Page 3 of 3

Re: Path tracing in games

Posted: Mon Sep 24, 2018 12:04 pm
by cignox1
I think that many of the ideas expressed here several years ago could be now revisited with the new RTX in mind.
For example, the complexity level reached by rasterizazion engines is something that has been discussed a lot recently. Or the hybrid approach that I considered in the opening message seems to be the path choosen by nVidia to introduce RT in games.

What do you think about it?

Re: Path tracing in games

Posted: Tue Mar 26, 2019 6:25 am
by andersll
Things seem to be progressing the exact same way we did in the offline world: start introducing raytracing to capture certain effects that are difficult to get looking right with rasterization techniques, then when the hardware catches up move to full path tracing and do away with the massive ball of complexity that's caused by all the hacks upon hacks to try and recreate stuff that just falls out of the rendering equation naturally.

Might take a little longer for games though - in feature work render times have gotten significantly longer since the shift to path tracing but we put up with it because the results are better and things generally require less artist time to reach a certain level of quality. For games you've got a hard frame time limit to hit.

Re: Path tracing in games

Posted: Tue Mar 26, 2019 8:08 am
by cignox1
Things seem to be progressing the exact same way we did in the offline world
Yes, that's what I think it is happening (and will happen with some luck = if nVidia don't give up due to the warm reception).
Might take a little longer for games though - in feature work render times have gotten significantly longer since the shift to path tracing but we put up with it because the results are better and things generally require less artist time to reach a certain level of quality. For games you've got a hard frame time limit to hit.
And on the top of that, in the offline world you can "just" upgrade your hw so that it is up to the task for your next feature film, but you can't force 10 millions players to upgrade their hw 2 or 3 times in a few years. And since you still need to keep your "old" rasterization code unless your target players already have capable hw, many years might be required before we see the first RT only AAA game (but I bet we will see some experimental/indy games in the meantime).

Re: Path tracing in games

Posted: Tue Mar 26, 2019 1:33 pm
by friedlinguini
I think the hope is that it will follow roughly the same trajectory as programmable shading, which was also once an extravagant feature from the offline world. Unlike with Cg, I think there’s a bit more of a push for multi-vendor adoption of the interface as seen from getting it in DirectX early, along with various game engines.

Disclaimer: NVIDIA employee expressing my own opinion, and not privy to the meetings where this might have been discussed.

Re: Path tracing in games

Posted: Wed Jun 26, 2019 2:45 pm
by capagot
Interesting discussion. As friedlinguini has pointed out, it seems that rendering, for interactive applications, is following the same path as in the movie industry. Also, I agree that, as pointed out by cignox, it may not be that easy due the hardest constraints imposed by the gaming industry (frame time and the heterogeneous hardware at the hands of the consumers, which can not buy new hardware at will).

However, it would be interesting to consider into this scenario the possibility of game-streaming (e.g. Google Stadia), which would reduce the hardware cost for the end user. In this scenario, the game-streamers would be in charge of maintaining and upgrading its own hardware. It would be a situation similar to that of the movie industry (one company responsible for the software and hardware) except that, in this case, the rendering demands would be *much* (more emphasis here!!!) higher because the rendering farm would have to render every frame of every game for every (potentially thousands) user.

At a first glance it may suggest that, given the higher rendering power available at the rendering farms, the evolution of the rendering engines towards pure ray tracing would be faster. However, every small increase in the computational demands of the rendering engines must be scaled by thousands of game players which would, in turn, keep the evolution of the rendering engines as slow as it would be if the consumer would be the responsible for his own rendering hardware. Certainly these distributed rendering engines will take advantage of coherence in order to reduce computational demands, by I have no idea how well that would scale.

All in all, I am really curious and excited about what is going to happen, but I am really bad at predicting :)