Luminescent said:
Scali, would you consider the Doom 3 renderer revolutionary in any way?
I read your opinion in the UE 3.0 thread and partially agree that the technical term "unified" (not Carmack's definition) does not properly describe Doom 3 for a number of technicalities. Would you agree, though, that it does present a significantly more unified approach than any of its predecessors?
Yes, but progress is only natural. Every few years we get a new generation of games with more advanced approaches in rendering, and more realism.
So Doom3 happens to be the first game with both shaders and dynamic shadows (with selfshadowing) on pretty much everything.
Big deal, the parts of the puzzle have been around for years, and have just been waiting to be put together in a game when the hardware was ready for it. I guess people are dazzled by the visual impact of shadows and bumpmaps everywhere... I don't think it's such a big deal anymore, since I did that years ago on my GF2.
I believe we are gradually growing closer to diminishing returns in terms of accurate rendering methods relative to performance and visible change. Therefore I wouldn't predict anything presenting a greater paradime shift, in terms of visual quality/accuracy relative to performance, than Doom 3 (in the short term, at least). If we want unified, according to your argument, we would need something that produces results on the level of raytracing, which I don't see as being plausible for some time;
Indeed, I have said this myself in the past, regarding hardware accelerated raytracing. We can currently get visuals that rival raytraced images, in realtime, using our rasterizer hardware.
The Quake3RT game demonstrates the point perfectly. It doesn't look anywhere near as good as Doom3, and it is about all that first-generation hardware raytracers would be capable of. A good reason to completely ignore raytracing or other 'unified' or 'elegant' solutions. If it's not better or faster, don't use it, simple as that.
not even UE 3.0 provides a robust solution for dynamic indirect lighting in comparison to its direct lighting model. That said, I would consider Doom 3 as revolutionary as can be, at this stage of the rendering game.
I don't, really. As I said before, the things that Doom3 does, have been done to death already. Doom3 may be the first *game* that does it, but the technology is by no means revolutionary anymore. In fact, because Doom3 was delayed this long, it is not even taking advantage of current high-end hardware. There's hardly any visual difference between a GeForce256 and a GeForce6. Things like parallax mapping could have been applied for high-end hardware. And as I mentioned before, using the vertexshaders for skinning/shadows could greatly improve polycount aswell.
As for dynamic indirect lighting... not even regular raytracers can handle that. You'd need additional processing, like Monte Carlo path tracing or photon mapping. The problem is just that there is no known algorithm that will allow us to render indirect light in realtime. The problem is just too complex. Don't expect it to happen anytime soon.
I would think that HalfLife2 will be more revolutionary. It actually introduces a novel way to get bumpmapped radiosity (as far as I know, it has never been used before), and it uses shaders extensively to model complex surfaces such as water, stained glass and fire. It will also use HDR lighting. The only bad thing about HalfLife2 are the shadows, I suppose. I believe it just uses projected shadows, no selfshadowing. Now if either Doom3 had the advanced shading of HalfLife2 or if HalfLife2 had a nice robust dynamic selfshadowing system, then we'd have the ultimate renderer at this time, I suppose. UE3.0 should more or less be this engine.
I find Doom3 very poor in terms of lighting. That's probably mainly because it was originally aimed at GF256-class hardware, but it was delayed far too long, without readjusting the lighting model.
HalfLife2 will actually give the shaders on more recent hardware a run for their money.
Personally I think the next big advance in rendering will be introduced by 3DMark05 (mainly filtered shadowmaps and HDR, perhaps some other eyecandy that I haven't seen yet). And after that, probably UE3.0. Let's just hope that UE3.0 doesn't make the same mistake as Doom3 of being delayed for years without updating the renderer to implement some new features.
To conclude: Doom3 is cute compared to previous games, but it is released in a time when we have the GeForce 6800, and we have no idea what it is capable of yet... a lot more than Doom3, that is for sure. 3DMark05 should give us a first glimpse of things to come.
My bad. I just played some legacy doom and that has jumping and I was pretty sure that there were some areas that you could only reach by jumping so I figured Doom had that originaly. So I fired up WinDoom just to be sure and nope, no jumping. Still, it introduced "true" 3d levels (i.e. stairs and lifts) so it is significant in that respect.
