daglivewire
Newbie level 1
My friend posed a question to me today that I wasn't quite able to answer. I work in hardware but am not an expert on CPUs so I'm hoping someone can provide me with some insight.
Let's ignore economic viability for a minute. What if we took an Ivy Bridge CPU, slapped on a state of the art cooling system, and overclocked it. OK, so we get up to 5 or 6 GHz and eventually the power dissipation becomes too high and the processor burns up.
So, back to the drawing board. Let's take the same architecture, same 22 nm technology, but now let's make the wires between transistors a little longer. Still short enough that the CPU meets timing requirements, but a little bit longer so that there is more space between transistors. When we overclock it again, presumably we can make it run a little faster before it burns up because it has a larger surface area for the heat to dissipate out of.
My question is this: if we continued to move the transistors further and further apart, what physical limit would eventually prevent us from increasing the CPU speed?
Would it be switching speed due to carrier mobility in the semiconductor? Parasitics of the junction? Parasitics of the transmission line?
Would it be signal's propagation speed through the wire from transistor to transistor?
Would the limit still be thermal?
Might it be something that I missed?
TLDR: What physical limit would first prevent us from increasing the clock rate on the current gen of CPUs?
Let's ignore economic viability for a minute. What if we took an Ivy Bridge CPU, slapped on a state of the art cooling system, and overclocked it. OK, so we get up to 5 or 6 GHz and eventually the power dissipation becomes too high and the processor burns up.
So, back to the drawing board. Let's take the same architecture, same 22 nm technology, but now let's make the wires between transistors a little longer. Still short enough that the CPU meets timing requirements, but a little bit longer so that there is more space between transistors. When we overclock it again, presumably we can make it run a little faster before it burns up because it has a larger surface area for the heat to dissipate out of.
My question is this: if we continued to move the transistors further and further apart, what physical limit would eventually prevent us from increasing the CPU speed?
Would it be switching speed due to carrier mobility in the semiconductor? Parasitics of the junction? Parasitics of the transmission line?
Would it be signal's propagation speed through the wire from transistor to transistor?
Would the limit still be thermal?
Might it be something that I missed?
TLDR: What physical limit would first prevent us from increasing the clock rate on the current gen of CPUs?