"Yes, but how many polygons?" An artist blog entry with interesting numbers

Absense of post processing effects and of better lighting effects was something common in DC games.
DC games looked flatter in general. I think it was a limitation of the hardware. Even games like Shenmue that pushed it looked flat with barely much if any in terms of post processing.

Whatever the PS2 lacked in image resolution and texture quality it made up with effects, post processing and complex (for its time) lighting.
I doubt the DC could produce anything remotely close in terms of lighting and materials to TTT
Totally agree. The Dreamcast just lacked the computational power to do really good lighting with, at the time, high polygon counts. While I would bet the Dreamcast could come somewhat close to the polygon counts of some PS2 games, it would have to sacrifice all the lighting and effects that make those games look so good. I would bet that the PS2 could have a crazy polygon count game, if it sacrificed lighting to the extent most games made for Dreamcast did. I think that is as much to do with developers not really putting the time on Dreamcast development, as it was Dreamcast lacking a bit of computational power in the 6th gen console war. The DC was an amazing console, I really did enjoy mine, but the SH-4, as powerful as it was, could not replicate the vector processing of the VU0/VU1 of the PS2.
 
Totally agree. The Dreamcast just lacked the computational power to do really good lighting with, at the time, high polygon counts. While I would bet the Dreamcast could come somewhat close to the polygon counts of some PS2 games, it would have to sacrifice all the lighting and effects that make those games look so good. I would bet that the PS2 could have a crazy polygon count game, if it sacrificed lighting to the extent most games made for Dreamcast did. I think that is as much to do with developers not really putting the time on Dreamcast development, as it was Dreamcast lacking a bit of computational power in the 6th gen console war. The DC was an amazing console, I really did enjoy mine, but the SH-4, as powerful as it was, could not replicate the vector processing of the VU0/VU1 of the PS2.

But thats what i was pointing out, it does not have to. With one or 2 directional lights( which is what doa2 does) provided just enough to prevent from looking too flat. If the it can push enough geometry to keep up with soul calibur 2( probably does considering doa2 characters are 1.5x to 2x more polygons than tekken tag) and soul calibur 3 minus the multi texturing and no more than 2 lights it would still make it eaiser for multi platform because it keeps same texture and model assets. It wouldn't look that much different.

Post processing as long as it does its crossfade and basic non- interactive motion blur( like wall deaths on doa2) its more than enough.
 
Dead or Alive 2 ( limited edition )- Dreamcast
Extracted using naomilib for the model files. Out of curiousity to compare Doa2 to other games , I looked through other peoples dump of naomi 2 version vf4 and vf5 FS character models. Results were surprising. Naomi 2 virtua fighter 4 character models ranged from 12k to 14 triangles(double of its ps2 counterpart) and Virtua fighter 5 FS ranged from 10K( dural) - 20k triangles. At around 10k triangles per model DoA2 is impressive. especially for an early game.

Tina costume 2 - 10,028 triangles ( no shadow volumes)


ein costume 4 - 9,005 triangles ( no shadow volume)



Demons Church( lower floor) - 17,358 triangles
So circling back on this tried the debug version of Dead or Alive 2 Dreamcast , which lets you control the light sources. Turns out the game by default uses 3 light sources, 1 per fighter and 1 for the stage /props(each light has ambient). you can then further activate 4 more lights( it affects both fighters) and change their type from directional or point lights. Its hard to control the color so turned each light on and off and they indeed add to the models. So in total you can achieve 7 lights but it drops the framerate to around 30 fps which makes the game run in slow mo. ( tried on a real dreamcast)
Probably why the cutscenes are locked to 30 fps on the dc i guess it uses more lights.
The situation isnt as dire as you all made it seem to be honest if it can do doa2 detail with 7 lights and not be a slide show.

https://i.ibb.co/5jJyJ2D/deadoralive2lights.jpg
 
Totally agree. The Dreamcast just lacked the computational power to do really good lighting with, at the time, high polygon counts. While I would bet the Dreamcast could come somewhat close to the polygon counts of some PS2 games, it would have to sacrifice all the lighting and effects that make those games look so good. I would bet that the PS2 could have a crazy polygon count game, if it sacrificed lighting to the extent most games made for Dreamcast did. I think that is as much to do with developers not really putting the time on Dreamcast development, as it was Dreamcast lacking a bit of computational power in the 6th gen console war. The DC was an amazing console, I really did enjoy mine, but the SH-4, as powerful as it was, could not replicate the vector processing of the VU0/VU1 of the PS2.

