3 ms·

Once again, absolutely amazing. Those are more details of a really interesting internal CPU bug than I could have ever hopes for. Ken, do you think in some fut

by anyfoo 4y ago

Once again, absolutely amazing. Those are more details of a really interesting internal CPU bug than I could have ever hopes for. Ken, do you think in some future it might be feasible for a hobbyist (even if just a very advanced one like you) to do some sort of precise x-ray imaging that would obviate the need to destructively dismantle the chip? For a chip of that vintage, I mean. Obviously that's not an issue for 8086 or 6502, since there are more than plenty around. But if there were ever for example an engineering sample appearing, it would be incredibly interesting to know what might have changed. But if it's the only one, dissecting it could go very wrong and you lose both the chip and the insight it could have given.[1] Also in terms of footnotes, I always meant to ask: I think they make sense as footnotes, but unlike footnotes in a book or paper (or in this short comment), I cannot just let my eyes jump down and back up, which interrupts flow a little. I've seen at least one website having footnotes on the side, i.e. in the margin next to the text that they apply to. Maybe with a little JS or CSS to fully unveil then. Would that work? [1] Case in point, from that very errata, I don't know how rare 8086s with (C)1978 are, but it's conceivable they could be rare enough that dissolving them to compare the bugfix area isn't desirable.

molticrystal 4y ago
>some sort of precise x-ray imaging that would obviate the need to destructively dismantle the chip I don't know about a hobbyist, but are you talking about something along the lines of "Ptychographic X-ray Laminography"? [0] [1] [0] https://spectrum.ieee.org/xray-tech-lays-chip-secrets-bare https://spectrum.ieee.org/xray-tech-lays-chip-secrets-bare [1] https://www.nature.com/articles/s41928-019-0309-z https://www.nature.com/articles/s41928-019-0309-z
- anyfoo 4y ago
  I haven't ever looked into it itself, but what you just pasted seems like a somewhat promising answer indeed. Except for the synchrotron part I guess? Maybe?
  - sbierwagen 4y ago
    >Except for the synchrotron part I guess? Maybe? They were using 6.2keV x-rays from the Swiss Light Source, a multi-billion Euro scientific installation. 6.2keV isn't especially energetic by x-ray standards, (a tungsten target x-ray tube will do ten times that) so either they needed monochromacy or the high flux you can only get from a building-sized synchrotron. Given that the paper says they poured 76 million Grays of ionizing radiation into an area of 90,000 cubic micrometers over the course of 60 hours suggests the latter. (A fatal dose of whole-body radiation to a human is about 5 Grays. This is not a tomography technique that will ever be applied to a living subject, though there is some interesting things no doubt being done right now to frozen bacteria or viruses.)
    - anyfoo 4y ago
      No question, but: "Though the group tested the technique on a chip made using a 16-nanometer process technology, it will be able to comfortably handle those made using the new 7-nanometer process technology, where the minimum distance between metal lines is around 35 to 40 nanometers." The discernible feature size required for imaging CPUs of 6502 or even 8086 vintage is much lower than that. The latter was apparently made with a 3 µm process, so barring any differences in process naming since, that's about 3 orders of magnitude less. Plus, as Ken's article says, the 8086 has only one metal layer instead of a dozen. So my (still ignorant) hope is that you maybe don't need a synchrotron for that.
- kens 4y ago
  That X-ray technique is very cool. One problem, though, is that it only shows the metal layers. The doping of the silicon is very important, but doesn't show up on X-rays.

by anyfoo 4y ago

molticrystal 4y ago

>some sort of precise x-ray imaging that would obviate the need to destructively dismantle the chip I don't know about a hobbyist, but are you talking about something along the lines of "Ptychographic X-ray Laminography"? [0] [1] [0] https://spectrum.ieee.org/xray-tech-lays-chip-secrets-bare https://spectrum.ieee.org/xray-tech-lays-chip-secrets-bare [1] https://www.nature.com/articles/s41928-019-0309-z https://www.nature.com/articles/s41928-019-0309-z

anyfoo 4y ago
I haven't ever looked into it itself, but what you just pasted seems like a somewhat promising answer indeed. Except for the synchrotron part I guess? Maybe?
- sbierwagen 4y ago
  >Except for the synchrotron part I guess? Maybe? They were using 6.2keV x-rays from the Swiss Light Source, a multi-billion Euro scientific installation. 6.2keV isn't especially energetic by x-ray standards, (a tungsten target x-ray tube will do ten times that) so either they needed monochromacy or the high flux you can only get from a building-sized synchrotron. Given that the paper says they poured 76 million Grays of ionizing radiation into an area of 90,000 cubic micrometers over the course of 60 hours suggests the latter. (A fatal dose of whole-body radiation to a human is about 5 Grays. This is not a tomography technique that will ever be applied to a living subject, though there is some interesting things no doubt being done right now to frozen bacteria or viruses.)
  - anyfoo 4y ago
    No question, but: "Though the group tested the technique on a chip made using a 16-nanometer process technology, it will be able to comfortably handle those made using the new 7-nanometer process technology, where the minimum distance between metal lines is around 35 to 40 nanometers." The discernible feature size required for imaging CPUs of 6502 or even 8086 vintage is much lower than that. The latter was apparently made with a 3 µm process, so barring any differences in process naming since, that's about 3 orders of magnitude less. Plus, as Ken's article says, the 8086 has only one metal layer instead of a dozen. So my (still ignorant) hope is that you maybe don't need a synchrotron for that.
kens 4y ago
That X-ray technique is very cool. One problem, though, is that it only shows the metal layers. The doping of the silicon is very important, but doesn't show up on X-rays.