hotaru.hino
Glorious
If we're talking about cinematic quality real-time 3D graphics, I would argue in some instances, we've been there and without the help of ray tracing (mostly because artists were using a lighting model that's physically based). You can get people to look at a game render and think it's real life footage without batting an eye.Actually you're both right. GPU capabilities such Ray Tracing DLSS along with faster and newer techniques should continue. However, Ray Tracing began with CGI Films using tons of RAM CPUs and now GPUs on servers. Realtime CGI is still years away from achievement if not decades.
If we're talking about real-time path tracing... well again we're already there. Quake 2 RTX and Minecraft RTX use fully path traced renderers. Now to achieve that without DLSS or whatever to clean up the noise, sure, it'll take a while, but what we have is good enough. And that's what we're used over the decades.
The problem with multi-GPU setups is they only add one thing: more GPU compute performance. That's the only pro. The cons are many:Multi GPUs are key and should not have been abandoned. Diminishing returns can be worked out with different methods but the resistance to change the current gaming platform will be much more costly than trying and keep trying different things. That's how success is achieved.
- Multiplies the cost of a video card setup (and in the current pricing climate...)
- Requires a beefy PSU to even run it
- Needs a good cooling setup to keep the thing from coming down in flames (so good fans, good case that promotes healthy airflow, or worst case, you'll have to resort to water cooling)
- May require a more expensive motherboard to get the most out of it
- VRAM doesn't combine
- Issues with load balancing the rendering
- Issues with additional complexity to divvy up the work
- With gamers demanding 120+ FPS, any amount of additional latency is going to add up and hurt