0
\$\begingroup\$

I have a question regarding to hardware signal. There is a rule says ["] Smaller is faster. [."], which in the link states that

  1. In high-speed machines, signal propagation is a major cause of delay;

    Q: I don't believe this. I think this is just academia explanation. Or is there any diagram can show the details of this? e.g. the x-axis is number of registers/memory unit and the y-axis is the delay/time Actually you may but you don't have to answer my Q1

  2. In most technologies we can obtain smaller memories that are faster than larger memories. This is primarily because the designer can use more power per memory cell in a smaller design;

    Q: Why more power will make signal faster? Isn't that the speed of current the same regarding to the same medium?

\$\endgroup\$
  • \$\begingroup\$ I feel like it has something to do with capacitance, among other things; if you scale down a capacitor equally in all dimensions, \$A\$ decreases quadratically and \$d\$ decreases linearly, meaning that \$C\$, which is proportional to \$\frac{A}{d}\$, decreases linearly.. \$\endgroup\$ – Hearth Nov 30 '18 at 0:53
  • \$\begingroup\$ Consider the Dk of silicon is something like 10 or 12 (IIRC) and in really fast logic we care about picoseconds of delay, if you want to route a signal across a whole chip (3-10mm, say), the speed of light delay really is significant. Whether it's the dominant limit on switching frequencies, I'm not sure. \$\endgroup\$ – The Photon Nov 30 '18 at 1:28
  • 1
    \$\begingroup\$ Silicon Dk = 11.7 according to the Ioffe Institute...glad I didn't misremember that number too badly. \$\endgroup\$ – The Photon Nov 30 '18 at 1:29
  • \$\begingroup\$ @ThePhoton: Great, thank you then I must be wrong in my 1, according to your comment. \$\endgroup\$ – ptr_user7813604 Nov 30 '18 at 1:30
  • 1
    \$\begingroup\$ Yes, it's a common symbol for dielectric constant. \$\endgroup\$ – The Photon Nov 30 '18 at 2:55
4
\$\begingroup\$

Regarding your Q2, keep in mind that this is in the context of logic gates, which in CMOS logic are switched by driving the gate of a MOSFET. MOSFET gates behave like small capacitors. To change the state of these gates, i.e. to charge or discharge the capacitor, you need to apply a current to it. The time it takes to charge or discharge will depend on the magnitude of the current applied, so more power gets you faster switching.

\$\endgroup\$
  • \$\begingroup\$ This makes perfect sense! Thank you! Every time I read textbook it made me felt like I'm fooled... \$\endgroup\$ – ptr_user7813604 Nov 30 '18 at 1:18
  • 2
    \$\begingroup\$ The change of state also involves resistance of connecting wire (RC anyone?) That's one reason why smaller is faster - shorter wire has less resistance. This is the same reason why silicon manufacturers switched for more conductive copper interconnect instead of aluminum. \$\endgroup\$ – Ale..chenski Nov 30 '18 at 1:47
  • \$\begingroup\$ @Ale..chenski Didn't know they were using copper now, neat! \$\endgroup\$ – Hearth Nov 30 '18 at 2:34
  • \$\begingroup\$ @Hearth, yes, about for 15 years already, since 130nm process, Intel is using copper interconnect and low-K dielectric, Pentium-4/Northwood, web.stanford.edu/class/ee311/NOTES/… \$\endgroup\$ – Ale..chenski Nov 30 '18 at 3:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.