4 ms·
I'm super curious -- how does chip manufacturing yield go with node size? Do they need to design even more defect-tolerant designs if the density of random env
by supernova87a 6y ago
I'm super curious -- how does chip manufacturing yield go with node size?
Do they need to design even more defect-tolerant designs if the density of random environmental contamination stays about constant? (Is that true?) Or do they need the clean rooms, solvents, filtration, etc. to improve just as much with every jump in node size?
I was also wondering, since Apple is not marketing on Ghz and playing the Intel game, how do they deal with variable chip manufacturing yield? Is there a narrow band or minimum chip speed that they'll accept and toss anything performing below?
Does the SoC become so physically larger that yields also go down by any significant amount, and any defects at such small sizes become much more troublesome?
- Tuna-Fish 6y ago> Do they need to design even more defect-tolerant designs if the density of random environmental contamination stays about constant? (Is that true?) Or do they need the clean rooms, solvents, filtration, etc. to improve just as much with every jump in node size? We are way past that already. Any external contamination, even one stray atom, can cause a defect in any reasonably modern lithography process. There are parts of chips where thicknesses are measured in "monolayers" -- that is, the thickness of the material in that position is one uniform layer of atoms.
- willis936 6y agoDo the photolithography machines operate in vacuum? If not, why not?
- genericone 6y agoYes, they do, all layering steps additive and subtractive are done in vacuum environments.
- themule 6y agoIf it can be it is in a vacuum. Almost everything is crazier than you think. Many times whenever yields work they just blindly repeat the process - it's a freaking awesome science.
- Tuna-Fish 6y ago> Almost everything is crazier than you think. This is a very good rule of thumb about modern leading edge fab technology. Imagine the craziest, most impractical, most expensive things you can come up with for methods to manufacture things. The actual processes used are crazier than that. For a reference point, just look up how the light source for the EUV steppers work. [0] tl;dr: They are melting tin, dripping the molten tin in droplets of carefully controlled size, shooting those droplets into a vacuum chamber at 70m/s, and then hitting them simultaneously with so many lasers that they are instantly ionized and then emit radiation of the desired wavelength. And they are using this Rube Goldberg-esque contraption as a glorified light bulb. Everything is like that. [0]: https://www.osa-opn.org/home/articles/volume_29/march_2018/features/euv_light_sources_for_next-gen_lithography/ https://www.osa-opn.org/home/articles/volume_29/march_2018/f...
- willis936 6y agoWow. That's a lot of extra steps compared to what they do with other discharge lamps. I'm not an authority in this area and I don't know proper terminology, so forgive me if I'm mistaken here. There must be some good reasons for all of this. Probably a long list of things to do with optics, power density, and contamination. Here is a video of a rubidium standard teardown that I like to share. There aren't many places to actually see a discharge lamp. https://www.youtube.com/watch?v=ymV9LwhD0W0 https://www.youtube.com/watch?v=ymV9LwhD0W0
- supernova87a 6y agoIs the washing and dissolving with solvents in a vacuum too? Doesn't that cause stuff to boil?
- dogma1138 6y ago
- swalsh 6y agoMy understanding is that part of the reason AMD went with a chiplet design was to work around this issue.