renz496
Champion
finally the architecture detail is coming out. to me this is more interesting that seeing the RTX series performance in games haha.
some of TPU summary
AMD probably did not like nvidia to lead what next graphical feature to be adopted by game developer but MS and Sony probably will force AMD to come up with something similar to nvidia RTX (especially sony).
hmm typical nvidia? in general i think this is the main principle for nvidia when they design their GPU. design a feature that easy to take advantage of by game developer to the point developer does not need to support anything specific to reap the benefit. it reminds me of AMD and nvidia approach in increasing GPU utilization in games. AMD push async compute to increase their utilization while nvidia make more significant architecture changes to improve architecture overall performance (fropm kepler to maxwell). the end result was AMD specifically need low level API (Mantle/DX12/Vulkan) before they can use async compute while pretty much any existing games can reap the benefit from maxwell design without the need to support specific API.
some of TPU summary
Raytracing adoption will also heavily depend on developers and publishers. It's great to see several API standards for raytracing, backed by all the major industry players. In my opinion the deciding factor will be adoption in consoles, though. Most titles today are developed for consoles first, which run on AMD hardware, that have no support for raytracing at this time. No doubt AMD, Sony and Microsoft, are working hard to bring raytracing capabilities to consoles, which would intensify developer effort and convince publishers to allocate significant money and development time for these features.
AMD probably did not like nvidia to lead what next graphical feature to be adopted by game developer but MS and Sony probably will force AMD to come up with something similar to nvidia RTX (especially sony).
The shader engine has been upgraded to concurrently support execution of integer and floating point instructions, which will give a nice boost in every single game, without the game developer having to make any code changes.
hmm typical nvidia? in general i think this is the main principle for nvidia when they design their GPU. design a feature that easy to take advantage of by game developer to the point developer does not need to support anything specific to reap the benefit. it reminds me of AMD and nvidia approach in increasing GPU utilization in games. AMD push async compute to increase their utilization while nvidia make more significant architecture changes to improve architecture overall performance (fropm kepler to maxwell). the end result was AMD specifically need low level API (Mantle/DX12/Vulkan) before they can use async compute while pretty much any existing games can reap the benefit from maxwell design without the need to support specific API.