We did a lot of analysis of what it would really mean to run a game at 4K with 60fps and then to scale that down to 1440p at 60fps.
The reality is you don’t need as much memory bandwidth because you’re not loading the highest level MIP levels into memory. You don’t need the same amount of memory as well.
Developers have a whole host of different techniques, whether that’s changing the resolution of their title, things like dynamic resolution scaling frame to frame — that’s something we’ve seen a lot of adoption of, especially towards the end of this generation.
And obviously the ability to enable and display different visual effects, without actually implementing the fundamental gameplay.
With RDNA 2 we get basically a 25 percent performance uplift over GCN with no work by developers at all.
There’s a significant amount of efficiency we’re getting out of RDNA2 relative to GCN. Then we look at other things like using float 16 or variable rate shading, and we’re seeing on the order of 10-20 percent performance benefits from there as well.
It is really difficult to compare raw numbers like teraflops across generations, because we think about them differently.
In general on the GPU side the Xbox Series S is effectively the same performance as an Xbox One X GPU, but it brings all the next-gen features like ray tracing, VRS, mesh shaders, and obviously when you look at the massive leaps in CPU performance and I/O performance, that’s why Xbox Series S is designed to deliver that true next-gen experience just targeting a lower resolution than Xbox Series X.
There are also opportunities where we can enhance the titles on Xbox Series S even further than what we can do on Xbox One X.
If you look at the raw power of the Xbox Series S, if a title wants to go in and double its frame rates it’s actually really easy, because we’ve more than doubled the GPU performance and more than doubled the CPU performance, so it’s relatively easy for a developer to go in and enable that if they choose to update their title.
Historically, console generations have really been defined by how games look.
This generation a lot of it is going to be how do they feel? How do they play? When you think about these large open worlds and keeping players immersed in it, I don’t want pop in, I don’t want load times.
When we think about the Xbox velocity architecture or sampler feedback streaming, those are areas where we expect a lot of innovation over the generation.
With something like sampler feedback streaming, it can deliver performance well beyond the raw hardware specs itself. So much of this generation is about efficiency.