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

Amazing bro!! First of all, i´d like to thank for your work in one of the Dreamcast games i love the most. It gave me many great gaming moments 21 years ago!!

Headhunter really impressed me at the time for the quality of the stages specially. At the beginning, remember the Stern Mansion and being shocked cuz to me it looked prerendered quaility, but in real time. Also the Grey Wolf boss fight...The rain effect, lightings and reflexions of stage objects still look unreal to this day....You really felt back in the day that with Headhunter, Rez, Shenmue 2, Test Drive Le Mans, Virtua Tennis 2, Ikaruga and DOA2, DC wasnt that far on the visual level from PS2.

Now you proof us you can make a fantastic looking game with less than 15k polys on screen at once....Question is: Do you think DC could have been pushed further? Graphical level of the likes of PSP´s GTA LCS, Need For Speed Most Wanted or Burnout Dominator, could have been reached by DC? or even that is out of league?
Funny you should mentions just those, as its them I was directly involved with (and a bit Wolfpack HQ) as they where the first environments we had in the game so later on when the project and company grew I unfortunately could not do any more hands on work.

And yeah it proves how much art direction helps. I think we was quite early also starting to use photos over handpainted textures. That made it look more realistic I think. We had bought two 1.3mPixels (was A LOT back then) cameras and two guys in the art team went to LA for two weeks to shoot around 10k photos. DC was great in that photos worked really well with the PVR compression. No limits to color and as the packing could make them a but noisy was less visible as the photos themselves was a bit noicy by nature :)

Regarding prerendered I also think we was quite early in baking in lighting not just into vertex colors, that had been done quite a lot PS1 era but directly into textures. So we basicallt lite the scene in 3DS max where we had many object mapped to textures space was used over all object. It took a bit more texture use but allowed us to, especially on floors, bake lighting into kind of lightmaps. Then of course we compromised. Like those leather chairs in Angelas place have the same texture lighting on both with a little adjustment of vertex color. And then we added fake reflections (shiny floor) nd a lot of transparency (curtains etc) to make those things look more real. Same with the greywolf bossfight. Its just particle systems and animated textures for the rain FX (and mirrored objects for the fake reflection) I mean objects reflect but Grewolf and Jack do not.. still few seems to mention or notice that :D
The falling rains itself is just a cylinder with an raindrop texture that I attached to Jack. So the rain always is around him. And then by code the texture is animated. In that way we could change the scrolling direction and speed very easy. To simulate the direction the rain is falling due to wind etc.. I think it worked pretty well.
Now you proof us you can make a fantastic looking game with less than 15k polys on screen at once....Question is: Do you think DC could have been pushed further? Graphical level of the likes of PSP´s GTA LCS, Need For Speed Most Wanted or Burnout Dominator, could have been reached by DC? or even that is out of league?

I think games like Shem Mue II and Dead or Alive pushed it pretty far tbh. A lot of people compare DC with the gap of early and late PS2 games but the PS2 is a completly different beast. Also there is memory limitations that actually caps the DC from reaching such hi poly count (simple version: the polygons to draw takes place in DC graphics memory so to achieve the combination of varied detailed textures and tons of polygons like say GOD of WAR 2 is not possible)

That said I think it would have been possible to make better games for sure. I mean we learned a lot from Headhunter, and with better art tool and direction we could make something LOOK more even if it wasnt more if you get what I mean?
So yeas I think games that used DC smarter and maybe even was optimised a little little better would have been possible. But it would not have been huge differences.
 
Graphical level of the likes of PSP´s GTA LCS, Need For Speed Most Wanted or Burnout Dominator

Speaking about that. I have started to talk with some people in the DC indie dev scene so we are planning to (still need to iron out things and it will for sure be time before it might end up as a Kickstarter project) do a smaller game that is feasable for a few people. To avoid a lot of art production and advanced game systems we are thinking of making a F Zero / FAST type of futuristic racer.
Goal there would be to try to really work along with the DC hardware, do smart art stuff, fake things as hell to make it look like those games you mentioned. Like racing with lots of nice lighting, bloom, specular stuff, reflections, to make it look like later PS2, PSP games. Burnout Dominator is actually in my list of visual references :)

So if you are up for a new shiny DC racer in the future, please spread the word as we need to see if there is any interest for this type of game... Atm I myself just put the temporary worktitle "Project DreamStream" (a nudge to the worktitle of Metropolis Street Racer / MSR that was "project Gotham" for a long while) and as said talking with people to form a small team.
One REALLY neat thing is I have already got thumbs up from composer Richard Jacques (MSR, Headhunter, Jet Set Radio, Sonic etc) do do music for a track or so :)
 
Funny you should mentions just those, as its them I was directly involved with (and a bit Wolfpack HQ) as they where the first environments we had in the game so later on when the project and company grew I unfortunately could not do any more hands on work.

And yeah it proves how much art direction helps. I think we was quite early also starting to use photos over handpainted textures. That made it look more realistic I think. We had bought two 1.3mPixels (was A LOT back then) cameras and two guys in the art team went to LA for two weeks to shoot around 10k photos. DC was great in that photos worked really well with the PVR compression. No limits to color and as the packing could make them a but noisy was less visible as the photos themselves was a bit noicy by nature :)

Regarding prerendered I also think we was quite early in baking in lighting not just into vertex colors, that had been done quite a lot PS1 era but directly into textures. So we basicallt lite the scene in 3DS max where we had many object mapped to textures space was used over all object. It took a bit more texture use but allowed us to, especially on floors, bake lighting into kind of lightmaps. Then of course we compromised. Like those leather chairs in Angelas place have the same texture lighting on both with a little adjustment of vertex color. And then we added fake reflections (shiny floor) nd a lot of transparency (curtains etc) to make those things look more real. Same with the greywolf bossfight. Its just particle systems and animated textures for the rain FX (and mirrored objects for the fake reflection) I mean objects reflect but Grewolf and Jack do not.. still few seems to mention or notice that :D
The falling rains itself is just a cylinder with an raindrop texture that I attached to Jack. So the rain always is around him. And then by code the texture is animated. In that way we could change the scrolling direction and speed very easy. To simulate the direction the rain is falling due to wind etc.. I think it worked pretty well.


I think games like Shem Mue II and Dead or Alive pushed it pretty far tbh. A lot of people compare DC with the gap of early and late PS2 games but the PS2 is a completly different beast. Also there is memory limitations that actually caps the DC from reaching such hi poly count (simple version: the polygons to draw takes place in DC graphics memory so to achieve the combination of varied detailed textures and tons of polygons like say GOD of WAR 2 is not possible)

That said I think it would have been possible to make better games for sure. I mean we learned a lot from Headhunter, and with better art tool and direction we could make something LOOK more even if it wasnt more if you get what I mean?
So yeas I think games that used DC smarter and maybe even was optimised a little little

While I do agree that the numbers of God of war 2 wouldn't be possible on DC, I still think psp levels of detail or even better would be perfectly fine. That's what @xaeroxcore was asking, btw, what do you think of psp (playstation portable) numbers on Dreamcast?
 
While I do agree that the numbers of God of war 2 wouldn't be possible on DC, I still think psp levels of detail or even better would be perfectly fine. That's what @xaeroxcore was asking, btw, what do you think of psp (playstation portable) numbers on Dreamcast?

I dont have such knowledge about PSP, still when I look at the most polygon pushing PSP games it seems they are a bit over what the DC can achive. I always Take Dead or Alive 2 as an example of what is possible more or less for DC and as you can see its not a lot of textures there. Then when you have games with more textures the polycount goes down.

What many stil seems to overlook or ignore (I dont know which as its rather easy to calculate it) is DC GPU need to store the framebuffers, textures and display list in the texture memory. When that is full thats it. Do not matter how optimised that CPU stuff is. While from what I get from PSP its similar to PS2 that as long as you can feed it textures and polygon data it just keep drawing them. So if coded in a good way it can achieve more polygons.

So if level of detail comes down to polygons. Then nope I dont think so. If it comes down to a combination of textures and polygons that achieve a visual that look as detailed, then yes I think you can make DC games look on pair with great PSP and PS2 games.

But again, I have not worked on PSP so I am a bit in the dark there. Sorry :)

And on a personal note, who cares really? I mean its different machines. I am more interested in pushing as much on respective machines. So the question, could the DC be pushed more (in comparison to itself) is A LOT more interesting. Also its a question that is actually possible to answer if people would develop more for it in the homebrew scene. Same as we have seen amazing games for Atari 2600, C64, Megadrive, Intellivision etc I would love that the same happened to DC :)
 
Doricatch - FunCargo (Toyota Doricatch series for Dreamcast)

One of the possible Car configurations - 39.576 polygons.

image.png


RIGHT_P.png



image.png

Sin_titulo-1.png

View attachment 9403

Yes the interior is modeled.

I guess they like doing cars. The toyota supra from the very rare sh4 demo disc for dc is 41,000 triangles. with like 2 mb of textures. impressive considering it was a somewhat interactive demo at 60 fps.

Sh4 tech demo disc- dreamcast

toyota supra -41,702 triangles
toyodc.jpg
These are absolutely Amazing! I’m still trying to get that Ridge Racer demo to compare. Cars have been well served by tech demos, I wish fighter jets got a similar treatment!


Some video of the tech demo, clearest footage. It’s stunning to look at, I would have thought it higher then 41k which is amazing for its time!
 
I dont have such knowledge about PSP, still when I look at the most polygon pushing PSP games it seems they are a bit over what the DC can achive. I always Take Dead or Alive 2 as an example of what is possible more or less for DC and as you can see its not a lot of textures there. Then when you have games with more textures the polycount goes down.

What many stil seems to overlook or ignore (I dont know which as its rather easy to calculate it) is DC GPU need to store the framebuffers, textures and display list in the texture memory. When that is full thats it. Do not matter how optimised that CPU stuff is. While from what I get from PSP its similar to PS2 that as long as you can feed it textures and polygon data it just keep drawing them. So if coded in a good way it can achieve more polygons.

So if level of detail comes down to polygons. Then nope I dont think so. If it comes down to a combination of textures and polygons that achieve a visual that look as detailed, then yes I think you can make DC games look on pair with great PSP and PS2 games.

But again, I have not worked on PSP so I am a bit in the dark there. Sorry :)

And on a personal note, who cares really? I mean its different machines. I am more interested in pushing as much on respective machines. So the question, could the DC be pushed more (in comparison to itself) is A LOT more interesting. Also its a question that is actually possible to answer if people would develop more for it in the homebrew scene. Same as we have seen amazing games for Atari 2600, C64, Megadrive, Intellivision etc I would love that the same happened to DC :)
Well said, my man! That's exactly how I feel! The Dreamcast is it's own beast and is powerful relative to its contemporaries. It would have gotten better with time, just like every other console in the world. According to what I've seen even of doa2 mods on DC, tecmo wasn't even done with showing what their engine could truly do! I'm so ready to join a team of dedicated DC lovers and atleast develope a well crafted tech demo!
 
I dont have such knowledge about PSP, still when I look at the most polygon pushing PSP games it seems they are a bit over what the DC can achive. I always Take Dead or Alive 2 as an example of what is possible more or less for DC and as you can see its not a lot of textures there. Then when you have games with more textures the polycount goes down.

What many stil seems to overlook or ignore (I dont know which as its rather easy to calculate it) is DC GPU need to store the framebuffers, textures and display list in the texture memory. When that is full thats it. Do not matter how optimised that CPU stuff is. While from what I get from PSP its similar to PS2 that as long as you can feed it textures and polygon data it just keep drawing them. So if coded in a good way it can achieve more polygons.

So if level of detail comes down to polygons. Then nope I dont think so. If it comes down to a combination of textures and polygons that achieve a visual that look as detailed, then yes I think you can make DC games look on pair with great PSP and PS2 games.

But again, I have not worked on PSP so I am a bit in the dark there. Sorry :)

And on a personal note, who cares really? I mean its different machines. I am more interested in pushing as much on respective machines. So the question, could the DC be pushed more (in comparison to itself) is A LOT more interesting. Also its a question that is actually possible to answer if people would develop more for it in the homebrew scene. Same as we have seen amazing games for Atari 2600, C64, Megadrive, Intellivision etc I would love that the same happened to DC :)
I will say, I think DC's output relative to PSP would probably be higher poly atleast in fighting games.
 
Funny you should mentions just those, as its them I was directly involved with (and a bit Wolfpack HQ) as they where the first environments we had in the game so later on when the project and company grew I unfortunately could not do any more hands on work.

And yeah it proves how much art direction helps. I think we was quite early also starting to use photos over handpainted textures. That made it look more realistic I think. We had bought two 1.3mPixels (was A LOT back then) cameras and two guys in the art team went to LA for two weeks to shoot around 10k photos. DC was great in that photos worked really well with the PVR compression. No limits to color and as the packing could make them a but noisy was less visible as the photos themselves was a bit noicy by nature :)

Regarding prerendered I also think we was quite early in baking in lighting not just into vertex colors, that had been done quite a lot PS1 era but directly into textures. So we basicallt lite the scene in 3DS max where we had many object mapped to textures space was used over all object. It took a bit more texture use but allowed us to, especially on floors, bake lighting into kind of lightmaps. Then of course we compromised. Like those leather chairs in Angelas place have the same texture lighting on both with a little adjustment of vertex color. And then we added fake reflections (shiny floor) nd a lot of transparency (curtains etc) to make those things look more real. Same with the greywolf bossfight. Its just particle systems and animated textures for the rain FX (and mirrored objects for the fake reflection) I mean objects reflect but Grewolf and Jack do not.. still few seems to mention or notice that :D
The falling rains itself is just a cylinder with an raindrop texture that I attached to Jack. So the rain always is around him. And then by code the texture is animated. In that way we could change the scrolling direction and speed very easy. To simulate the direction the rain is falling due to wind etc.. I think it worked pretty well.


I think games like Shem Mue II and Dead or Alive pushed it pretty far tbh. A lot of people compare DC with the gap of early and late PS2 games but the PS2 is a completly different beast. Also there is memory limitations that actually caps the DC from reaching such hi poly count (simple version: the polygons to draw takes place in DC graphics memory so to achieve the combination of varied detailed textures and tons of polygons like say GOD of WAR 2 is not possible)

That said I think it would have been possible to make better games for sure. I mean we learned a lot from Headhunter, and with better art tool and direction we could make something LOOK more even if it wasnt more if you get what I mean?
So yeas I think games that used DC smarter and maybe even was optimised a little little better would have been possible. But it would not have been huge differences.

Speaking about that. I have started to talk with some people in the DC indie dev scene so we are planning to (still need to iron out things and it will for sure be time before it might end up as a Kickstarter project) do a smaller game that is feasable for a few people. To avoid a lot of art production and advanced game systems we are thinking of making a F Zero / FAST type of futuristic racer.
Goal there would be to try to really work along with the DC hardware, do smart art stuff, fake things as hell to make it look like those games you mentioned. Like racing with lots of nice lighting, bloom, specular stuff, reflections, to make it look like later PS2, PSP games. Burnout Dominator is actually in my list of visual references :)

So if you are up for a new shiny DC racer in the future, please spread the word as we need to see if there is any interest for this type of game... Atm I myself just put the temporary worktitle "Project DreamStream" (a nudge to the worktitle of Metropolis Street Racer / MSR that was "project Gotham" for a long while) and as said talking with people to form a small team.
One REALLY neat thing is I have already got thumbs up from composer Richard Jacques (MSR, Headhunter, Jet Set Radio, Sonic etc) do do music for a track or so :)

So we are witnessing Dreamcast history in the making on this thread!!

My man! We are very excited to see the results of your project DreamStream! Our Dreamcasts need to be pushed forward beyond its limits finally!

I´ll spread the word of course!

And, finally, you´re completely right, the focus is not how DC compares with other systems, is how to take it beyond it´s own limits. Probably this approach will end up in greater results!

Thanks for keeping the dream alive!!!


Cheers from Colombia, South América!!
 
These are absolutely Amazing! I’m still trying to get that Ridge Racer demo to compare. Cars have been well served by tech demos, I wish fighter jets got a similar treatment!


Some video of the tech demo, clearest footage. It’s stunning to look at, I would have thought it higher then 41k which is amazing for its time!
AH yes I remember this disc. Just when we had got started and had prety bad performance (before we got a great AMS coder onboard) one of the programmers complain that some of the numbers SEGA had been telling us was not realistic. And then one of us spoke with SEGA and got this disc and run the teapots and the progammer went silent. I think it was 8 or 12k tris in each teapot so a bit over 50k or 80k on screen with lighting and all. If course it demonstrates the importanse to keep stuff in the cache
 
So we are witnessing Dreamcast history in the making on this thread!!

My man! We are very excited to see the results of your project DreamStream! Our Dreamcasts need to be pushed forward beyond its limits finally!

I´ll spread the word of course!

And, finally, you´re completely right, the focus is not how DC compares with other systems, is how to take it beyond it´s own limits. Probably this approach will end up in greater results!

Thanks for keeping the dream alive!!!


Cheers from Colombia, South América!!
Yeah lets see. As its atm just a concept and a side hobby project like many os those probably a few years until something more tangible. I think next year will mostly be finding the right people and testing out things and geting it more down to specifics. But after that if we get that done I dont think its going to take so long or many people.

Actually I think the hardest part will be to make good tracks, so if anyone here has a lot of experience in Wipeout / F zero / FAST etc. And also can a bit modelling that for sure will be needed as a track / level designer.
As I think its super important to make something that does not just pretty but plays really well. I mean there was quite a few different racers wipeout style released in 1998-2001 som ugly and some pretty ok looking, but from what I rememebr most of them played really, really bad.

Also of course its about scaling, how many track to do.
For you personally, how many track would you say would be minimum for you to feel it would be a worthwhile purchase? How many different ships etc..

Also I assume 2 player head to head racing would be needed? Or would it be ok with just 1 player racing against AI? such things is very important for me to get a grasp on so we can target the best performance for the best experince. (like me personally love 4 player gaming. but I think not all DC owners have 4 controllers or do 4 player sofa gaming anymore. Online play is of course interesting but much, much bigger in scope to get playable.

So we need to be realistic in what we can do.
 

Attachments

  • DreamStreamTitle.png
    DreamStreamTitle.png
    293.6 KB · Views: 5
Yeah lets see. As its atm just a concept and a side hobby project like many os those probably a few years until something more tangible. I think next year will mostly be finding the right people and testing out things and geting it more down to specifics. But after that if we get that done I dont think its going to take so long or many people.

Actually I think the hardest part will be to make good tracks, so if anyone here has a lot of experience in Wipeout / F zero / FAST etc. And also can a bit modelling that for sure will be needed as a track / level designer.
As I think its super important to make something that does not just pretty but plays really well. I mean there was quite a few different racers wipeout style released in 1998-2001 som ugly and some pretty ok looking, but from what I rememebr most of them played really, really bad.

Also of course its about scaling, how many track to do.
For you personally, how many track would you say would be minimum for you to feel it would be a worthwhile purchase? How many different ships etc..

Also I assume 2 player head to head racing would be needed? Or would it be ok with just 1 player racing against AI? such things is very important for me to get a grasp on so we can target the best performance for the best experince. (like me personally love 4 player gaming. but I think not all DC owners have 4 controllers or do 4 player sofa gaming anymore. Online play is of course interesting but much, much bigger in scope to get playable.

So we need to be realistic in what we can do.
I for one think it would be great to have atleast 5-10 cars with atleast 5-10 tracks. A story for each character would be epic, 4 player split screen is definitely a viable thing for DC fans in my opinion. The online multi-player would just put it over the top! Something that takes advantage of the vmu would be epic! I have some ideas, but maybe I should talk to a vmu guy about that being possible 😅
 
I for one think it would be great to have atleast 5-10 cars with atleast 5-10 tracks. A story for each character would be epic, 4 player split screen is definitely a viable thing for DC fans in my opinion. The online multi-player would just put it over the top!

Ok great info! Many thanks! I myself was thinking minimum 8 tracks / vehicles so that maps nicely. Regarding characters idea was to not have any, as no racer really have any. And its a LOT more production effort to make those vs just ships

So idea would be this regarding players:

- 1 Player mode: 60 fps mode without some framebuffer FX (like glows, blurs, more dense particles and more close LOD switching) for those who really enjoys a 60fps racer. And a 30 fps mode with more nice FX
- 2 Player mode 60 fps with less detailed geometry / lighting and 30 fps with similar quality as the 60 fps single player mode
- 4 player mode 30 fps with lowest details

Does that sound like a good compromise for you as a player. So 1 player mode has the choice of 60 vs 30 fps mode. And if so what do you think would be best as default? The nicer looking 30fps mode or the more smooth running but less visually impressive 60fps mode?
Something that takes advantage of the vmu would be epic! I have some ideas, but maybe I should talk to a vmu guy about that being possible 😅

Actually that is a great reminder. I totally forgot about it. It would be great with VMU features that adds but the game does not rely on (if we want to release it later on other platforms)

If you have any idea you want to share, please do. Either directly here or if you prefer to tell in a PM that is also ok.
 
Ok great info! Many thanks! I myself was thinking minimum 8 tracks / vehicles so that maps nicely. Regarding characters idea was to not have any, as no racer really have any. And its a LOT more production effort to make those vs just ships

So idea would be this regarding players:

- 1 Player mode: 60 fps mode without some framebuffer FX (like glows, blurs, more dense particles and more close LOD switching) for those who really enjoys a 60fps racer. And a 30 fps mode with more nice FX
- 2 Player mode 60 fps with less detailed geometry / lighting and 30 fps with similar quality as the 60 fps single player mode
- 4 player mode 30 fps with lowest details

Does that sound like a good compromise for you as a player. So 1 player mode has the choice of 60 vs 30 fps mode. And if so what do you think would be best as default? The nicer looking 30fps mode or the more smooth running but less visually impressive 60fps mode?


Actually that is a great reminder. I totally forgot about it. It would be great with VMU features that adds but the game does not rely on (if we want to release it later on other platforms)

If you have any idea you want to share, please do. Either directly here or if you prefer to tell in a PM that is also ok.
I say definitely go all out for 30fps, then reduce where you need to reach 60 in all modes, like you said! The character thing was more of an illustration thing, not models in the ships, kinda like f-zero does. Maybe if there's time or someone that can do models, have a model viewer of different characters posed next to their racers as a trophy for winning races, kind of how they did with smash bros melee. Good incentive for replay value. Also, yes, I'd love to DM you about the vmu ideas, etc! I think this would be the most epic indie game out there if done right!
 
The character thing was more of an illustration thing, not models in the ships, kinda like f-zero does.
For sure we will consider it. Still its also not just about production but performance, if we want to show the vehicles looking really nice, with reflections effects, maybe bump mapping etc. We need to spill most of the power on the ships themselves. If its possible to both have nice looking character and ship All go for that! If we would end up with just ok looking character and ship, I would say we would go for just great looking ship. So that we need to see what is possible.

Also, yes, I'd love to DM you about the vmu ideas,
Cool I will read them now. Thanks!
 
I for one think it would be great to have atleast 5-10 cars with atleast 5-10 tracks. A story for each character would be epic, 4 player split screen is definitely a viable thing for DC fans in my opinion. The online multi-player would just put it over the top! Something that takes advantage of the vmu would be epic! I have some ideas, but maybe I should talk to a vmu guy about that being possible 😅
I might be interested doing some modelling and texturing for you if time is available when you decide to make it
 
Yeah I was tech art lead for Headhunter and sometimes it was even less htan that. Also regarding culling I would say most game of the time had some form of LOD and culling system. I mean there is reasons why levels often was built a bit mazy like, not just for the gameplay but to be able to block stuff. For DC/PVR that had deffered rendering it was not such and issue with overdraw but on other consoles like PS1, Saturn culling was needed just not for processing but draw reasons.

Also about shen Mue not having as advanced lighting as Code Veronica, the lighting in headhunter is super simple (I designed the system) its just maximum two directional (paralell lighths) and just one color (so not possible to have two colored lighting on objects) per object.
i was quoting what you said on some youtube video about 15k triangles on screen.

I dont have such knowledge about PSP, still when I look at the most polygon pushing PSP games it seems they are a bit over what the DC can achive. I always Take Dead or Alive 2 as an example of what is possible more or less for DC and as you can see its not a lot of textures there. Then when you have games with more textures the polycount goes down.

What many stil seems to overlook or ignore (I dont know which as its rather easy to calculate it) is DC GPU need to store the framebuffers, textures and display list in the texture memory. When that is full thats it. Do not matter how optimised that CPU stuff is. While from what I get from PSP its similar to PS2 that as long as you can feed it textures and polygon data it just keep drawing them. So if coded in a good way it can achieve more polygons.

So if level of detail comes down to polygons. Then nope I dont think so. If it comes down to a combination of textures and polygons that achieve a visual that look as detailed, then yes I think you can make DC games look on pair with great PSP and PS2 games.

But again, I have not worked on PSP so I am a bit in the dark there. Sorry :)

And on a personal note, who cares really? I mean its different machines. I am more interested in pushing as much on respective machines. So the question, could the DC be pushed more (in comparison to itself) is A LOT more interesting. Also its a question that is actually possible to answer if people would develop more for it in the homebrew scene. Same as we have seen amazing games for Atari 2600, C64, Megadrive, Intellivision etc I would love that the same happened to DC :)
see though your assumption of doa2 is wrong there. it actually uses alot of texture in memory. so the stages have their texture compressed in a .bin blob thats slightly less 1mb for each stage/stage layer. the actual textures are vq mipmapped textures around 70 per stage layer, so lets say the clock tower which has a top stage and a bottom stage thats like over 140 textures, alot of textures being 256x256. Same with the characters, they use around 40 to50 textures each, ranging from 32x32 to 256x256. if we were to do non-vq pvr version of all these textures there probably wouldnt be enough vram to hold it. polycount on average stage layer seem to be around 10k to 15k triangles and characters 8k to 9k triangles. True doa2 is the exception on dc rather than the norm.

psp assets are rather interesting, they arent that far apart from dc stuff at all. more or less the same range from the samples ive seen. theses a bigger difference when looking at ps2 vs dc than psp. if you dont feel like going through psp disc for assets just use jpcsp , rips things from the screen with textures to obj format.

which kind of bring me to my next question, 15k or less on headhunter seems on the low end though especially for 30 fps. considering ive found games assets who player character/car a third of you polybudget for the fullscreen of headhunter or certain games run 12k-18k triangles @ 60 fps.
 
Funny you should mentions just those, as its them I was directly involved with (and a bit Wolfpack HQ) as they where the first environments we had in the game so later on when the project and company grew I unfortunately could not do any more hands on work.

And yeah it proves how much art direction helps. I think we was quite early also starting to use photos over handpainted textures. That made it look more realistic I think. We had bought two 1.3mPixels (was A LOT back then) cameras and two guys in the art team went to LA for two weeks to shoot around 10k photos. DC was great in that photos worked really well with the PVR compression. No limits to color and as the packing could make them a but noisy was less visible as the photos themselves was a bit noicy by nature :)

Regarding prerendered I also think we was quite early in baking in lighting not just into vertex colors, that had been done quite a lot PS1 era but directly into textures. So we basicallt lite the scene in 3DS max where we had many object mapped to textures space was used over all object. It took a bit more texture use but allowed us to, especially on floors, bake lighting into kind of lightmaps. Then of course we compromised. Like those leather chairs in Angelas place have the same texture lighting on both with a little adjustment of vertex color. And then we added fake reflections (shiny floor) nd a lot of transparency (curtains etc) to make those things look more real. Same with the greywolf bossfight. Its just particle systems and animated textures for the rain FX (and mirrored objects for the fake reflection) I mean objects reflect but Grewolf and Jack do not.. still few seems to mention or notice that :D
The falling rains itself is just a cylinder with an raindrop texture that I attached to Jack. So the rain always is around him. And then by code the texture is animated. In that way we could change the scrolling direction and speed very easy. To simulate the direction the rain is falling due to wind etc.. I think it worked pretty well.


