Power-Density
Power density describes power dissipated per unit area or volume. For chips, an areal value such as watts per square millimeter is common, but the averaging region must be stated. Whole-die average power density and the local power density of a busy execution block can differ substantially. Neither quantity alone is the junction temperature.
A simple example separates power from area. A hypothetical 20-watt chip spread uniformly across 100 square millimeters has an average of 0.2 watts per square millimeter. Putting the same power into 50 square millimeters doubles that average. These are arithmetic examples, not measurements of any named processor or process generation.
Switching power connects circuit activity with electrical design. A common model expresses dynamic power using activity, capacitance, clock frequency, and the square of supply voltage. Increasing frequency or switching more circuitry can raise power even as individual features become smaller. Leakage adds another contribution that is not captured by the switching term alone.
Temperature depends on the heat-removal path and time history. Package materials, thermal interfaces, heat spreaders, cooling conditions, and the distribution of active blocks all matter. A small hot region can create a local temperature problem while the whole-die average seems moderate. Thermal interpretation therefore needs spatial information and a defined operating workload.
The deep-submicron lesson is that more transistors do not create a free power budget. Process scaling, voltage choices, clocking, and architecture must be considered together. Comparing two processors by nominal node size alone cannot establish which has lower power density. Report measured or modeled power, area, workload, and cooling assumptions before drawing that conclusion.
Sources and further reading. [Intel Technology Journal: power modeling](https://www.intel.com/content/dam/www/public/us/en/documents/research/2010-vol14-iss-3-intel-technology-journal.pdf).
Read Power density as the spatial concentration of dissipated power that must be interpreted together with workload and heat removal.