HalfLife 2 engine and Doom 3 engine comparison

Which would indicate a actual limitation in the engine.
Anyone wanting to create large levels with huge (or even average) draw distances would have to implement their own LOD mechanism.

It's okay for people to suggest that doom3 has massive draw distance capabilities in excess of anything that Source or CryEngine can achieve, but if you can't actually use it without modifying the engine yourself, then it's a fairly pointless thing to say.
If it can't realisticly do it without modification, then it's not a capability of the engine.

 

Don't worry, I'm sure we'll get to this stage were the Valve fans will force Gabrobot to add enough stuff to the map that it comes to a sufficient crawl that they feel vindicated...

 

I suppose the arguement for or against each engine (including the CryEngine) is flawed anyway.

Each engine is more or less capable than the othes in certain aspects. Trying to argue over which one is technically superior is futile because they all excel in different fields. You'll allways get doom3 fans who say "The doom3 engine is the best! Period!", or Source fans who say the opposite, but in reality, each engine is equally as impressive, but in different ways.

Doom3 engine is good at smaller darker environments which show off its shadow and lighting capabilities. It's probably capable of larger enviroments with modification, but these won't show off the lighting system for which it was designed, so it's unlikely to be used for anything other than doom3 style games. Not a bad thing, since there's plenty of games around like that could use this.
Developers that want to create this type of game will use this engine.

CryEngine is good at vast vistas which show off its impressive LOD system. It also has decent shader usage and a built in Havok physics system which can be put to good use (and yes, before you all jump on me, I know the doom3 engine has a decent physics system too).
Game developers that want to create these sort of large environments will use this engine.

The Source engine seems (to me anyway) to be a decent, scaleable, general purpose engine that doesn't nessiscarily excel in any one area, but is reasonably competent at a wide range of areas.
The majority of developers will probably use this engine because of its general competence and scaleability.

The other factor as far as developer choice will of course be pricing, developer tools and support.

I think all 3 engines are excellent and will help supply us gamers with some great gaming experiences for a good few years. I look forward to seeing what other developers can create using them

 

My stance on Doom3, Source, and CryEngine is that Source is most likely the most intelligent choice for todays games...large environments for expansive gameplay, decent shadows, lightmaps, which look awesome and are very cheap, and overall source runs well.

Doom 3 is a brute force, raw technology engine. It is the most advanced on a technical side, but that doesnt make it better necessarily. Yes, I am saying the Doom 3 Renderer (Not overall engine) is the most technically advanced, however for many games, do we really need real time lighting and shadowing? Do we really need to move away from lightmaps/shadow maps (not shadow buffers to avoid confusion)? I think when hardware is right and the time is right Doom 3's engine will show its true colors, and Doom3 "barely scratched the surface" according to id. But for now, Im glad games like HL2 are using lightmaps for cheap, great looking environments.

CryEngine in my opinion, is not as advanced as the Doom3 engine on a technical side.. it doesnt use real time lighting/shadowing, so lighting is relatively cheap, allowing developers to focus elsewhere, such as 1.2Km draw distance in an actual game, and beautiful tropical islands. Along with good character models and such.

I'd say the renderer is a tad more advanced than half life 2 on a technical side.

Anyway, Doom3's engine is "more technically advanced" but it doesnt make it better than HL2 or CryEngine. We don't really have a need for real time lighting on todays hardware if we have to sacrifice texture quality and polygon counts.

But I respect Doom 3 as a great technical achievement....

 

You know I think that his hacked together level actually looked kind of cool. If you were wandering through a forest of pillars it would seem very alien, I mean to make it go well you need more random crap around to further illustrate the contrast to the perfect pillars. (I mean rocks and grass and so forth) But I think something that simple could be an effective environment.

 

When I say "interesting or new" in relation to the D3 engine, I'm specifically speaking of that in relation to previous id software engines.

Game engine = the core engine around which a 3d-game is built. My problem with the D3 engine is in determining whether the limits in the engine as it is used in D3 (like no outdoor rendering of large, brightly illuminated areas, for instance) exist because of the engine, or because of the limits of id's creativity in designing the D3 environments.

I inserted "set com_fixedtic -1" into my autoexec and noticed that in certain areas I could move up to a wall and face it and my fps counter would zoom right up to my refresh rate, to hover near 100fps or so (vsync on), until I moved away from the wall. This implies that the limits to rendering large exterior environments are found in the engine itself, are performance concerns, and that the D3 engine has limits similar to those in Q3 for exterior rendering (as I recall a significant frame-rate performance jump in Q3 by disabling the "sky" textures--which was in most cases fine since you couldn't see them most of the time anyway.) As I say, until I can see the engine used by another developer it'll be difficult for me to discern which of the environment limits in D3 are contrived by the game design, and which of those limits were considered necessary by id because of limits inherent in the engine itself.

