Excuse me for nicking your diagram IDC.
What it is calling "Dark" silicon are the area that do less work (especially since if you are running discrete graphics and the iGPU is off), these areas are cooler than the active cores. For data to get across the chip there have to be traces between the various parts, and these are capable of carrying heat as well as data. As a result cores 1 and 4 (or 0 and 3 if you prefer), are capable of dissipating heat to the cooler silicon to their left/right respectively. Core 3 doesn't really have this option, so almost all the heat has to go straight up through the IHS.
(cache would be slightly cooler wouldn't it IDC? Some heat will also be dissipated out the back of the socket)
I don't understand, which part of your post is the nitpick? Sounds like you are further explaining one of the mechanisms by which those areas are "effectively' dark silicon, an explanation which I happen to agree with. :thumbsup:
The allure of dark silicon is that it helps with the lateral heat transfer through the silicon, remember the heatsink is attached to the backside of the CPU, not the topside.
So while some amount of heat is actively conducted through the BEOL wires forming the circuits themselves, in practice we are relying on the thermal diffusion that occurs within the silicon and that is a three dimensional situation.
The relevance of having darker areas of silicon is that it gives the heat more room to spread out, literally, which lowers the local temperature since temperature is literally heat density.
The definition of dark silicon is "under-utilized transistors". I refer to the non-core areas of the CPU as "
effectively dark silicon" by way of comparison to their utilization (and resulting power density) in comparison to the utilization of the circuits located in the CPU cores.
I think we are both saying the same thing here, just talking past one another? Or am I missing your point entirely?