I think games like Shem Mue II and Dead or Alive pushed it pretty far tbh. A lot of people compare DC with the gap of early and late PS2 games but the PS2 is a completly different beast. Also there is memory limitations that actually caps the DC from reaching such hi poly count (simple version: the polygons to draw takes place in DC graphics memory so to achieve the combination of varied detailed textures and tons of polygons like say GOD of WAR 2 is not possible)

That said I think it would have been possible to make better games for sure. I mean we learned a lot from Headhunter, and with better art tool and direction we could make something LOOK more even if it wasnt more if you get what I mean?
So yeas I think games that used DC smarter and maybe even was optimised a little little better would have been possible. But it would not have been huge differences.
I just want to say that you guys did a great job in Headhunter.

Like most people, I’m obviously particularly fond of the rooftop boss fight, very impressive-looking scene at the time. As I recall, you also used Modifier Volumes for the dynamic searchlights on the boat level; this was a nice touch as well, most DC games didn’t bother getting creative with MV. I generally really like the scenes where the team seemed laser-focused in playing the "light vs shadow" card. Scenes like these give Headhunter a more "modern" look compared to other games in DC’s library. Given that lighting was predominantly baked back then, it also showcases the importance of good art/visual design in providing "good-looking lighting".

(YouTube capture warning)
Surprisingly visually pleasing given its age IMO (despite that "old-school" seam under the door :p).​
HH.png
That said I think it would have been possible to make better games for sure. I mean we learned a lot from Headhunter, and with better art tool and direction we could make something LOOK more even if it wasnt more if you get what I mean?
It’s all in the rendering techniques IMO. Techniques which allow you to increase, not the raw quantity, but the quality of your on-screen polygons. 1998-2000 developed DC games didn’t get to benefit from these techniques because game development was still stuck in the PS1/N64/SAT era in terms of knowledge, tools and methodologies.

Case in point, some racing games during that generation use what I assume is an environment map, with basic UV calcs, to apply fake "specular" on the road. Believe it or not, I actually think it looks extremely cool, even today lol. It’s amazing what good old environment mapping can do when used creatively in key areas of the scene. As far as I know, there’s nothing that prevents the DC from doing that. Except of course for the fact that DC died too soon and this particular technique wasn’t applied / popularized until Gran Turismo 3 (IIRC). Since you’re making a racing game, please, try to implement this!
Also there is memory limitations that actually caps the DC from reaching such hi poly count (simple version: the polygons to draw takes place in DC graphics memory
You probably won’t remember after all these years, but did you guys ever experiment with PVR’s Strip Buffers to compensate for this?

From what I can grasp of the docs; instead of storing a full 640x480 double-buffered frame buffer in VRAM, Strip Buffers "compress" it at 256, 128 or 64 screen lines and render the screen in 2, 4 or 8 "parts" respectively, in-sync with the video out, allowing you to conserve lots of VRAM. Was it too troublesome to implement?

There was a rumor back in the day that Soul Calibur used Strip Buffers (but it was, probably, an urban legend).​
 
i was quoting what you said on some youtube video about 15k triangles on screen.
Yes that was what I assumed :)

though your assumption of doa2 is wrong there. it actually uses alot of texture in memory. so the stages have their texture compressed in a .bin blob thats slightly less 1mb for each stage/stage layer. the actual textures are vq mipmapped textures around 70 per stage layer, so lets say the clock tower which has a top stage and a bottom stage thats like over 140 textures, alot of textures being 256x256. Same with the characters, they use around 40 to50 textures each, ranging from 32x32 to 256x256. if we were to do non-vq pvr version of all these textures there probably wouldnt be enough vram to hold it. polycount on average stage layer seem to be around 10k to 15k triangles and characters 8k to 9k triangles. True doa2 is the exception on dc rather than the norm.
i was quoting what you said on some youtube video about 15k triangles on screen.


see though your assumption of doa2 is wrong there. it actually uses alot of texture in memory. so the stages have their texture compressed in a .bin blob thats slightly less 1mb for each stage/stage layer. the actual textures are vq mipmapped textures around 70 per stage layer, so lets say the clock tower which has a top stage and a bottom stage thats like over 140 textures, alot of textures being 256x256. Same with the characters, they use around 40 to50 textures each, ranging from 32x32 to 256x256. if we were to do non-vq pvr version of all these textures there probably wouldnt be enough vram to hold it. polycount on average stage layer seem to be around 10k to 15k triangles and characters 8k to 9k triangles. True doa2 is the exception on dc rather than the norm.

psp assets are rather interesting, they arent that far apart from dc stuff at all. more or less the same range from the samples ive seen. theses a bigger difference when looking at ps2 vs dc than psp. if you dont feel like going through psp disc for assets just use jpcsp , rips things from the screen with textures to obj format.

which kind of bring me to my next question, 15k or less on headhunter seems on the low end though especially for 30 fps. considering ive found games assets who player character/car a third of you polybudget for the fullscreen of headhunter or certain games run 12k-18k triangles @ 60 fps.
I think we might be talking about differenth things or well not exactly the same.