Yep and it sticks out like a sore thumb, Shenmue is specially poor as every object has the same flat lighting and there's no shadows.

It's quite a poor looking game so I have no idea why it's graphics have always been hyped up as better than most of what PS2 could manage.

It took just 12 months to kill anything DC had graphically as it had Silent Hill 2, Devil May Cry, MGS2 and GT3..........those 4 games surpassed anything on DC by a good margin IMO.
 
So circling back on this tried the debug version of Dead or Alive 2 Dreamcast , which lets you control the light sources. Turns out the game by default uses 3 light sources, 1 per fighter and 1 for the stage /props(each light has ambient). you can then further activate 4 more lights( it affects both fighters) and change their type from directional or point lights. Its hard to control the color so turned each light on and off and they indeed add to the models. So in total you can achieve 7 lights but it drops the framerate to around 30 fps which makes the game run in slow mo. ( tried on a real dreamcast)
Probably why the cutscenes are locked to 30 fps on the dc i guess it uses more lights.
The situation isnt as dire as you all made it seem to be honest if it can do doa2 detail with 7 lights and not be a slide show.

https://i.ibb.co/5jJyJ2D/deadoralive2lights.jpg
For fighting games you need 60fps. So those extra lights are a no no.
In addition those games running at 30fps would probably kneel by enabling a few more lights. This shows that the DC was really struggling with more complex lighting in highly detailed scenes.
 
For fighting games you need 60fps. So those extra lights are a no no.
In addition those games running at 30fps would probably kneel by enabling a few more lights. This shows that the DC was really struggling with more complex lighting in highly detailed scenes.

In the case of DoA2 on DC, the game was almost certainly double buffered, with vsync, so any drop below 60 would automatically go to 30 fps. So the fact that the game with multiple additional lights drops to below 60 fps only means that you are now in a frame time of between 16.7 ms (60 fps) and 33.3 ms (30 fps).

In the case of Shenmue, the game in significant part predates the hardware itself. Much of Shenmue 1 (and 2) was produced before launch, and both polygon and texture budget use was far from ideal. With more experience and better tools, Sega would have produced even more impressive results on the same hardware. It was insanely ambitious to try and bring a game like that out so close to launch - asset creation and engine development was still in the Saturn age, and bug tracking was done manually on Excel! They were in some cases developing the tools for the system along with the game.

Shenmue's poly counts are in some cases well below what the system is capable of, with modders able to insert a massively more detailed Ryo model into areas where GPU memory allows (if it doesn't crash due to running out of memory for the GPU's poly binning stuff). If something like Panzer Orta had seen life on the DC in ~2003 as originally planned, I think it would have shown real progress over Sega's first gen stuff.

Years apart, DC (1998) simply didn't have the BOM to really match the PS2 (2000). A second SH4 would have easily allowed 2 ~3 times the poly counts in DC games, but that would have also required more GPU ram for the deferred geometry, and probably more main ram for more complex models. Normal maps could have made a big difference too, but these would also have required more vram. That wasn't happening in a $200 console in 1998 / 1999, especially not from Sega's broke coffers.
 
For fighting games you need 60fps. So those extra lights are a no no.
In addition those games running at 30fps would probably kneel by enabling a few more lights. This shows that the DC was really struggling with more complex lighting in highly detailed scenes.
Some Fighting games in the past were below 60 fps so that argument is void. Especially when theres money to be made from releasing to a lower platform in a multiplatform release, if they felt they could make money from a dc release you could probably bet they would have just kept same assets with half framerate and minus multi texturing.

