| vivster said: And you're telling me AMD isn't able to beat that on a smaller node without "useless" crap on the chip? The last time I checked the goal was to beat your competition and not to dally around in the lower high end. |
A smaller node is useless if they don't decide to go the full way which is approaching towards the reticle limit. I would say your constraints are pretty similar in expecting a GTX 1060 to come out on top of a GTX 980 Ti ...
I don't think you think you know what exactly AMD's goals are in mind. They only have political will to remain in the gaming market, not to dominate it like how Nvidia rolls out large specialized gaming GPUs. Technological superiority plays a role but at the end of the day what separates AMD and Nvidia is pure political will in which one will always use the bigger die. Just as gladiators live and die by their sword so too do the chips within their silicon but more specifically the size of it in many cases ...
With Radeon VII, their main objective wasn't the gaming market but it was high end compute in deep learning. The Radeon VII is the exact same chip design found in their MI50/60 line of compute cards so AMD is obviously in it to make a quick buck for it in the gaming market despite a couple of flaws like half rate FP64 which only makes power consumption worse, disabled PCIE 4.0 which is still built into the chip, and then we have deep learning specific instructions which are somewhat similar to what Nvidia calls their "Tensor Cores" except baked into the instruction set as special instructions instead of a separate unit like Nvidia's implementation ...
With AMD there are a several flaws with their approach when it comes to gaming like either not using a big die size or catering to the wrong market so without hitting these two notes properly they practically never are able to take the performance crown ...
