Stop blaming the supply for your dissipation woes
It's not news that everyone is interested in efficiency these days. Whether it is because of "green" incentives, run time, or thermal limitations, using less power is a good thing and often a priority. As a consequence, engineers are looking at sources of inefficiency and trying to minimize them, of course.
So, where do they look first, and sometimes almost entirely? At the power supply, of course. But that can be somewhat misleading and self-deceptive.
Go through some simple numbers and you'll see why, using a nominal 100 W supply. A modern AC/DC or DC/DC switching supply has efficiency of between 80% and 85% when operating at half load and above. (These are "typical" numbers, and using them can be a little dangerous - but this is an illustrative example to make a point, not a formal design review.) That means the supply itself is dissipating between 15 and 20 W.
So management says "go get a more efficiency supply." You search and find a supply that is a little more expensive, but has efficiency of 90% at the same load rating. Great: you have cut your supply dissipation down to 10 W. Or looking at it the other way, you have decreased the heat from the supply from between 15 to 20 W down to 10 W - thereby cutting it as much as 50%. Very impressive, for sure, and that's less heat being wasted and needing to be removed.
But wait: even that 20% loss represents only a small fraction of the total heat dissipation of your system's electronics, which is 80 W. Getting a better supply hasn't changed that; it has decreased the supply + circuitry dissipation by only a few percent (I'll let you work out the numbers!). In short: your supply is not the major source of power loss and subsequent dissipation, your circuit is. Maybe you need to work on decreasing the dissipation of the electronics as much as that of the supply, and get yourself down to a 50 or 75 W total load (supply + circuitry) - that would be nice!
The problem is that in most designs that's a lot of work to achieve. Substituting lower-power components, lower-power circuit topologies, and lower-power I/O is hard work, may require new and untried components or designs, and is risky. It's easier to try to get a big chunk of savings in one shot, rather than work watt by watt (or mW by mW) through the circuit.
It's like the old story of the drunk who is looking for wallet he dropped by the lamppost, even thou he lost it in the dark corner on the other side of the street Why does he do that? Because it's easier, although not as productive in the end.
Don't be defensive when people start pointing at the supply's inefficiency as the source of your consumption and dissipation problems. It's OK to point out that such conventional, simplistic thinking may have you looking for trouble in the wrong places. Plus, power demands by the circuitry often expand to consume available resources, and any increase in the supply's efficiency will often become the rationale for adding more functions and features, thus keeping the overall thermal load pretty much where it was before!
Have you ever been in a situation where management and even your team members focused blame the on aspects of the design (e.g., BOM, component selection, or topology) when the numbers showed the problem was primarily elsewhere, either wholely or in part? What did you do about it?