# Need help understanding the concept of "power wall"

I am currently studying computer architecture and I am having some trouble understanding the concept of the "power wall".

I came across the following statement:

The term power wall typically means that the maximum clock speed has been reached and that chips would be too hot if they were clocked at higher frequencies.

I am not quite sure what this means or how it relates to computer architecture.

As far as my understanding goes, clock speed refers to the frequency at which a processor executes instructions, and that higher clock speeds lead to higher performance. I am not clear on how clock speed relates to the concept of power wall and why increasing clock speed would lead to higher heat generation.

I would appreciate any help or clarification on this topic. If anyone could explain the concept of power wall in simpler terms or provide me with any references or resources to better comprehend this topic, I would be grateful.

• This is a duplicate of the SuperUser question superuser.com/q/163567/252680 Feb 20, 2023 at 19:42
• Someone in that thread is really going on about gate capacitance, which is a thing, but the other wrinkle is that MOSFETs dissipate a lot of heat (relatively) in the transition between high-resistance/low-current and low-resistance/high-current states (off and on). Faster clock speeds mean more transitions and more heat, in addition to the higher average currents (and higher resistive heating) from charging and discharging that gate capacitance.
– vir
Feb 20, 2023 at 19:47
• power wall ... I think that there may be a word missing in the middle ... maybe something like power dissipation wall ... it may have been written by a person who refers to a USB flash drive as USB Feb 20, 2023 at 20:39
• (@jsotola USB flash drive vs. USB drive, not just USB. (But, occasionally, just USB stick when USB memory stick was prominent.) Feb 21, 2023 at 5:18

It simply means you can build enough computing power into so small space that you can't use all that computing power because the thing would heat too much and melt.

The switching power P dissipated by a chip is proportional to capacitance (C), voltage squared (V^2), and frequency (f). Therefore, the higher frequency it runs at, the more power it consumes and turns into heat. Also higher frequencies need higher voltage to work, and that effect from voltage is to the power of 2.

• The term "power wall" does refer to a limit on the clock speed of processors due to the amount of power consumed and the resulting heat dissipation. However, it is not solely a matter of chips being too hot if they were clocked at higher frequencies. The term actually refers to a combination of factors that limit a processor's performance, including power consumption, heat dissipation, and other technological constraints. Feb 20, 2023 at 20:03
• These factors can make it difficult or impossible to continue increasing the clock speed, leading to a "wall" that limits the processor's performance. So while heat is certainly a factor in the power wall, it is not the only factor and should not be considered the sole cause of the power wall. So I would say that the statement "The term power wall typically means that the maximum clock speed has been reached and that chips would be too hot if they were clocked at higher frequencies." is incorrect. Feb 20, 2023 at 20:03
• @bitts where are you quoting that from, professor? I didn't realize you were "holding class". I thought you were a student. Ever hear the term "sophomoric"? Feb 20, 2023 at 20:06
• @Harper-ReinstateMonica It's the statement from the question. Feb 20, 2023 at 20:20

MOSFETs only consume power when they change state so the higher the clock rate, the more power consumed. More transistors means more power.

Moore's Law is still going, but the line is flattening and at some point we will reach the laws of physics. The physics wall.

Maximum frequency and corresponding power have flattened out. Dennard scaling which states power density stays constant as transistors get smaller. Dennard scaling is over. The power wall.

One use for extra transistors is multi-core processors, but multi-core require operations that can be paralleled. Portions of software can be run in parallel, but some must be run in sequence. Single-thread performance is flattening out. Amdahl's law means potential improvements to performance by multi-core processors are limited by the software which can't be run in parallel.

Multi-core processors are a waste for most users, so it could be argued that making the transistor smaller has no benefit unless power dissipated by individual transistors can be removed.

Remove heat and processors can go faster and performance can increase.

Cell phones attempt to get around this by utilizing specialized cores. Cores only consume power when feature is in use.

Chip Design Shifts As Fundamental Laws Run Out Of Steam

• Actually, transistor geometries are so small that a significant amount of static power is consumed, so there is power consumption even if dynamic switching power consumption is zero when switching is stopped. Feb 20, 2023 at 21:44
• Tunneling through the oxide layer, which is definitely a power problem but not directly related to clock speed - the original statement might need some updating.
– vir
Feb 20, 2023 at 22:07