The reduction in the process size could mean a lower TDP, although that is not really guaranteed. So far AMD has had to crank up the power to get performance from their top end cards and Artic Islands could end up in the same boat. AMD is switching from HBM1 and its 4GB limitation to the much more mature HBM2. Doing this gives their GPUs a much larger available bandwidth and more memory to play with as HBM2 and run up to 32GB. Of course, AMD has to figure out how to use all of that properly or it is just an expensive waste of time.
On the other side of the coin, HBM2 is a lot like HBM1: there is not much of it out there and probably won’t be for a while. This means that AMD will have another card that is starved for parts just to get them on the streets. AMD will also have NVIDIA to contend with for the HBM2 supply. Here AMD is at a disadvantage as NVIDIA has much deeper pockets and can afford to pay a premium to get the parts they need.
HBM is a very cool computing component but, in the end, if there no parts available to manufacturers it is of little use to the consumer. AMD made the decision to move in this direction hoping that the innovation and potential performance aspect of it would be enough to win back market share. The downside is that the type of consumer that will pay the premium for a top-end GPU is not likely to wait around until they can get one especially when there are other options that come close to the same performance readily available on the market.
AMD is on the right track, but their efforts might be a little misguided in the rush to push out HBM based products. We do hope that they are able to work something out with their GPUs (and CPUs for that matter) to regain market share and revenue. AMD has some talented engineers on staff, but without the money to run R&D properly they always seem to be one step behind or pushing out one off products that do not have the same return as having a solid full product line would.