Doom3: Will it have a separate SM3 codepath ?

DaveBaumann said:
What was JC's quote? "Doom 3 is designed around the features and capabilities of GeForce2 style products, my next engine will be designed around nv30/nv40".

If I recall correctly, he stated that it was originally based around the Geforce256.
 
His 17 May, 2000 .plan update specifically said that the Geforce was his "baseline for current rendering work". On the first of June he confirmed that he was working on Doom 3.

It's kinda interesting to re-read some of those old .plans. . .
 
Guden Oden said:
He's already down to 1 rendering pass per light using plain DX9-level functionality...

Actually, he'd need at least two passes. One for the shadows (two if double-sided stencil isn't supported), and one for the lighting.
 
I think it all boils down to whether there's many situation under which SM3.0 or its equivalence in OpenGL can boost peformance. Anyone knows what kind of light does he use? Are they range-limited?
 
991060 said:
Are they range-limited?

Yes.

[edit: If you need more than my word on it, nVidia's "Robost Stencil Shadow Volumes" paper from 2001 specifically states this on slide 26. Plus, you pretty much have to do it this way, or else determining which lights need to be processed is extremely difficult (with finite-radius/range-limited lights, the objects you have to do lighting/shadowing calculations on is simply the union of objects within the "attenuation sphere" of all light's whose attenuation sphere intersects the view fustrum. If you don't cap the range of lights, this becomes a much much harder problem since you no longer have a closed sphere, but an entire 3d space of influence - you have to either calculate shadowing and lighting from every single lightsource in the level, every frame, or somehow determine that a light source in no way casts light or shadows anywhere in the view fustrum, and cull them away, one by one.)]
 
Ostsol said:
His 17 May, 2000 .plan update specifically said that the Geforce was his "baseline for current rendering work". On the first of June he confirmed that he was working on Doom 3.

It's kinda interesting to re-read some of those old .plans. . .

Is this bad???
 
Ilfirin said:
991060 said:
Are they range-limited?

Yes.
Thanks, good to know. Then I think dynamic branching COULD do sth good. But considering the lighting model used in DOOM3 is relatively simple, I'm not sure if the benifit is worth the penalty.
 
991060 said:
mito said:
Is this bad???
Good for Geforce1/2/3/4 owners, bad for NV40 owners. ;)
Bleh. Even though it'll run on Geforce 1s, it'll be at low resolutions and slowly. If you want decent speed you'll probably want to turn off alot of the cooler features. The middle-spec for the game will definitely be the Geforce 3/4 or Radeon 8500/9000/9100/9200.
 
Ostsol said:
991060 said:
mito said:
Is this bad???
Good for Geforce1/2/3/4 owners, bad for NV40 owners. ;)
Bleh. Even though it'll run on Geforce 1s, it'll be at low resolutions and slowly. If you want decent speed you'll probably want to turn off alot of the cooler features. The middle-spec for the game will definitely be the Geforce 3/4 or Radeon 8500/9000/9100/9200.
yeah, but it's not my point. What I meant is that if game developers only have a fixed target machine, they can put much more energy in refining the visual effect, rather than in paying effort on fall-back stuff. Just take a look at what they achieved on XBOX, you'll know what I mean.
 
991060 said:
yeah, but it's not my point. What I meant is that if game developers only have a fixed target machine, they can put much more energy in refining the visual effect, rather than in paying effort on fall-back stuff. Just take a look at what they achieved on XBOX, you'll know what I mean.
Given how long ago Doom 3's engine was started, I find it hard to be surprised at the fallbacks it has. The console performance vs PC performance issue is an old argument discussed by people far more qualified than myself, so I feel no obligation to try and reiterate it.
 
If I recall right didn't Carmack say something along the lines if you want to use a Geforce 256 you will have to run at a res of like 320x240 if you want decent framerates. I might be a little off but I recall something really low res like that for FPS around 30 he was estimating.

Edit:
"· The slowest cards will be the 64 bit and SDR ram GF and Radeon cards,which will really not be fast enough to play the game properly unless yourun at 320x240 or so. "
http://www.newdoom.com/newdoomfaq.php

Edit 2: Finally found the real original source for that :p
http://archive.gamespy.com/e32002/pc/carmack/index2.shtml
 
tEd said:
Waltar said:
DaveBaumann said:
What was JC's quote? "Doom 3 is designed around the features and capabilities of GeForce3 style products, my next engine will be designed around nv30/nv40".

Fixed.

now you got it wrong :?

daves post didn't need fixing
Dave was almost spot-on. JC's quote, IIRC, was that Doom 3 was based on the GF1 featureset (which is ~ GF2), and that his next engine will be based on NV30. I'm guessing he may change the base model of his next engine, given how NV30 shook out, but that may depend on its user base (which may be insignificant by the time his next engine debuts).
 
DaveBaumann said:
http://www.sharkyextreme.com/hardware/videocards/article.php/3211_1434621__2

well at least we know for his next engine the r3x0 and x800s should be fine running i ;)
 
jvd said:
DaveBaumann said:
http://www.sharkyextreme.com/hardware/videocards/article.php/3211_1434621__2

well at least we know for his next engine the r3x0 and x800s should be fine running i ;)

Like a GF will run Doom III ;)
Seems that he likes the flexibility of the NV30 architecture with no limit on dependant texture reads, and a large number of instructions. I think that's why he said his next engine will be designed around that level of features, but since I've read that the NV40 will be the base for his next engine.
 
Back
Top