"Total Board Power" (while sometimes referred as TDP which has a pretty generic and loose definition nowadays) generally refers to the power the whole thing (GPU + memory) can burn at full utilization. Not sure what you meant by "planned power usage".If you want to match chip for chip, you have to look at the number of shaders & the planned power usage...not the TDP which represents the max the chip is capable of. As you announced, Vega & Polaris are going hand in hand. The only difference between the two (apart from shaders) is the updated design to optimize Vega. So why not update the Med & Low end chips with that? Because it won't be worth the money. Why didn't they release Polaris chips with the same max shaders as the high end previous gens? Because nVidia beat them to the punch & they want faster high end chips with the same chip size. That's how they make their money.
They are on the (more or less) same process node, so TDP tells you the relative value preposition. I wouldn't call a 232 mm^2 Polaris a "high-end" chip in any sense either. Polaris is just a quick attempt to recapture some lost ground in notebook and mainstream PC. Not brilliant, but it saved AMD some ground.
Vega 10 is expected to be more than a double of the size (and more or less the architectural scale) of Polaris 10 (>480 mm^2), with a more expensive but compact & efficient HBM memory. That's what really high-end is. Moreover, AMD already said there are two Vega GPUs, and if you assume both are using HBM, obviously the second Vega would not be anything weaker than Polaris 10. This means Vega is not going to cover the entire GPU spectrum, but only the high-end. So what fills the hole? Polaris.
Last edited: