Why have graphics stagnated for almost 20 years?

Soldato
Joined
16 Aug 2009
Posts
7,747
Agreed. 2004 was very good also (Far Cry, Doom 3, HL2) but those games are starting to age a bit now.
I think the main differences are textures are typically higher resolution now (a lot more VRAM) and you've got ray tracing to throw in the mix. Plus just generally more detail due to enhanced processing power.

When you try games from 20 years ago you realise just how bad they looked visually.

Its mainly textures and higher resolution, things look shinier and more detailed but on the whole its incremental rather than revolutionary those who remember the change from 2D sprites in the 80's to 3D wireframe and then fully skinned 3D models in the 90's... which went to high poly 3D models by 2005 and... not a lot a lot since, not really only higher resolution, more detail and better lighting. The next big shift looks like ray tracing but its long way off becoming the main method of rendering its still only tinkering around the edges. Moore's Law is dead or at least in a care home.
 
Soldato
Joined
14 Aug 2009
Posts
2,772
Games are still pushing on in the visuals department, it's just more subtle. More polygons aren't going to offer more at this point. It's animation, lighting and shaders that are where the gains are being made.

Horizon Forbidden West has pushed human faces agonisingly close to out of the uncanny valley. The skin shaders in particular are incredible. With a bit more advancement in the facial animation, human features will be nailed.

AI is an interesting one (mainly because I am a game AI programmer :p)...relatively easy to make genuinely 'smart' and 'clever' AI. For a shooter you could use neural nets and let the AI run around in simulation and get extremely good at killing players and exploiting every single flaw in your engine....not much fun to play though, and there's little scope for tuning.

Making game AI feel 'smart' is actually mostly about smoke and mirrors, rather than it actually making smart decisions. Animation and audio tics that help the player read what the agent is doing and why, etc.

Wouldn't Nanite from UE5 invalidate your point about polygons? Not to mention lots of assets at high poly count, close, but also far away, when you have more of the gameplay outside, in real open spaces
 
Soldato
Joined
20 Dec 2004
Posts
15,838
Wouldn't Nanite from UE5 invalidate your point about polygons? Not to mention lots of assets at high poly count, close, but also far away, when you have more of the gameplay outside, in real open spaces

Nanite is mainly addressing the big challenge in game dev these days, which is the content pipeline. There's very little visual fidelity to be gained from more polygons beyond a certain point. Polygons are just one relatively minor input, it's the shaders and lighting that make the visuals.
 
Man of Honour
Joined
13 Oct 2006
Posts
91,112
AI is an interesting one (mainly because I am a game AI programmer :p)...relatively easy to make genuinely 'smart' and 'clever' AI. For a shooter you could use neural nets and let the AI run around in simulation and get extremely good at killing players and exploiting every single flaw in your engine....not much fun to play though, and there's little scope for tuning.

Making game AI feel 'smart' is actually mostly about smoke and mirrors, rather than it actually making smart decisions. Animation and audio tics that help the player read what the agent is doing and why, etc.

Yeah AI is as much an art as it is a science. While I can understand the reluctance from a quality assurance/testing perspective I find a lot of game developers are too shy of using small amounts of random in AI programming to simulate fuzziness of behaviour which in a lot of cases fakes up a more "alive" AI than can be achieved with the limitations of neural nets, etc. in current tech.

For instance with Quake 2's AI which is incredibly limited - not having the time to rip it all out and reimplement I just added a few randoms to various parts and now instead of always running down one path, often then getting stuck and not making it to the player, half the time they do that, and the other half is split between accidentally making a smart decision and accidentally making a dumb one but to the player on balance it appears like the AI has some degree of path finding despite it doesn't at all.
 
Soldato
Joined
17 Jan 2016
Posts
8,768
Location
Oldham
I would say a lot of the reason is down to them using the same game engine.

I know only recently some games are being programmed in the most recent game engines, UE5 for example.

I'd say the reason for the stagnation is people get used to one system and think they can squeeze the last drops out of it rather than start afresh on a new system. Thankfully this is changing. Maybe a generational training thing that newer people are brought up learning newer languages.

Being an old timer game though, graphics shouldnt be the main thing we look for in a game. Gameplay should be king. I'd rather have low quality graphics in a game that is playable over and over than some good looking game that has no re-playability.
 
Soldato
Joined
14 Aug 2009
Posts
2,772
Nanite is mainly addressing the big challenge in game dev these days, which is the content pipeline. There's very little visual fidelity to be gained from more polygons beyond a certain point. Polygons are just one relatively minor input, it's the shaders and lighting that make the visuals.

I agree with you, after a certain point there isn't much difference between a higher and lower polygon asset (over tessellated flat surfaces in Crysis 2 comes to mind). Is just that we're not there yet. There are plenty of assets, even in newer games, that ar rather low poly, easier to see where something needs to be round. Poly count alongside lighting, shadows, etc., they all come together to create a nice looking image.

With that said, stuff like the sound system hasn't really changed much. AMD hasn't pushed things further with True Audio and probably it won't change in the near future as well.

Game design (gameplay stuff), still has elements from many, many years ago. You're still limited to X pieces of object Y arbitrary, because otherwise the developers would actually need to create a workable economy of sorts and AI routines to take into account these changes. "Police" AI still spawns around you in open world games (sometimes more obvious than others), you can still go nuts and kill people, hide and everyone forgets about your deeds, etc. Bullets just do magically more damage thanks to 1 point assigned to a skill or just because you're crouched and the other guy hasn't seen you and so on...

Funny thing is, graphics is one of the places where things actually improved.
 

V F

V F

Soldato
Joined
13 Aug 2003
Posts
21,184
Location
UK
I would say a lot of the reason is down to them using the same game engine.

I know only recently some games are being programmed in the most recent game engines, UE5 for example.

I'd say the reason for the stagnation is people get used to one system and think they can squeeze the last drops out of it rather than start afresh on a new system. Thankfully this is changing. Maybe a generational training thing that newer people are brought up learning newer languages.

Being an old timer game though, graphics shouldnt be the main thing we look for in a game. Gameplay should be king. I'd rather have low quality graphics in a game that is playable over and over than some good looking game that has no re-playability.

In a way Assetto Corsa/Competizione is like this. The graphics are great but will never be as polished as Gran Turismo but it has stellar physics for that experience feeling. More so the force feedback. Which is why so many people keep playing it.

Where Gran Turismo becomes stale and boring very quickly.
 
Back
Top Bottom