Tech Tuesday for 10/23 Blue Phosphorus and the Flying Dandelions

Oscar Gordon

A Navy Turbine Tech who learned to spin wrenches on old cars, Oscar has since been trained as an Engineer & Software Developer & now writes tools for other engineers. When not in his shop or at work, he can be found spending time with his family, gardening, hiking, kayaking, gaming, or whatever strikes his fancy & fits in the budget.

Related Post Roulette

7 Responses

  1. Michael Cain says:

    TT14: A little booster: 150 kg to 500 km orbit max. Although if they’re cheap, some of us may live to see an actual Kessler Syndrome as everyone with a couple million dollars throws something into LEO.

    Separately, reports have appeared over the last few days that Intel is scrapping its 10nm fabrication process, at least for the time being. Rumors are that they have been unable to get the yields up to the level where the they can make money. I have long said Rock’s Law — the cost of a leading-edge fab doubles every four years — would be a limit sooner than the more famous Moore’s Law. A number of the Far East foundry companies stopped at 20nm because of the costs to go smaller. I have a bet that 10nm is the bottom for everyone except companies backed by national governments (eg TSMC (Taiwan) and Samsung (South Korea)), and even they won’t ever go below 7nm.Report

    • Oscar Gordon in reply to Michael Cain says:

      150kg is a whole bunch of cubesats. Governments are going to have to start requiring that what goes up must be able to reliably bring itself down.

      So what is the reason for the cost increase for IC fab? And if we can’t get below 10nm, what does that mean for IC (besides the obvious of, we can’t make smaller devices).Report

      • Technical note: the X-nm designations have lost a great deal of their meaning except relative to each other within a manufacturer’s fab family. 10nm means “twice as many transistors per square mm as 14nm”. Some of that may be due to smaller features; some of it may be due to stuff like finFETs where transistor structure is 3D rather than 2D. Samsung announced a 7nm process yesterday. In terms of transistor density, it’s about the same as Intel’s 10nm process. In a year or so we’ll know whether they can find enough customers and get high enough yields to make money at it.

        IC fab is expensive because it sits at the top of a huge pyramid of engineering tech in many fields. Consider just physical positioning: for some 10nm features, macro-scale objects have to be positioned relative to each other with sub-nm accuracy in three dimensions. On the order of a hundred times to get one chip, if it’s a complex one. In some cases, to overcome the limits of the optics, moving macro-scale objects are positioned with sub-nm accuracy relative to one another.

        If it’s not profitable to go below 10nm? More parallelism, more special-purpose processors. NVidia’s latest offerings are done using someone’s 12nm process. Big dies; pushing 20B transistors; special processors for ray-tracing; special processors for neural networks; thousands of shader cores. More and more supercomputers are being built from GPUs, not CPUs.Report

  2. bookdragon says:

    TT5: Potential is there still a bit pie in the sky. ARL has actually been working on this for over a decade.

    TT6: I don’t know that it’ll get anywhere, but the CFD is cool and I love the images – very scifiReport

  3. Oscar Gordon says:

    Notable passing: Raye Jean MontagueReport