* * If you predicted how future games engines improve...

[g__day]

What would your main thoughts be?

I guess this comes in a few parts, because we sometimes spectulate on what, where and how fast 3d hardware is going. But what of the game engines, like Doom 3, Source, Cry-Engine, Krass or X-Ray?

How to we rank the best game engines and where do we see they are heading feature wise? What will we see in the future?

I assume a game engine has several components nowadays, multiplayer LAN management sub-engine, physics and AI sub-engines, sound sub-engine and of course graphics sub-engine.

Nowadays the LAN and sound engines are pretty good and even the graphics engines can scale from DX 7 -> DX9.

So where are things headed? If you had to predict the best innovations we will see for say 3 years out, what will they be?

 

[Polarbear53]

I hope more things on screen at once, with larger levels, with better graphics. That will make me happy.

 

[nutball]

So where are things headed? If you had to predict the best innovations we will see for say 3 years out, what will they be?

I know where I'd *want* them to be -- AI. The eye candy in the current state-of-the-art engines is pretty impressive, and the UE3 screenies almost qualify as Underpant Disturbing.

But if those pretty monsters are moping about, walking into walls and generally being a bit local, it kinda spoils the whole experience really, dunnit?

 

[RejZoR]

First thing would be to get rid of old rendering style with polygons (triangles).
Increasing the number of polys to make curved surfaces won't last forever.
Also different filtering methods than Bilinear/Trilinear. One such method was presented few years ago with some SiS Xabre tech demo. Don't know where is now...

 

[bloodbob]

First thing would be to get rid of old rendering style with polygons (triangles).
Increasing the number of polys to make curved surfaces won't last forever.
Also different filtering methods than Bilinear/Trilinear. One such method was presented few years ago with some SiS Xabre tech demo. Don't know where is now...

Umm thats a hardware issue and I'm guessing its probably doing less filtering then trilinear so thats making the graphics worse.

 

[Scali]

Well, I suppose if we look at Hollywood's offline rendering, that's a good indication of where games will be going.

And in general, I don't think we are that far behind anymore. There are only two main differences, and that is a REYES rendering architecture instead of triangle rasterization, and the use of shadowmaps.
The first is not really a big deal, I suppose, because they can get similar results. The second is just around the corner. We are already seeing shadowmaps being used, with softshadows and everything.

Apart from that, I guess the differences are just in the scale. Movies have shaders for pretty much every material, the shaders are more complex than those used in games, the resolution is higher, the polycount is higher, etc.

Oh, and another thing is that sometimes raytracing is used for special effects. We can already raytrace on current hardware, but it's still very slow and primitive.

So the way I see it, we're well on track with engines like UE3.0, and the hardware mainly has to get bigger and faster. Other than that the hardware may eventually move to a REYES architecture, I don't know how attractive that will be in a hardware implementation. And there may be better support for raytracing in the future.

 

[jpr27]

I know where I'd *want* them to be -- AI. The eye candy in the current state-of-the-art engines is pretty impressive, and the UE3 screenies almost qualify as Underpant Disturbing.

nutball is right on with this. AI should start evolving just like graphics has. I believe M$ is making a big push with future development AI in its next generation of gaming (XNA ?). Im not sure how extensive it is but any improvement on AI focus would be a welcome one.



*** There is no luck...Only the will and desire to succeed ! ***

 

[Diplo]

The successful game engines of the future will be the ones that are flexible and have the best content-creation tools (editors, material constructors, landscape generators, physics, plug-in-integration with major modelling packages, extensible object-orientated scripting languages, cross-platform compatibility, efficient net-code). Game assets are increasingly become the major factor in how long games take to develop which in turn reflects on their development cost. The engines that makes this the easiest and can target multiple platforms (especially consoles) will emerge as the winners, as these will allow maximum profitablity.

It's no use having an amazing looking engine if the level editor is buggy and unintuitive and you need programmers to construct materials. It's no use having the potential for complex AI if there is no high-level language available to make this easy to utilise. There is far more to a successful engine than flashy graphics.

 

[RejZoR]

Thats why Unreal Engine is so successful...

 

[London Geezer]

We need to get our heads out of our asses and stop worrying about how many pixels per second per frame per poly per texture we can push, and start animating stuff properly.
Until things move like they're supposed to move in real life, a game can look as photorealistic as u like, but it won't ever be realistic.

 

[The549]

We need to get our heads out of our asses and stop worrying about how many pixels per second per frame per poly per texture we can push, and start animating stuff properly.
Until things move like they're supposed to move in real life, a game can look as photorealistic as u like, but it won't ever be realistic.

Yeah, I think that there are certain things that make a gamer feel like he or she is actually in the world of the game.....things like ai, interacting w/ environment, remembering bullet marks etc, mouths moving like something realistic, people moving without gliding etc is far more inmportant. I think that polys will come with all of that.
And yeah Doom, Unreal 3, some of Far Cry and a lot of HL2 are making me go . Love the big levels in Joint ops too, like 4 miles long or something.

 

[Cyberdigitus]

We need to get our heads out of our asses and stop worrying about how many pixels per second per frame per poly per texture we can push, and start animating stuff properly.
Until things move like they're supposed to move in real life, a game can look as photorealistic as u like, but it won't ever be realistic.

Indeed. I have a feeling we will already experience this in hl2. While it has relatively good animation and ai calling upon it; watching some video's revealed some choppy reactions. this comes more and more obvious when the visuals get better.

It's not only about better animation loops though, also how they are blended and called from the logic, there will have to be much more inbetween animations or ik.

generally i'm with Diplo, the tech isn't that important anymore. actually this animation issue is an example of it.

 

[NeoCool]

TBPH, IMO, I think we need in the future, engines that make the DooM³ engine's architecture, tech and code look laughably simplistic.

And yes, granted this means a true 100x100 perfectly mapped HDR rendering pipeline, true per-pixel accurate soft shadows and a real unified lighting model system based on the laws of area lighting, radiosity lighting and real specular / diffuse components all per-pixel accurate and lit. Dynamic light refraction, characterizing every pixel at accuracy would also be a plus.

We need engines that will allow game developer's work to be not only much easier - but to enable much more complex achievements.

Onto textures.....to be frank I'd like interactive, real dynamic textures(like in real life), like being able to scratch the paint off a texture and having displacement maps with at least 4096x4096 resolution textures with life like specular highlights. We also need real-time reflection and refraction maps that reflect and refract everything no matter what and look perfectly sharp and clear - that can also be characterized in cool ways by programmable pixel and vertex shaders(or fragment and vertex programs if we're talking the OpenGL renderer).

Also, if game devs are going to bother to implement realistic waves and oceans into games, we absolutely need ultra-realistic diffraction in waves. That would be surely awesome.

That's basically what I'd majorly like to see in a future engine in the visual effects category.