Honey I Shrunk the Chips: How die shrinks help make processors more powerful

2»

Comments

  • Reply 21 of 22
    knowitallknowitall Posts: 1,648member
    JWSC said:
    Even when UV lithography becomes the norm, you’ll still have to deal with quantum tunneling. That means another significant change to transistor design is in the cards.
    I see this as added noise, so this is in conflict with lowering voltage levels.
    A bigger problem is that heat (dissipation) is  proportional to the inverse square of the feature size.
    This means the chip will burn up at certain hot spots when below a certain feature size, unless counter measures like lower voltage are taken.
     0Likes 0Dislikes 0Informatives
  • Reply 22 of 22
    maestro64maestro64 Posts: 5,043member
    knowitall said:
    maestro64 said:
    Part of the reason Apple is able to keep jumping to the next process technology is due to the fact they do not have to worry about hardware backward compatibility and they have complete end to end control over the entire design. When you shrink the process technology on a specific chip design, sometimes there are unforeseen side effects on how the chip runs. If you have to make sure the current chip works the same way as the previous chip (like Intel Processors, and Memory Chips) it can make it very difficult to process jump. Intel can not all of a sudden jump process technologies and tell everyone the processor now behaves differently and Computer companies have to redesign the boards and Microsoft has to change how Windows works.

    This is why with every generation of the Apple A processors there is new hardware around it and new software. Apple only has to worry about Apple. Other companies do not have this luxury so they have to make sure when they make a process change the new chip works like the old one, or they have to modify the chip to work the same way in the new process technology as it did on the previous technology and this is not trivial.
    The instruction set has to behave exactly as specified and cannot deviate. Sometimes chips have hardware (instruction) errors possibly in combination with a specific run of instructions in advance. This can be corrected by not letting the compiler emit such sequences. But some code is written in assembly and must behave as specified for the ARM instruction set.
    So, it might be possible to release A processors with instruction defects, because App programmers cannot write code in assembly (language) and fix it by using adapted swift and C/ObjectiveC compilers, but this wouldn't work for macOS programming.
    I sincerely doubt Apple releases instruction defect A processors, because of the complexity involved in fixing it (I know because I've done that).
    But it is possible that some A processors were defective ...

    This is why Apple does not want developers writing directly to the hardware, this is why Apple developed the Mac OS toolbox calls as well as API, Apple can modify the underlining code to fix hardware defects and timing issues with our developers not needing to change their code.
     0Likes 0Dislikes 0Informatives
Sign In or Register to comment.