Just guessing I'd say that the limits in the D3 rendering environments are closely constrained by performance concerns involving the engine itself. But perhaps in the hands of another developer not so preoccupied with shadows and lighting extremes the engine might do much better with exterior environments--that aspect of engine performance capability just doesn't seem clear from looking at D3 itself. Yes, I was talking primarily of the rendering engine and how its limits seem to dictate the design of the D3 game environments.

 

To be honest the majority of developers will probably license the engine that is the most cost-effective and has the best support and content-creation tools.

 

You mean developers won't take into consideration the valid opinion of Beyond3Drs?

 

Ylandro, I mostly agree with your assessment. However, it seems that you have bought Valve's marketing sh*te regarding 'real-time radiosity'.

If they had real-time radiosity - which is currently not feasible - then they would have perfect soft shadows automatically and there would be no need for the cheap fake shadows we've seen in the movies. Valve's GDC presentation of HL2 shading has been linked several times already in this thread; have you looked at it?

 

Some input from what is evident from playing CS:Source :

• Source's dynamic shadows are the glaring weakness in the engine compared to Doom 3. They do cast fairly well on changing surfaces, and are cast by every moveable object I recall, but the fixed nature doesn't allow them to play a significant role in the "life" of the scene AFAICS.
  I couldn't find any way to increase the amount of light sources casting shadows (or a way to have the flashlight interact with the dynamic shadows at all), but the shadow casting direction does seem at least to be dynamic.
• The surfaces in the updated de_dust seem to all have bump mapping effects (except the wood and baskets, I think, and the "physics toys" strewn about), though much more subtle than the "ant lion/roof top" video might lead you to expect.
  The only surface where this is prominent without a flashlight is the tile surface inside the central area, which also has a bit of a shine (at least the shine is believable, though dust covering in spots would help I think).
  Where the engine stands out is in making such a surface believably interact with footsteps, lighting, and bullet impacts.
• The physics interaction with the net code is something that impressed me, and seems to work well for large player count servers. It doesn't seem as lag free as the prior netcode (the physics inclusion would have to be "free"), but with all the physics events synchronized for the players I found it surprisingly efficient. There are a LOT of "physics toys", including trash, cement blocks, basket lids, tires, and, the most tactically prominent outside of players and weapons, rusty barrels.

My own impression is that Source is aiming for realism (as in, really trying to give the illusion of REALITY), and succeeds very well in lighting, physics, and material interaction, is "typical" in sound positioning but succeeds in other aspects of "soundscaping", and is quite limited in dynamic shadowing behavior and presentation (I'm curious if there are any plans to address this moving forward, since it is such a prominent lack even without comparison to Doom 3, given how well other things seem implemented).

...

Briefly commenting on Doom 3, the only place realism seems to have been the target was in the shadow behavior (not the graphic presentation of the shadows) and some advanced shader touches (heat haze, flames). The rest of the graphics weren't aiming for realism, they were aiming for detail in an "off reality" style (though, to the confusion of comparing to Source, this seems to be the reviewer definition of "realism").
Actually, this is primarily in comparison to Valve, who have stood out in attempting more nuanced atmospheric lighting since the original HL...though Doom 3 the game doesn't use sophisticated lighting much (lighting, not shadowing), I don't think there is a lack in the ENGINE in comparison to something like the current UnrealEngine or its dynamic shadowing modified engines, which also strike me as lacking from my own exposure to the demos of the applicable games (of course, UE 3.0 seems on track to directly address this).

...

To compare some other engines, the Crytek engine seems to be similar in lighting sophistication to other engines (behind Source in my impression), though that's because I attribute the successful tropical daylight/sunset/night time atmosphere to artistic implementation since I found the model lighting unrealistic. Also, it seems ahead of everything (as far as licensed engines) in geometry proliferation (or shortcuts for simulating them). I do, however, have a faint impression of the current UE having flexibility in this regard that I just haven't seen used in a game or demo, and Source's capabilities aren't evident yet (nor Doom 3's for that matter, but focus on that would seem an odd fit for its intent).

I also don't know the extent of the dynamic object shadowing functionality implemented in the Crytek engine by the upcoming/recalled patches, but in 1.1 it is at "typical levels" like Source outside of some isolated (though pretty) dynamic light movement drop-ins (like at the end of the Fort).

