Carmack on low polies on models and other things

Displacement mapping adds additional geometry at the 3D card level, it would cause the Doom3 engine similar problems TruForm would and JC already said he wouldn't/couldn't use that. So from Carmacks comments I think its safe to say we can forget about Doom3 using displacement mapping, at least for now - who knows how much the engine might still evolve before or even after Doom3's release, which is obviously still off in a rather distant future...

 

displacement mapping works only with newer cards, which aren't even on sale yet, but Carmacks current approach works on most cards (not all, because there are pleny of computers sold with tnt2 even today). Therefore no DM for another couple of years me thinks...

 

Reverend,

This is actually console related, but can you ask JC if it is still true that that if Doom III were ported to the Xbox that he would not need to turn off any graphical features (with the exception that the levels would have to be smaller)? Many people in various Xbox fan websites forum are saying that Doom III could run on Geforce3 if graphical features were turned off. Others are saying that JC's original target of Geforce3 to run Doom III has probably changed since last year, since Id software was running Doom III on a R300 at E3 this year.

 

The deployment of displacement mapping is dependent on how many OOHs and AAHs it will produce.

Just like the current Doom III engine, with it's dynamic lights/shadows, shifts emphasis towards stencil performance.

Installed base is, of course, always a concern. But with DM being included in DX9, I think most next generation cards will feature it.

Cheers
Gubbi

 

two questions i'd like to ask


1. Is there any other technic than stencil buffer used for shadows? It looks like from the movies/screenshot that the shadows of some part of the maps aren't stencil shadows but shadowmaps. (or maybe I just dreamed it)

2. Less graphic related, I've read that JC wasn't convinced by C++ compared to C because of code effenciency/speed. But it seems DOOM III is written in C++. Why this sudden change? And what's JC's current opinion on C and C++?

 

There's one question that's been bugging me since I saw the demo...most of the scenes are indoors and with an emphasis on dark corners and such, the number and type of lights is limited. This works well for that atmosphere but how does the engine work for outdoor scenes, with light coming from a not so specific source (the "sun" in the sky, if you will). How will this affect the shadows rendered and will everything be bumpmapped, eg., foilage, terrain, etc.

 

I'd like to know what resolution the demo was run in and whether or not they used FSAA and or ANSIO because the screenshots are amazingly clean and have no(?) nasty stepping.

I'd also like to know if all those amazing lighting effects etc are present in multiplayer?

If possible will he say how high the GF4 was clocked?

on another note, screw the slightly pointy heads it looks amazing

 

Reverend,

here are some questions, you could ask:

- which unit will calculate the shadow volumes (GPU / CPU or both - is fixed or can the user change the workload?)

- what about multi CPU systems? any andvantages for the shadow volume calculations possible?

- what about a linux version?

- will the engine use some higher precision surfaces (like the new ones in DX9) - through new OpenGL extensions?

- what about gamma correction? will there be advantages for DX9 class hardware (linear gamme ...) ?

- is there a benchmark option in the engine?

- whats the average visible polygon count?

- is there any use for volumetric textures in the engine?

- whats the average size of texture memory required from doom 3 ? will 64 mb cards have some problems with the highest texture quality settings?


Regards,
Thomas

 

Revrend,

Is this what he said? becuase as i read this it sounds to me like you are combining a coupple things he previously stated about 2 different products. Specifically, he stated that the GF4 Ti was faster than a R8500 *until anisostrpoic filtering is turned on*. Also, He clearly stated that The Gf4 they used was not an off the shelf model, but one running Very Fast, or tweaked. Another thing.. Are you simply assuming that the Anisotropic filtering on the R300 is *different*? with 8 pipes and 4 textures per pipe i kind of doubt its doing Aniso the same old radeon way.

Are you answering this based on your own opinion/assumption? or is there another part of your email you did not share with us.


As for a JC question... I am simply curious why the complete lack of even mentioning the Parhelia. It is a little odd.

 

Reverend,

I don't doubt your sources, yet there seems to be a slight oxymoron here:

Unless of course JC's comments on performance comparisons between NV25/R300, weren't at all related to the final presentation at E3.

 

From the videos of doom it looks like there is only shadowing from one source of light. I'd be interested in finding out how many simultaneous light sources we can typically expect in a scene, and how much the extra lighting and shadowing will slow down the frame rate.

 

I don't think he's using shadow maps due to lack of hardware support, plus that it's quite limited in it's uses anyway (only spotlights). There's obviously a lot of projective shadows in there though.

 

Quite obviuosly they used 2x FSAA as shown on this zoomed image:

 

If they were indeed demoing with FSAA turned on (even if only 2x) then I am even more impressed with the current performance of the [unoptimized] engine (and beta R300).

 

I hope I'm not too late, and this one is a bit of a loaded gun...

I'm not even sure if it's a driver or an application question, so if anyone can answer it ASAP that would be cool...

Will Doom III support single pass multitexturing up to eight layers if it's supported by the card, a 'la the Kyro series?

 

Doom3 will pretty much take as much advantage of the cards capabilities as possible. I don't think there much hope for Kyro series cards though with the lack of cubemaps.