Youre kind of moving the goal post now, regardless of genre why wouldnt 30k to 40k triangles per scene with 4 to 7 lights wouldnt be enough at 30 fps? At that time on pc video drivers had like 8 hardware lights limit.( or was it fixed function driver limit?) Even sonys own doc of how far we got says average polygon per scene was about 50k with 60% of the games running 30 to 25 fps at 2003. Dc would have been just fine especially with sdk improving also it DOESNT HAVE TO render that amount of lights.Can the ps2 do better? Yes but not necessarily every dev bothers with it at that time frame.

Screenshot-20220703-194420-Adobe-Acrobat.jpg
 
I'm confused now, is the discussion now about DC vs PS2 graphically?

Is it DC vs PS2 in fighting game graphics?

Is it DC vs PS2 in games with more than 3 lights?
 
Some Fighting games in the past were below 60 fps so that argument is void. Especially when theres money to be made from releasing to a lower platform in a multiplatform release, if they felt they could make money from a dc release you could probably bet they would have just kept same assets with half framerate and minus multi texturing.

Youre kind of moving the goal post now, regardless of genre why wouldnt 30k to 40k triangles per scene with 4 to 7 lights wouldnt be enough at 30 fps? At that time on pc video drivers had like 8 hardware lights limit.( or was it fixed function driver limit?) Even sonys own doc of how far we got says average polygon per scene was about 50k with 60% of the games running 30 to 25 fps at 2003. Dc would have been just fine especially with sdk improving also it DOESNT HAVE TO render that amount of lights.Can the ps2 do better? Yes but not necessarily every dev bothers with it at that time frame.

Screenshot-20220703-194420-Adobe-Acrobat.jpg
Which fighting games are you comparing to?
Soul Calibur, DOA, Tekken, VF and SF always targeted 60fps.
Having these franchises run at 30fps would have been a first and one issue that would have affected gameplay greatly.
Now if you want some generic fighting game at 30fps, its fine. Then of course we got a lot of generic non-fighting games that were below even 30fps anyways.
If Tecmo could do DOA at 30fps with more lights why didnt they do it?
I am pretty sure other genres with more lights would have tanked the framerate below 30fps
 
I'm confused now, is the discussion now about DC vs PS2 graphically?

Is it DC vs PS2 in fighting game graphics?

Is it DC vs PS2 in games with more than 3 lights?
Well you are correct, best to nip it in the bud. Next time i post will be about more polycounts since its going off rail.
 
Which fighting games are you comparing to?
Soul Calibur, DOA, Tekken, VF and SF always targeted 60fps.
Having these franchises run at 30fps would have been a first and one issue that would have affected gameplay greatly.
Now if you want some generic fighting game at 30fps, its fine. Then of course we got a lot of generic non-fighting games that were below even 30fps anyways.
If Tecmo could do DOA at 30fps with more lights why didnt they do it?
I am pretty sure other genres with more lights would have tanked the framerate below 30fps
Virtua fighter ran below 60 in the past. Vf1 was 30 fps in the arcades if you didnt know. Which i guess you didnt bother.

They did bother, they used the extra lights from cutscenes, you can see the light shift when the match seamlessly starts. It runs them at 30 fps. And i showed you if they wanted they could have done the same for gameplay. 30 fps wouldnt be the death of a fighting game especially if theycare getting extra money from a lower sku. Your bias is showing is all.

And even for other genres they turn lights on and off in proximity, yes even on ps2 . 7/8lights isnt an issue because most games then didnt go above that with the lights around them. They can have many lights but all inactive unless near.Why do think shreks deffered lighting was such a big deal on xbox 1 huh? Forgot that too?
 
Last edited:
Someone told me about a Soul calibur 2 importer recently ,so here it is i guess.

