Tim Sweeny interview over @ BeyondUnreal

If you only have a 256 meg video card you will be running the game one step down, whereas if you have a video card with a gig of memory then you'll be able to see the game at full detail.
:oops:
 
Wunderchu said:
If you only have a 256 meg video card you will be running the game one step down, whereas if you have a video card with a gig of memory then you'll be able to see the game at full detail.
:oops:

By the time the game comes out, it might not be so far fetched. Still very high end, but probably a reality.
 
I like this part...

"It will be just another animation source for the hiarchy. If a player gets gets shoot in the arm it will go flying back, but then he will recover and continue animating on. Something we haven't been able to do in the past.

There'll be more intigration between hand created animation, physics animation and proceedural animation. Now days, all our joints have breaking strength in them. So if you actually shoot a character's arm with enough force it will break his arm off and go flying. You can pull arms or body parts off characters nowadays. It has some sick possibilities, but you can do it."
 
MasterBaiter said:
http://www.beyondunreal.com/content/articles/95_1.php

-edit- P.S. When is Beyond3D gonna have theirs up? :?
I've bugged Tim a couple of times now coz I'm a little upset that some of the questions I sent him a month ago has been answered in interviews elsewhere. I just reminded him again. Ours is a 17-question interview so perhaps it's simply a matter of Tim having to take extra time (compared to the other interviews) or he wants to give Very Good Answers.
 
Reverend said:
MasterBaiter said:
http://www.beyondunreal.com/content/articles/95_1.php

-edit- P.S. When is Beyond3D gonna have theirs up? :?
I've bugged Tim a couple of times now coz I'm a little upset that some of the questions I sent him a month ago has been answered in interviews elsewhere. I just reminded him again. Ours is a 17-question interview so perhaps it's simply a matter of Tim having to take extra time (compared to the other interviews) or he wants to give Very Good Answers.

I'm sure he's just busy.
Or maybe you sent him the questions twice and it pissed him off more than it did John Carmack? ;)
 
Isn't he going a little overboard with the 1 GB video card???

We aren't even breaking the 128 MB limit yet. Anything over 512 MB, even 2 years from now will be overkill. As humans our eyes can only see so much detail, especially in fast moving sequences. There will be a point where texture detail above a certain point will be unnecessary and unnoticed.

The Human eye can only detect close to TRUE 32-Bit color, anything higher wouldn't matter.
 
Ummm, even if they are extensively using 2048x2048 textures (which are, what, 12MB each?) that's still a lot of onboard RAM...
 
RingWraith said:
The Human eye can only detect close to TRUE 32-Bit color, anything higher wouldn't matter.

I often see banding the problem is the colours aren't distrubted over the 24bit colour range that best suits the eye. ( +/- 1 in luminance can often be seen but a small change in hue is often alot harder ).

grey.png


32-bit in all its glory.
 
Who's to say it's all textures? There's also high-poly models and the precomputed data needed for lighting techniques they're using. Just look at ATI's Ruby demo and its memory requirements. Also realize that the map used in Ruby is not particularily extensive. . .
 
@RingWraith

Actually, he's not going overboard with 1GB of RAM for graphics cards...
Texture detail today has little to do with color depth (which will probably stay put at 32 bit for a long time), but more with all kinds of instructions for extra effects, shaders, bump mapping and the like.

Plus, as Diplo pointed out, the texture resolutions are in fact growing beyond one's wildest dreams... I remember a time when 128x128 textures were used much to the satisfaction of all parties.
Even id thought of using 640x640 textures as a VERY high detail option for Doom III only a couple of years ago... of course, they don't need that much memory for textures anyhow, seeing as they chose to make very high polygon character models covered with relatively low res skins, effectively letting the T&L engine do most of the work in order to produce a crisp overall impression. The golden rule of videogame industry is 'there is never enough texture memory'.

BTW, I have found it to apply in case of any given PC component.
When I was playing behind a 386DX 40MHz computer with 4MB RAM, I would have had a really tough time imagining the applications that would need 512MB RAM at MINIMUM to run. Now it seems these applications are here. :rolleyes:
 
It's not only texture resolution but texture variety. Things look a lot more real with less repetition.

The other thing is the whole "polybump" technique of Doom3. Each models needs its own bump map texture (3Dc should help this a bit), and there's no repeating (except maybe left-right on some body parts).

Combining these, and a good virtual reality will need craploads of texture memory. 3D textures might be used more as well.
 
erick said:
Plus, as Diplo pointed out, the texture resolutions are in fact growing beyond one's wildest dreams... I remember a time when 128x128 textures were used much to the satisfaction of all parties.
Even id thought of using 640x640 textures as a VERY high detail option for Doom III only a couple of years ago... of course, they don't need that much memory for textures anyhow, seeing as they chose to make very high polygon character models covered with relatively low res skins, effectively letting the T&L engine do most of the work in order to produce a crisp overall impression. The golden rule of videogame industry is 'there is never enough texture memory'.

They don't cover a high detail model with a low resolution skin.

The ingame models are relatively low resolution from 1k polys to an assumed 5k polys (take that for any variant of "boss" level enemy).

The visible detail comes from normal mapping which has been discussed to great length here by more educated members than I am.

The high resolution models id software uses to extract normal maps, may it be for a wall panel texture or a character, aren't UV mapped at all.
It's beyond any nightmare to properly UV map a 3D character in the million poly range.

Typically id software still sticks to Quake3 texture resolutions which means 256* for a standard texture, 512* for a detailed wall texture and 1024* for a special decoration texture.

The visual outcome of Doom³ relates strongly to the resolution of the normal map (IMHO).

Everybody I know who's toying around with the leaked build of Doom³ has found out, that the diffuse map plays an important yet secondary role comapred to the resolution of the normal map.

I'm willing to think that an Unreal engine 3 model will look very similar in Doom³ when you use the same polycount and texture map resolution.
Of course there are nice additional touches like HDR lighting effects, but the base output looks similar.

A while (late 2002) ago I made some models between 5k and 15k polys with corresponding high resolution models up to 4million polys.
Then I extracted the tangent space normals and the surface colors of the high resolution models to 2048* texture maps, made a shader and viewed them in the leaked Doom³ build.

While such a model did use up to 96mb or texture data (4 texture maps at 2048* for each - mechanical and organic parts of the model, diffuse map, normal map, specular map and height map) it looked miles ahead of "current" technology.

The beloved Hellknight in Doom³ is textured with 512* maps on the body, now imagine how great he would look if you would increase the texture resolution to 2048* for each texture layer.

04.jpg


Tim Sweeny said they're using a texture resolution of 2048* for all and everything, so it's no wonder the models and the game world looks very nice and detailed.

So I totally agree with you that there can't be enough texture memory :)
 
Okay, BNA! I'm not going to get into an argument with you, seeing as we seem to be of the same mind concerning the possibility of using 1GB memory for graphics cards.

However, when John Carmack explained his method some 2 years ago, I was left under the impression, that the only REVOLUTION about Doom III would be the same method for increasing texture detail, as I described above. I admit that I haven't kept myself up to date with the newest developments, and indeed JC has had ample time to revise the engine (if he should wish to do so) as the launch date was pushed forward by a year.

And yes, I have heard the term 'normal mapping' before, and seen the diagrams and explanations as well. I think its great. So it doesn't really matter, which way Doom III works, as long as it looks sooo good ;)
 
erick said:
Okay, BNA! I'm not going to get into an argument with you, seeing as we seem to be of the same mind concerning the possibility of using 1GB memory for graphics cards.

No need to argue over anything.
It was just a more general post than a direct reply to your post.

However, when John Carmack explained his method some 2 years ago, I was left under the impression, that the only REVOLUTION about Doom III would be the same method for increasing texture detail, as I described above. I admit that I haven't kept myself up to date with the newest developments, and indeed JC has had ample time to revise the engine (if he should wish to do so) as the launch date was pushed forward by a year.

I think Carmack envisions his unified lighting model as the key point of Doom³. The rendering techonlogy is finished since quite some time, at least he said so a while ago.
But there's only little coming out of id software, only hype...
I miss the days of Carmack's plan updates.

And yes, I have heard the term 'normal mapping' before, and seen the diagrams and explanations as well. I think its great. So it doesn't really matter, which way Doom III works, as long as it looks sooo good ;)

Exactely :)
 
Back
Top