Engine minor features

[Zaldron]

Lately there has been a lot of talk regarding the D3 engine. T&L, bump-mapping and some other high-end OpenGL features had been considered. But still there are some things related to the engine that would be cool if Carmack find time to implement them.
Things that would enhance the experience, for example :

1) Alpha-blended transitions between mipmaps. It would help visual quality without falling onto trilinear filtering.
2) Bump-mapping minimum distance for activation. Could look ugly sometimes, but a giant performance boost.
3) Video options could be partially split between models & map. There could be also another category for multiplayer sessions (something like STV:EF).
4) NURBS refinement iterations.
5) I don´t know if this is currently supported in Q3, but will help a LOT some kind of surface texture log, to prevent useless buffer cleaning. (don´t load textures already loaded in the previous level) Might help in loading times.
6) A particle system that understands collision against brushes. I hate smoke getting merged with the ceilling.
7) EAX!!!! A3D gives some weird problems on a Live!
8) A light algorithm as much polished as it can be. I can´t believe that real-time raytraced lightning will be fast enough if calculated from scratch. Some light compilation would be needed pre-distributing the map. And it will be a huge chunk of data, slowing loading times greatly.
9) An easy to modify maxpolys in scene var and clipping plane distance var, to adjust perfectly to every level.
10) Index of light reflection for every texture, linked bump-map. A opacity map would be great too.
11) Visportals.
12) A new brush controller, entities have some limitations.
13) Brushes would be still limited to convex forms?? This is a way to avoid self-shadowing?
14) Could be very, very difficult, but a real-time volumetric fogging system would be great. Imagine the shape of an imp blocking some intense light.
15) NURBS on models? seems hard, but it´s iD.


And now some minor things :

1) If there is a flashlight, please make it a proyector system, not like HL or DX normal-align lightmap.
2) A good looking lens flare system. Cheap way to get great graphics.
3) light decay = 1/r*r. Just like real life, no crappy linear decay with out-limits.
4) Skin modifier skeletal system. No ugly links.
5) LOD
6) A little more intuitive console. The cvarlist and cmdlist need to be modified.


Well, that´s all I can think now. If you understand every item, then I congratulate you.

Any more ideas?

[Dima]

1)Leave the perfomance boost things out! (As long as they decrease the graphic quality).
2)It will be nice if Carmack also implements pixel shaders I think he will cause it is the only technique that allows the creation of perfectly realistic shadows (together with stencil buffer).
3)I also hope that Carmack will find a way to eliminate the problem of torn borders on shadows (this is pixel shaders main problem).
4)Basically pixel shaders allow to build your own multitexture effects! YES it takes quiet a while but looks at today's tech demos that were written in a couple of days!
5)It will be also nice if Carmack implements realistic cloth,hair moving and not like in Quake 3 engined games where hair looks like a paint!
6)Some NURBS on models and enviroments will be cool also (huge graphic resources required).
7)More interactivity in levels and big outdoor areas will greatly enhance the expirience.
8)DIFFERENT EDITOR! (I am tired of Worldcraft and Qradiant).
That's all I can think of right now when I'l have more ideas I'l add them right away!
Before I finish here is a little note:
Carmack focuses on Singleplayer in Doom 3 that means that he will focus on the INCREDUBLE graphics concern and not speed concern like he did in Quake 3 (because it was multiplayer only and bandwith limited).
I hope more ideas will be posted here!

[Zaldron]

"Carmack focuses on Singleplayer in Doom 3 that means that he will focus on the INCREDUBLE graphics concern and not speed concern like he did in Quake 3 (because it was multiplayer only and bandwith limited)."


You´re right, but that´s not excuse to do things bad. Ever played Ultima IX?? No actual machine could run that game at 50-60 FPS, and that´s because it has no optimization.

All the things that can increase performance are optional, i rather prefer, for example, ranged bump-mapping instead of turning it off.

The shadow´s problem could be fixed with some kind of scan that create a progressive blur to the stencil image.

Pixel shaders can do incredible things, remember that DooM texture of merged faces?? You could animate them, and make the perfect extrusion in each one, fitting to the present lightning in the scene. A simple texture, with the help of an procedural render and a set of bump-maps could convert that lousy effect into something like the Diablo II´s Portal to Hell as seen in (Act III)

About cloths and hair, that´s made by a secondary "model-skin" implemented over the skeletal basic model. No problem here, just need some animation work.

I like BIG outdoors, there sould be some special entities that modifies the present maxpoly and clipping plane dist. to adjust to every distinctive place in the level.

A different editor, could be, if it has to be built-in, then make a port of Qradiant is a waste of time. But iD is greatly adapted to Qradiant, so, anything new won´t be too far from the main concept.

[Dima]

1)My first question: Zaldon do u use ICQ?? connect sometime it will be easier to chat than in the forum!
2)Realistic cloth and hair moving is not as easy as u think! For realitic hair moving there should be at least 30-40 bones only in the head!!

[EllipsusD]

Zaldron said:

Damn... I've gotta learn more about engines... maybe then I can keep up.

[Zaldron]

Dima said:

1)My first question: Zaldon do u use ICQ?? connect sometime it will be easier to chat than in the forum!
2)Realistic cloth and hair moving is not as easy as u think! For realitic hair moving there should be at least 30-40 bones only in the head!!

That´s the nice but tough way to do it, if that represents too much "secondary work" for iD, I believe that a simple frame animation system for the hair and cloth would be enough. Take for example Heretic II, Corvus had a lot of that stuff, and it was all framed.
You´re going a little out the road, 30-40 bones are enough to make facial expression.

About ICQ, every night, but i have disabled all the annoying sounds, that´s why i never get alerted about your presence.

[Dima]

Still look at the latest Unreal Technology: http://www.dailyradar.com/features/game_feature_page_1444_1.html
It doee allow 40 bones only in a head and a bunch of more amazing stuff!
I hope that the new Doom 3 Engine will do much more than the latest Unreal Tech!

[Zaldron]

C´mon, that´s an insult to iD. Look at Unreal I engine, yeah, it was pretty, much more than Q2. Now look at UT engine, and compare it to Q3A one. JA! that can´t stand in the same step than Camarck´s last creation. In fact, Unreal Engine has some problems that Tim Sweeny will fix in the future, i don´t know if you already read that. I can´t remember were I readed it, but it was write by John.

[Dima]

1)I am sorry about the jpeg I sent u!
I forgot that u were the one who sent it to me!
2)Carmack said that Unreal had a bad visibility alghoritms that limited the geometry and the ammount of polygons!
Tim said that it was ok for the first Unreal and for the hardware back then but for UT he said it was questionable!

[Dima]

Here are some more suggestions!
1)It would be cool if Carmack implemets 4 skinning matrices (used for vertex blending).
2)Carmack also said that he will use 8 passes of rendering or more if there will be a need to use full reflections (and there will be a need).
3)He should should use textures of 2048x2048@32bit.
4)3d texture support should be implemeted (Only the ATI RADEON support them today but when Doom 3 ships it will be an important feature since it allows the creation of complex dynamic lightmaps,as well as volumetric fog,smoke and fluids and simpllifying handling the objects with deformable geometry (GEOMOD anyone?) Example of 3d texture use:Parts of objects can be chipped or cut away without having to update textures.
5)Keyframe Interpolation is also an intresting feature to look at (Cause it provides Priority Buffer used for realistic shadow mapping from invidual light sources and it also provides 3d textures the only problem is that it's supported only by ATI RADEON!
6)Since DOT 3 BUMPMAPPING is used it will be VERY cool to have finally REAL WATER with bumpmapped surfaces utilizing diplasement maps technology!
7)He should also add versatile texture transforming methods (cubic enviroment mapping,projective texturing,etc).
8)The modelling of reflectice materials (water,metal,wood etc) should be modelled accurately (this can be accomplished with the use of pixel shaders and dot 3 bumpmapping).
9)The game will have to be T&L only to implement many of these features while keeping solid framerates.
That's all I can think off!
Gosh I am EXCITED!

