By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Soundwave said:
sc94597 said:

I'm thinking the 10th Generation (2026->2030) will be a short one. Basically pushing decent ray tracing, and fully achieving the promises of this generation. 

The 11th Generation (2030-2037ish) probably will be when neural rendering comes into its own and will be a huge (for today's standards) generational leap somewhere between the SD ->HD leap (6th -> 7th Generation) and 2D -> 3D (4th ->5th Generation) in noticeable difference. 

Neural Rendering will likely come with new gameplay genres/types and styles. That is probably what the current generation lacks most. A genre/sub-genre that can only be done now that couldn't be done very well in the past. 

Nvidia is laying out the blueprint, but AMD and/or Intel will likely catch up by then. If they don't Nvidia will become a real monopoly that might even try to expand into non-GPU consumer processors.  If thay happens, they probably should be broken up.

Nvidia and AMD don't really give a shit about PC GPUs anymore. Sure they will make them and they will continue to improve, but the AI server market is way, way, way larger. I think in Nvidia's recent earnings report, like only 8% of their total revenue was from "gaming" and that includes the Switch + PC GPUs. 

Gaming GPUs are a small potatoes business already. 

This is what essentially neural rendering (as a workload) solves. How can you maximally utilize hardware that is designed with specialized modules for GPGPU compute for gaming? 

Nvidia and AMD might not care about gaming enough to prioritize it (although that doesn't mean they can't also target that market), but Nintendo, Microsoft and Sony do, and Nvidia and AMD are happy enough to give them solutions that are useful for their needs.

It only looks like Nvidia is abandoning the market when one takes a narrow view of what gaming workloads can look like, by basically assuming the current methods of rendering 3D graphics are what should exist in the future and hardware needs to be designed for these workloads specifically.