Originally Posted by Opcode
Please correct us and tell me where we are wrong (AMD and I).
Heat transfer from GPU -> CPU, heat transfer from CPU -> IHS/Heatsink.
I'm really finding it hard to start an explanation for this seeing the logic exposed above...
Screw it, I'm making a starting point myself.
You do know how does the heatspreader work, right? It collects heat from a focus of heat, and evens it out around a larger surface so a heatsink can pick it up easily (lower thermal density>easier to keep up with the energy needed to be transferred). Heatsinks work almost equally as well across all of their surfaces.
Thing is, unless it has a built in heatspreader (only waterblocks and some thick base heatsinks actually do work as heatspreaders too), a heatsink is very inefficient if the heat load is focused on a small area of its total base surface. That happens because heat needs to be moved across a very large horizontal space, and then onto the dissipation area (fins, usually).
If you use a heatspreader, aka a relatively thick slab of highly conductive metal (copper, p.ex), the heat from the focus (die) is evenly spread across the heatsink base, allowing it to function to its maximum performance.
What AMD is trying to do here, is use the inactive (or less active) part of the die, to let some heat from the other part conduct through it and onto the heatspreader. So, it is basically acting as a heatspreader, since it only increases the amount of surface in contact with the die that's conducting heat towards the heatsink.
It is not dissipating any heat by itself. It merely acts as a support medium to reduce thermal resistance between the part that generates the most heat, and the heatsink.
AMD is not wrong. You misunderstood AMD's explanation and went anal about it when we put it in doubt.