I might have been the person who "introduced" this concept that better heat transfer means a hotter amplifier case. It is not strictly true as there are some (important) nuances to it that need explanation.
If you have two devices of equal size and shape and generate the same amount of heat, at steady state, the average chassis temperature will have to be roughly the same. The reason is: At steady state, heat generated = heat dissipated.
Heat transfer to the surrounding is governed by the difference between the chassis surface temperature and the ambient air temperature. If the amounts of heat dissipated to the surrounding are the same, with amp chassis of similar similar dimensions, their surface temperatures must also be similar (when averaged over the chassis surfaces), regardless of how well or poor the heat conduction paths from the heat sources to the amp chassis are.
What a poor thermal conduction path will affect is the time response, i.e. how quickly the chassis temperature will rise. To demonstrate I'll show simulation results of both cases (good and poor thermal conduction paths to the chassis).
Only a quarter of the geometry needs to be modeled due to symmetry. The heat released by the amp chip is transferred to a heat spreader (or heat sink). The heat spreader is mounted to the chassis, with thermal interface materials (TIM) applied to fill the gaps between mating surfaces of the heat spreader and the chassis. The chassis is cooled by convection (convection heat transfer coefficent of 7.2 W/m²·°C and 25 °C ambient temperature). We'll look at the differences between effectively applied TIM and poorly applied or no TIM (50x higher thermal resistance).
View attachment 541948
In the first scenario, the amp chip released 22 W of heat for 300 seconds and then the heating stopped. Below is an animation of the 2 temperature charts. When thermal conduction was poor, we can see that the amp chip temperature rose quickly and to a level much higher than when conduction is good. However, the chassis temperature rose slower and the peak temperatures didn't reach as high as with the good heat conduction case. Therefore, in this scenario, the amp with the poorer heat conduction will have a cooler amp chassis.
View attachment 541949
In the second scenario, we kept the heat generation on continuously. We can see that for this scenario the eventual steady state chassis temperatures are the same for poor and good thermal conduction. It just took the poor conduction case longer to reach the same temperature. Notice that with poor conduction the amp chip temperature went off the chart. That means the amp would probably shut itself down or destroy itself.
View attachment 541950
Therefore, the more accurate statement is that for a system with high thermal conduction to the chassis (or heat sink), the chassis will heat up faster than one with poor heat conduction. If you have enough patience and keep the test running for a long time, the one with poor heat conduction will eventually catch up. But you may need to keep the test on for half an hour or longer.
Below are the temperature profiles with heating on for 300 seconds.
View attachment 541951
View attachment 541952