One thing I never understood about processors is why exactly do they have to be so small, especially if quantum tunneling is a problem?
[QUOTE=Laserbeams;51040166]One thing I never understood about processors is why exactly do they have to be so small, especially if quantum tunneling is a problem?[/QUOTE]
sum efficiency increase, if i understand correctly
[QUOTE=Laserbeams;51040166]One thing I never understood about processors is why exactly do they have to be so small, especially if quantum tunneling is a problem?[/QUOTE]
because transistors and processors are the foundations of every silicon-based technology, so things like phones need smaller and smaller transistors in order to sustain more processing power. compare the original iphone or smartphones to the ones we have now, and compare how strong they are to perform even the simplest of tasks, like smoothing swiping an image over and performing all the math for that smooth calculation, while still having accuracy to how strong the image was pushed by the person's finger. everything you do, from typing this sentence and giving the computer a command to press a key, to the screen refreshing and drawing this new text, is done in a split second and with all the different commands a computer or phone or whatever is running, it adds up very quickly. hense, we need smaller components so we can put more power in every one of them and keep a comfortable size.
[QUOTE=Laserbeams;51040166]One thing I never understood about processors is why exactly do they have to be so small, especially if quantum tunneling is a problem?[/QUOTE]
I'm guessing maybe heat problems? Maybe it's really tricky to design a larger microprocessor that has in-built cooling.
[QUOTE=Gamerman12;51040208]because transistors and processors are the foundations of every silicon-based technology, so things like phones need smaller and smaller transistors in order to sustain more processing power. compare the original iphone or smartphones to the ones we have now, and compare how strong they are to perform even the simplest of tasks, like smoothing swiping an image over and performing all the math for that smooth calculation, while still having accuracy to how strong the image was pushed by the person's finger. everything you do, from typing this sentence and giving the computer a command to press a key, to the screen refreshing and drawing this new text, is done in a split second and with all the different commands a computer or phone or whatever is running, it adds up very quickly. hense, we need smaller components so we can put more power in every one of them and keep a comfortable size.[/QUOTE]
This is a nutshell, smaller transistors (see Feature Size) => higher scale of integration (see VLSI), which also means faster turn on time => ergo faster frequency of operation => more performance.
Miniaturization is the key to the information age.
[QUOTE=LoneWolf_Recon;51041678]This is a nutshell, smaller transistors (see Feature Size) => higher scale of integration (see VLSI), which also means faster turn on time => ergo faster frequency of operation => more performance.
Miniaturization is the key to the information age.[/QUOTE]
Its also important to remember that you can't simply expand your silicon die size and expect improved performance. Say you wanted to increase the performance of an existing processor, you may think that you could simply expand your die and use the new area to add in more hardware and therefore increase performance. However, you need to take into account the propagation delay of signals between transistors. If it takes more time for a signal to move across a chip than for a given transistor takes to stabilize its output, you can run into all sorts of timing hazards where transistors are outputting signals based on outdated inputs. You can fix this by reducing the clock speed in order to allow the signals more time to reach their destinations, but at that point the performance you gain from additional hardware is offset by the performance lost from reducing the clock speed.
[QUOTE=amos106;51042554]Its also important to remember that you can't simply expand your silicon die size and expect improved performance. Say you wanted to increase the performance of an existing processor, you may think that you could simply expand your die and use the new area to add in more hardware and therefore increase performance. However, you need to take into account the propagation delay of signals between transistors. If it takes more time for a signal to move across a chip than for a given transistor takes to stabilize its output, you can run into all sorts of timing hazards where transistors are outputting signals based on outdated inputs. You can fix this by reducing the clock speed in order to allow the signals more time to reach their destinations, but at that point the performance you gain from additional hardware is offset by the performance lost from reducing the clock speed.[/QUOTE]
True, and also there's talk and some progress into 3D FETs by stacking layers of transistors vertically to reduce said propagation delay but add complexity into manufacturing. (Also imagine the leakage current involved. :sick: )
[QUOTE=amos106;51042554]Its also important to remember that you can't simply expand your silicon die size and expect improved performance. Say you wanted to increase the performance of an existing processor, you may think that you could simply expand your die and use the new area to add in more hardware and therefore increase performance. However, you need to take into account the propagation delay of signals between transistors. If it takes more time for a signal to move across a chip than for a given transistor takes to stabilize its output, you can run into all sorts of timing hazards where transistors are outputting signals based on outdated inputs. You can fix this by reducing the clock speed in order to allow the signals more time to reach their destinations, but at that point the performance you gain from additional hardware is offset by the performance lost from reducing the clock speed.[/QUOTE]
IIRC it's mainly because of this that you can't overclock something to an extremely high frequency, even if you have some sort of super-cooling that takes care of all the extra energy
Sorry, you need to Log In to post a reply to this thread.