4 ms·

transistor != chip

by dspeyer 17y ago

transistor != chip

pohl 17y ago
True. Would someone familiar with the industry explain what it means that they "created a wafer-scale device"?
- sparky 17y ago
  They manufactured a whole wafer of the the things at once, rather than one researcher hovering over a microscope to painstakingly construct one or a handful of transistors. It speaks to the manufacturing-readiness of the process they use.
hga 17y ago
And transistor != logic family. I want to know what the fan-in and fan-out of the ensuing logic family will be (that's what killed Josaphen (sp.) junctions as the next cold thing). But they did fabricate it on a silicon wafer, which is an advance. Not strictly related to Moore's Law (which talks about the increase in the number of devices you can put on a piece of silicon), but if we could exit the CMOS speed trap we're currently in that would be very nice.
- jonsen 17y ago
  Josaphen (sp.) junctions I think you mean Josephson junction: http://en.wikipedia.org/wiki/Josephson_effect http://en.wikipedia.org/wiki/Josephson_effect
  - hga 17y ago
    Yes. You can make very fast parts with it, and for a long while the hype was that it was the future for high speed computers, but that never came to pass. I read in Science that the problem was insufficient fan-in/fan-out, it just wasn't practical to design computers based on Josephson junctions, and on top of all the practical problems (keeping one cold, issues of going from normal temperatures to cold and back, fabrication) that was that.
- jules 17y ago
  What is the CMOS speed trap?
  - hga 17y ago
    That's my term for how the Pentium 4 Netburst "marchitecture" hit an unexpected dead end. Externally they said they'd push it to 10GHz, I've heard that internally they expected at least 5GHz, but they were only able to push it to 3.8GHz (and more often 3.6GHz; 8GHz if you used liquid nitrogen cooling as reported by some overclockers). Heat dissipation scaled in a way they didn't expect and this forced a return to the wider Pentium Pro architecture approach (one AMD never left). If you look at Intel's current processors (at least on the Wikipedia lists) you'll see they tend to run between 2GHz to 3GHz, with only a few at the latter and above, with none more than 3.3GHz. Their future roadmap doesn't top 3.6GHz.

transistor != chip

by dspeyer 17y ago

transistor != chip

pohl 17y ago

True. Would someone familiar with the industry explain what it means that they "created a wafer-scale device"?

sparky 17y ago
They manufactured a whole wafer of the the things at once, rather than one researcher hovering over a microscope to painstakingly construct one or a handful of transistors. It speaks to the manufacturing-readiness of the process they use.

hga 17y ago

And transistor != logic family. I want to know what the fan-in and fan-out of the ensuing logic family will be (that's what killed Josaphen (sp.) junctions as the next cold thing). But they did fabricate it on a silicon wafer, which is an advance. Not strictly related to Moore's Law (which talks about the increase in the number of devices you can put on a piece of silicon), but if we could exit the CMOS speed trap we're currently in that would be very nice.

jonsen 17y ago
Josaphen (sp.) junctions I think you mean Josephson junction: http://en.wikipedia.org/wiki/Josephson_effect http://en.wikipedia.org/wiki/Josephson_effect
- hga 17y ago
  Yes. You can make very fast parts with it, and for a long while the hype was that it was the future for high speed computers, but that never came to pass. I read in Science that the problem was insufficient fan-in/fan-out, it just wasn't practical to design computers based on Josephson junctions, and on top of all the practical problems (keeping one cold, issues of going from normal temperatures to cold and back, fabrication) that was that.
jules 17y ago
What is the CMOS speed trap?
- hga 17y ago
  That's my term for how the Pentium 4 Netburst "marchitecture" hit an unexpected dead end. Externally they said they'd push it to 10GHz, I've heard that internally they expected at least 5GHz, but they were only able to push it to 3.8GHz (and more often 3.6GHz; 8GHz if you used liquid nitrogen cooling as reported by some overclockers). Heat dissipation scaled in a way they didn't expect and this forced a return to the wider Pentium Pro architecture approach (one AMD never left). If you look at Intel's current processors (at least on the Wikipedia lists) you'll see they tend to run between 2GHz to 3GHz, with only a few at the latter and above, with none more than 3.3GHz. Their future roadmap doesn't top 3.6GHz.