[Zaldron]

just replying :

1) You´re right, and it´s not that difficult.
2) 8 passes, mmh, a voodoo5 6000 takes some advantage there, but i still prefer Dot3. As many passes as he need, no thing to argue.
3) you know how much video RAM will use an uncompressed texture of that type?? 16 MB. Pretty hard.
4) He said that there would be available in the game. But there is no need to use them for lightmaps (they aren´t used in D3), fog & fluids are important.
5) No harm, software-emulated. It isn´t too power-consuming
6) Dot3 can´t do that, you need to code an displacement subsystem. Shadowing a pixel doesn´t modify its Z-buffer value by something usefull. The only problem is that there has to be vertex manipulation. I recommend some NURBS to make water, so you can adjust iterations subdivisions.
7) Cubic can be usefull, proyected only have some meaning for some FXs.
8) A GeForce and a r_mirroralpha are enough to do that. Mmh, but if you want reflections ALTERED by bumpmapping, then you have some serious speed problems.
9) T&L only, well, that is a must have. But, hey, no problem if we were talking of something like 2002. Remember when the VooDoo1 came out? 2 years later everyone had a 3D accelerator.


there are some other things to do :

1) this is easy, bump-mapped wallmarks : scratches, bullets and explosions would make believable damage without having to do some GEOMOD. That Red Faction is an interesting game, too bad that the engine will suffer in high-detailed scenes. Would never get the same detail than Q3A powered games.
2) this is complicated : collision detector explosions. I would like to see some explosion model mesh that fits exactly in the surrounding area. I don´t want more clipped explosions through walls and floor.
3) this takes some time : entities that defines a set of brushes as "reals". They can be pushed, affected by gravity, affected by mechanics, they have to spin, slide and move as Newton Law´s dictates.

well, with all this that we´ve been talking one can make the GAME of the GAMES.

[Dima]

1)Yeah I agree that bumpmapped wallmarks will make believable damage without using GEOMOD.
2)Red Faction is an intresting game indeed it will be nice to finally destroy the enviroments(especially in hot deathmatch battles) but it will require some hefty computer with plenty of ram cause full destruction of enviroments requires some hefty horse power!
3)Yeah I agree collision in general is one of the main problems of today games (ever played Tombraider 4?) Quake 3 engined games seem to suffer from that also (Fakk2,Elite Force) it will be VERY hard to implement what u said but if it will be implemnted why not removing the collision problem at all?? (I am not sure how to accompish this) If Carmack removes the collision problem at all that alone will cause a REVOLUTION!!
4)Entitles that defines a set of brushes as "reals" hmmm...
T&L will definetly help in this cause it will free the computer for these kind of effects (gravity etc).
5)Why not use diplasement maps for the water?? EMBM provides nice water effects but still... but diplasement maps allow to deform almost any type of surface (waves on the water) that will make the water look INCREDUBLE!!! Yeah I agree that NURBS will make the best result but it will be VERY HARD to imlement NURBS even generally! But since it's Carmack I would say it's possible!
6)Voodoo 5 6000 won't take some advantage of 8 rendering passes or more in 2002! Doom 3 will look bad on this card anyway!
7)Yes uncompressed texture of this type will take plenty of ram but it will make the best image quality!(compressed textures SUCK!(Quake 3 Arena with the Geforce based card is a good example) By that time NV25 with 256mb ram will solve this problem.
8)Carmack also said that 64 bit color is the right way to go why?
32 bit is enough!
9)Yes I forgot that there are no lightmaps in D3 so 3d textures will be used for something else.
Now moving a little from the graphics of Doom 3!
Here is an article written in russian about the Pentium 4 Benchmarks (Don't pay attention to the words only for the benchmarks themselves) http://3dstyle.hardware.ru/cgi/news.pl?last=1#971236954
All I can say is PENTIUM 4 WHO?

[Zaldron]

64 bits is the obvious evolution. 32 bits only gives you RGBA information. The rest of the bit string could be used for whatever Carmacks wants. From saving extra information to avoiding color sampling problems. Imagine, 3DSMAX uses 64 bits images as the virtual framebuffer.