Soul Calibur 2 - multiplat

Yoshimitsu - 7,134 triangles( seems to be one of the higher ones)
https://i.ibb.co/Q6fk157/sc24.jpg


Taki - 6,046 triangles
https://i.ibb.co/W5pH2Fw/sc23.jpg


Village of Wind - 23, 328 triangles
https://i.ibb.co/PQgwtdd/sc22.jpg


Shrine - 31,635 triangles
https://i.ibb.co/X5nYrwC/sc21.jpg
 
Virtua fighter ran below 60 in the past. Vf1 was 30 fps in the arcades if you didnt know. Which i guess you didnt bother.

They did bother, they used the extra lights from cutscenes, you can see the light shift when the match seamlessly starts. It runs them at 30 fps. And i showed you if they wanted they could have done the same for gameplay. 30 fps wouldnt be the death of a fighting game especially if theycare getting extra money from a lower sku. Your bias is showing is all.

And even for other genres they turn lights on and off in proximity, yes even on ps2 . 7/8lights isnt an issue because most games then didnt go above that with the lights around them. They can have many lights but all inactive unless near.Why do think shreks deffered lighting was such a big deal on xbox 1 huh? Forgot that too?
I dont think Sega would ever want to have the framerate of primitive 3d fighting games
 
They did bother, they used the extra lights from cutscenes, you can see the light shift when the match seamlessly starts. It runs them at 30 fps. And i showed you if they wanted they could have done the same for gameplay. 30 fps wouldnt be the death of a fighting game especially if theycare getting extra money from a lower sku. Your bias is showing is all.

Are those the same DOA2 cut scenes that run at 60fps on PS2 but 30fps on DC?
 
Are those the same DOA2 cut scenes that run at 60fps on PS2 but 30fps on DC?

The meaning of that is only that DC couldn't maintain 60 while PS2 could. Performance in ms could have been anywhere between 33.3 and 16.7ms per frame. I personally have no idea where it landed, and I don't believe anyone else in this thread does.

Best to stick to what we know. DC games used only a relatively small percentage of CPU for T&L because there's a lot of other stuff going on in a game that needs processing. Estimates from actual DC devs that I've seen talked about roughly ~20%-ish, though I'm sure that moved a bit depending on the game.

Every other system of that gen had a greater proportion of processing time that was naturally directed towards T&L. That's neither good nor bad, it's just how the DC with its architecture had to work. And with a relatively tiny power and silicon budget, DC was pretty incredible. When you can bolt on extra vector units that run in parallel to core CPU, things can change a lot.
 
Cosmic Smash - Dreamcast

Cosmic Smasher - 1,016/2,032 triangles( he has an outer shell that is exactly the same but larger, gives outline in game. furthermore he has an internal visible "skeleton" that I didnt bother getting so he might be a tad higher.)
cosmic1


Racket and ball - 512 triangles
https://i.ibb.co/hXv2Z9m/cosmic2.jpg


Cosmic bus - 560 traingles
https://i.ibb.co/4ds5qPf/cosmic4.jpg


Cosmic bus( i suspect this is a mesh used to control how the specular is seen in game, its overlaid unto the actual bus model)- 348 triangles
https://i.ibb.co/Pw2qN30/cosmic6.jpg


Block - 156 triangles( didnt post but has shadow casters and morph target to animate destruction)
https://i.ibb.co/9YKqbVW/cosmic5.jpg


Cosmic bus dark - 220 triangles
https://i.ibb.co/QMPyWSq/cosmic7.jpg


cosmic bus from cutscene - 1,594 triangles
https://i.ibb.co/SPxfmnF/cosmic3.jpg
 
Virtua Athlete 2000 - Dreamcast - Hitmaker

Highest athlete LOD ( used for single person event , closest to camera in multi person event or all in replay multiperson,8) -2,240 triangles
https://i.ibb.co/F4nqKXC/vathlete2k1.jpg


Low LOD - 1,172 triangles
https://i.ibb.co/hfbsWVh/vathlete2k2.jpg


High Shadow LOD ( single person event) - 562 triangles
https://i.ibb.co/fn9kGPd/vathlete2k4.jpg


low shadow LOD ( multiperson event) - 152 triangles
https://i.ibb.co/DfxmrNr/vathlete2k3.jpg
 
  • Like
Reactions: snc
Back
Top