
    4
  27. Even if all ice on earth melted sea level will not rise more than 2 meters. Floating structure does not work for heavy industry nor high density residential. Singapore just don't have enough space for a sprawling floating city. For now importing sand is cheaper if that become a problem then the next less expensive solution is to put their industry underground.
    4
  28. ​ @brodriguez11000 It is already the printing industry way. The modern printing industry uses lithography technology not the imprint technology use in Gutenberg prints.
    4
  29. ​ @Gameboygenius It's not economically unviable just economically not competitive. The issue is throughput but as the EUV tech becomes more and more expensive with hyper NA and quad patterning there should be a crossover point where nano imprint becomes competitive. NIL's yield issue is independent of resolution but EUV's yield becomes more problematic as you shrink the features.
    4
  30. Making complex 3d structure transistor requires several steps. This steps involves multiple passes to your expensive lithography equipment. That means each wafer will have to amortize a larger fraction of the equipments capital and maintenance cost.
    3
  31. LEGOS LEGOS LEGOS LEGOS LEGOS LEGOS LEGOS LEGOS LEGOS LEGOS LEGOS LEGOS
    3
  32. No EUV uses different resist and also uses different mask technology. What they learned on DUV does not apply to EUV.
    3
  33. With GAAFET they can move the source and the drain closer together shortening the channel length of the transistor. Also with nanosheet the fin are now sideways and on top of each other so they don't take as much area as current multi-fin transistor.
    3
  34. The problem is not propagation the problem is keeping the signal to noise ratio so that it is still possible to recover the signal in the other end.
    3
  35. All AI technology needs advance processors. We are now at the age of drone warfare. The smarter the drones the more significant the advantage. Imagine a swarm of drone carrying high explosive suicide bombing the very long Russian supply line of trucks stuck in the mud in Ukraine.
    3
  36. Tiny drones need tiny chips. If tiny drones are smart enough to search and destroy fuel trucks they can stop the advance of an armor division.
    3
  37. There is e-beam lithography but that is a lot slower than EUV order of magnitude slower even if they figure out the multi beam version.
    3
  38. AI demands more advanced nodes and they have the money to force going forward.
    3
  39. We do but mostly it is some form of witch brew that does something useful in the past. Microchip fabrication is in another level of complexity comparable to summoning the demon to do something for you.
    3
  40.  @h9hkk6155  maybe nano imprint if they can make that work. But multi-patterning is shoe in for future improvements.
    3
  41. There is no after lithography. The situation is complex but it is not the apocalypse. Yes there is no lithography roadmap after High NA EUV. But there are plenty in the transistor architecture. There is forksheet, CFET, Vertical FET, 2d material channel and many more. This transistor technology together with multi patterned EUV will allow performance improvement for the next decade. The main problem is the escalating cost.
    3
  42. linkzable Your X-ray laser was actually X-ray imaging not lithography and it is laboratory demonstration very far from mass production. China have no EUV technology. This industry cannot be forced with simple manpower. You need brains not muscle given China's severe corruption in academia smart people don't actually rise to the top to lead the research. China actually have to poach Taiwanese engineers to run their foundries. Then corruption sets in and the Taiwanese engineers quit because of unpaid salary. China is using speed to hand over billions to scammers who pocketed the money
    3
  43.  @johndoh5182  You don't know how to count you shrink around 30% per generation. So 3nm then 2nm then 1.8nm then 1.2nm and 0.8nm assuming the rate of 2 years per full node shrink will continue then at 2033 we will be at 0.6nm. I am accounting a slow down so 0.8nm in 2033 is a good prediction.
    2
  44.  @TheEVEInspiration  High resolution scanner are use for critical features but lesser resolution scanners are also use for less critical features. For example 7N process uses an EUV scanner for only the first 2 metal layers while DUV scanners are used for the rest of the 12 to 15 metal layers. The same thing will happen with High NA EUV. At first they will only be used for a few layers while Low NA EUV will be reused for the other layers. Then when Hyper NA becomes available the High NA EUV scanner will be reused for less critical layers.
    2
  45. b Low China needs to spend 10x of the entire silicon industry if they want to realistically catch-up and not take a decade to develop the technology that will be a decade old by time they manage to replicate it. So start counting the total combine R&D budget of Apple, Intel, Samsung, TSMC, IBM, Qualcomm and others. Multiply that by 10 and that should be the budget China needs to allocate to catch-up.
    2
  46. Flash and memory chip manufacturer.
    2
  47. E-beam is to slow. There are research for multibeam E-beam but it becomes complicated very quickly.
    2
  48. Anything Carbon base is dead end. There is a better 2d material for silicon replacement and it is not Carbon.
    2
  49.  EMAN67:RP forum  Better conductor but bad switching characteristic. This what most physicist miss. A transistor needs to be able to turn on and off. A transistor that is always on is useless. If you have a microchip that will switch a wrong transistor on by just talking loudly then that would not give you correct calculation. If your computer suddenly do 1+1=3 because the rickroll song is played near it that will be a very annoying defect.
    2
  50. ​ @CRneu I think silver is worst in diffusion. So you need thicker diffusion barrier which cut space for more conductivity.
    2