Hacker News Reader

Comment by hypertele-Xii | Hacker News Reader

Historic "next generation", not future prediction, as one might expect from the title. Detailing the evolution of 3D graphics 1984-1996 from Elite up to Quake.

View on HN

An incredible 12 years of progress. Looking back 12 years at 3D graphics from 2011, and… they’re kind of the same as we have now.

One reason is that up to quake almost every game had its own engine, hard coded by experts in pure graphics programming.

After halflife and unreal the game engines became standardised (later of course came unity etc.) and the focus of selling games turned to graphic style and storyline (and often multiplayer gameplay) over the pure mechanics.

As with most things in the beginning a lot of low hanging fruit allow for rapid innovation and progress. As things become more developed, progress tends to slow. At that time both hardware and software made rapid progress.

I think currently the big change that is happening is path based ray tracing. In the next couple years a lot of mainstream hardware should be able to do it and it makes a big impact on how games look.

Apart from that it's just more triangles I guess. We are approaching photorealistic looks and further development will try to run what huge desktop graphics cards can do, on smartphone GPUs. There we have seen actually significant developments in the past 10 years.

This is mostly due to the last console generation getting dragged out for so long. The PS4 came out in 2013, but thanks to a small intermediate upgrade and extended supply chain issues for the follow-up generation, by late 2022 you still had absolute high end games being made with decade old hardware in mind.

But what would you think will be the next big step after some of the current hyper-realistic visuals made possible by modern game engines?

AI procedurally generated content. AI npc's. And even AI supported render engines. Maybe at some point complex shader logic can instead be generated by some AI.

The recent Unrecord demo [0] looks amazingly realistic at parts, but it shows what I think we're still lacking, realistic non-mo-capped human visuals. To my eyes anyway, there's still something very fake about the way characters move their bodies, their eye saccades, their mouths, the way they breathe and tumble, at least in terms of what can be achieved by realtime game physics. Certainly given enough render time, studios can pretty convincingly fool the eye even now.

[0] https://www.youtube.com/watch?v=5qvVNzsJyB0

Anything related to Collision Detection and physics at large. That part of graphics is still at very early stages in realtime context.

Yes, turns out the things that greatly improves graphics rendering within non-exponential algorithmic where discovered quite fast. Either someone comes with a unconventional breakthrough, or we're stuck with heuristics optimizations and FSR/DSSL.

Are you referring to progress in rendering techniques or in the final visual result? At least in terms of the latter, games from 2011 look much cruder to me compared to games released today or even a few years ago.

Crysis with 2011 era mods looks about as good as a large chunk of modern AAA games, and even better in some cases such as compared to FF Strangers of Paradise.

Both real time ray tracing introduced in 2018, and real-time denoising with super-sampling / up-scaling / frame-interpolation, both of these have made meaningful visual quality improvements in the last 12 years. If you do a Google image search for “best games with ray tracing” and compare to “best looking games from 2011”, there is an obvious and sizeable gap in quality.

It’s still nothing like what we saw from say 2000-2010. I’m not sure exactly where the diminishing returns are given how much more powerful hardware is relative to a decade ago, been graphics development really has been disappointing. Maybe the explanation has a lot to do with training programmers on certain platforms that become outdated, or games with long development cycles being unable to take advantage of the tech that actually exists when they release.

I think games and graphics have some pretty inherent n^2 scaling which can make progress appear to have slowed down, despite large increases in computational performance.

Say your textures are 100x100px for example. If you double the size of your texture to 200x200px suddenly you have 4x as many pixels to deal with.

Take a look at Unrecord, for the difference from 2010, photo realism has come a long way

https://www.polygon.com/23691482/unrecord-bodycam-fps-steam-...

I think there does in fact exist a recent major breakthrough in real time rendering: Virtualized geometry.

https://advances.realtimerendering.com/s2021/Karis_Nanite_SI...

https://youtube.com/watch?v=eviSykqSUUw

It was invented by Brian Karis at Epic. Virtualized geometry allows for unbounded geometric detail while retaining the conventional rasterization approach with polygons and textures. It works like a dynamic LOD, where the game engine dynamically adapts the rendered geometric detail based on visibility and screen resolution. Essentially the maximum rendering cost only depends on resolution and frame rate, not on the detail of the underlying meshes.

However, the solution by Epic, dubbed Nanite, does not yet work for animated meshes. So virtualized geometry is not possible e.g. for game characters. Brian Karis says this is a solvable problem though.

Thank you. I showed up ready to argue that graphics plateau’d some time ago, and that unless you intend to use ML to make it look more real, it’s only going to look more detailed.

Nah, it's lighting because that's the most important thing. When the hardware is ready (at least 10 years).

Games have been adding more detailed textures and more geometry because that's all they could do. Textures just require RAM

Lightning and shaders.

Reply on news.ycombinator.com