Most places use the built-in benchmark for CP77, because it's easier, and the FPS is definitely higher — more GPU limited, less CPU limited. We are using a manual run through a portion of Night City right outside one of your apartments, the number 10 building I think? And we run across a crowded street toward Tom's Diner (no relation... LOL)
But it is still very much a GPU limited game, just like Dragon Age. What you're showing proves that DAVG with RT performance drops a ton, and that similarly the gap between various CPUs narrows a lot. That is the opposite of a CPU limited game. It is in fact the very definition of a GPU limited game. Why is your CPU maxed out, then? Probably because the programming behind DAVG with RT is less than optimal would be my guess.
FWIW, I'm seeing 84% GPU utilization on a 4090 at 1080p medium (no upscaling), 89% at 1080p ultra (without the "selective" RT effects), and 98% at 1440p ultra. The first of those indicates at least a modest CPU bottleneck — and not surprisingly, I see the 4090 and 4080 Super only separated by 5%. At 1080p ultra, the separation between those two GPUs grows to 10%, so still very CPU limited. 1440p ultra, there's a 19% gap, and 4K ultra gives a 25% gap. 4K ultra with RT maxed out, the 4090 ends up 28% faster. That's basically GPU limited. The 4080 Super is 92% utilized at 1080p medium, and 97% at 1080p ultra, with 99% at everything above that.
To prove DAVG is not GPU limited, as others noted, would require testing with a slower GPU like a 4080 and showing similar numbers to the 4090, across a variety of CPUs.
Incidentally, one of the things that can make RT games CPU heavy (heavier at least) is if they do more complex BVH builds on the CPU using a single thread (or maybe two?), which is what Diablo IV does. Diablo IV has a terrible RT implementation, though, and no one should look to it as an example of how to do RT properly.
Facebook
Twitter
Instagram