...

BTW, for the most dynamic aspect of the rendering, the description of the model shader for Source begins around page 57. It doesn't seem to propose real-time radiosity lighting calculations, but real-time usage of stored radiosity information (3 directional components) to models based on interaction with their normal maps, animation/position in the scene, while also interacting with dynamic light sources and the specular lighting "sampling nodes" stored throughout scenes...?

 

Not really. What does it really have to do with the engine? It has to do with what the engine has to do, not with how the engine does stuff. An LOD system simply takes out certain objects at a certain distance...these things are things like grass and high poly trees. It's not something that would be difficult to throw into a mod once the SDK is out. Hell, considering the results of my test and the fact that a real outdoor terrain implementation would probably be designed less open (so the level designer can control the player), it'll probably be fine without an LOD system. Enemies aren't as much of a problem since you're supposed to wait and spawn them at a certain point anyway instead of spawning them all when the level starts. If I'm not mistaken, Quake III didn't have an LOD system either, but that didn't stop game developers from throwing one in themselves...it's not a big deal.

 

It seems that the Source engine is just one step ahead of the direction paved by Splinter Cell, which was the first game to offer widespread soft projected shadows, limited HDR glow effects, and other interesting shaders like night vision which offered an overbright effect. These shaders gave the game quite a realistic twist but were not at the core of its rendering technology, at least in terms of lighting/robust shadowing; SC only offered simple vertex lighting with a diffuse component.

 

SC has soft shadows? I played the demo of PT and it had sharp shadows...I have an ATI card, is that why? I heard that because the shadow system was optimized for the Xbox it wouldn't work on ATI cards and they had to make a completely different one for ATI cards...sorry, not really relevant but I'm kinda curious.

 

SCs projected shadow algorithm does some sampling and extra passes on shadows produces soft shadows, although not quite as suffisticated as Max Payne 2's. In SC1 the projected texture path was the default path for shadows, although shadow buffers (from Xbox path) could be selected for Nvidia cards. Dany Lepage explained the implementation details of SC's projected shadow algorithm in the following excerpt:

Thread available here.

Thread available here.

He indicates that HL2 will use a simplified version of SCs shadow algorithm(!) and that shadow projectors are not a robust (consistent) means of producing accurate shadows, which supports demalion's CS Source claim.

 

I guess I wasn't completely clear. I certainly don't mean to say that HL2 has "soft shadows", or realtime radiosity. As you say, that's simply impossible.
I've had the pleasure of reading that PDF long before this thread came up. If I understand it correctly, it has a pre-rendered radiosity map of some kind for each 'room', which it can use to light dynamic objects.
Adding radiosity in the lighting, even if it is just static, adds the posibility to produce very subtle lighting.

Compare slide 4 and 5 of the PDF. That's illustrating it perfectly. Using radiosity, the shadows aren't absolutely black, like in Doom3. Instead, the shadow shows still a lot of very subtle shades of darkness and detail.

For a game like thief or splinter cell, that looks more interesting than pitch black dynamic shadows. When a guard is moving in a D3 shadow, it's simply invisible. But in the Source/HL2 static shadow, as shown in the slide 5 of the PDF, he would be just visible, if you were watching long and carefully. Sounds great for a secretive game.

I'm just wondering what will happen when you move through such a room with a torch.... Can't wait to see HL2...

 

Yeah and sounds like its the exact same thing as regular radiosity from the original Quake Engine except combining in the idea of normals to handle bump mapping. It seems to be doing the exact same thing for lighting dynamics with the radiosity mapping as in the original Quake Engine (besides its also color and prolly then handling point lights as dynamics to add specular light in).

This of course makes all sense when its just the Quake 1 engine modified out of the wazoo

Edit: Really bothers me that HL2 is based on a engine that was released almost a decade ago (hehe okay 8 years as of last June).

Doom 3 engine has promise for future expanding I would say if with conversion to shadow mapping but stenciling of course has limited expandibility.

 

Right, I momentarily forgot about how programable D3's shader system is. My bad.

 

HL2's Source Engine is a brand new DirectX engine that has nothing to do with HL1's modified Quake engine.

 

*Sigh...*

Who keeps telling you that? HL1 used the modified Quake engine.
The Source engine is a completely scratch-built one; it was in development for about 3-4 years (from when HL1 was finished until September 30, 2002), after which they started developing HL2. That means that if the game is released at September 30 this year, it'll only have been in development for 2 years.

Just an interesting little bit of trivia...