Maybe this is a better way to express it:

- We have the VRAM. Either we balance it equal or more textures or polygons (assuming the CPU could feed more than possible) All adds up together so you might come to a point where you need to downscale the space for textures to fit more polygons. Or if you want to have a lot of unqie big textures, you could not fit enough polygons (even if you could theoretyically T&L enought)

I mean if we look at DOA2 versus saw Soul Caliber its pretty obvious that its less textured. Part of it is the visual style, but that style is also done I think due to the Tech Restrictions. They favored more clean and hires texture vs Soul Caliber had more varied and "detailed" textures. For Shen Mue they have a lot of texture needed for the look. Also of course they cant hit so high poly count due to other things taking CPU time, so for that type of game it works pretty well for a balance of lots of textures, decent amount of polygons + stuff to do with the world etc. So in the end they use both the CPU power, the VRAM pretty PVR wise though those bigger non MIPMAP textures do not use the GPU the best. But if they would have opted for mipmap and more lower res textures for sure the world would not have as varied textures.

psp assets are rather interesting, they arent that far apart from dc stuff at all. more or less the same range from the samples ive seen. theses a bigger difference when looking at ps2 vs dc than psp. if you dont feel like going through psp disc for assets just use jpcsp , rips things from the screen with textures to obj format.

Yeah PSP I am like a lost kid regarding so I take your word for that. Speaking about ripping. I need to talk and get some help from you there. I am totally new to that so if you could get me started it would be a great favour. I tried to PM you but I seem not to get how to do it. If I click on your profile I cant find any button to PM/DM you. Maybe as a newcomer I dont have rights to do that yet. So if you are up for a direct chat please do message me directly :)

which kind of bring me to my next question, 15k or less on headhunter seems on the low end though especially for 30 fps. considering ive found games assets who player character/car a third of you polybudget for the fullscreen of headhunter or certain games run 12k-18k triangles @ 60 fps.

Yeah maybe my memory is just bad. I bet you know better than me about it. Or if it was like that it proves that our engine was not so good. I am sure that the render stuff was ok but a also know a lot of the AI game logic code was really slow which of course sadly affected how much stuff we could render.

Btw when you are talking the amount of tris. We are talking about culled right? so not the whole amount of tris inside the camera frustrum?

From my memory eah enemy was around 800 tris. and usually it was max 3 on screen so say worst caste 2.5k K for those
Jack I think was with effect etc... like 2k? maybe 2.5k (including the modifier volume shadow)

Then I think usually the game world that was left after culling max 5k tris. Plus UI and particles etc.. so yeah I think 10k is not to far off. But again I can be very wrong.
From my memory it was like 10k and often in the frustrum it would have been 15-20k tris if we would not had culled them. And then maybe the whole loaded level would have been of course more.
 
I just want to say that you guys did a great job in Headhunter.

Like most people, I’m obviously particularly fond of the rooftop boss fight, very impressive-looking scene at the time. As I recall, you also used Modifier Volumes for the dynamic searchlights on the boat level; this was a nice touch as well, most DC games didn’t bother getting creative with MV. I generally really like the scenes where the team seemed laser-focused in playing the "light vs shadow" card. Scenes like these give Headhunter a more "modern" look compared to other games in DC’s library. Given that lighting was predominantly baked back then, it also showcases the importance of good art/visual design in providing "good-looking lighting".​
I am not sure, I think it actually just was a top down projected light map mesh hull but not sure.Yeah I think the main difference was that we baked a lot of lighting into textures. While others just used vertex color light baking. Of course our apporached used a lot more texture memory as it meant we could not reuse textures in several places but many meshes had their dedicated meshes with the baked lighting. But it did exactly what you said. Play with light and dark.

Case in point, some racing games during that generation use what I assume is an environment map, with basic UV calcs, to apply fake "specular" on the road. Believe it or not, I actually think it looks extremely cool, even today lol. It’s amazing what good old environment mapping can do when used creatively in key areas of the scene. As far as I know, there’s nothing that prevents the DC from doing that. Except of course for the fact that DC died too soon and this particular technique wasn’t applied / popularized until Gran Turismo 3 (IIRC). Since you’re making a racing game, please, try to implement this!
Exactly. I mean that is basically the same as used for any type of reflections. Either fake lightmap on top. or if you want to control what part of a texture that is specular you can use it as a mirror box under the road. And its excactly what I plan to use in the racer. If you handle this with code you can also do it a bit smarter and optimized. This tech was underused. It can be used for so many things. And i fillrate is an issue sometimes it even works with Punch through tris on DC. Also lets not forget that DC supports offset/additive vertex color so you can use that also where the effects do not need to be so detailed.

FakeSpec (GoogleDriveClip)


You probably won’t remember after all these years, but did you guys ever experiment with PVR’s Strip Buffers to compensate for this?

From what I can grasp of the docs; instead of storing a full 640x480 double-buffered frame buffer in VRAM, Strip Buffers "compress" it at 256, 128 or 64 screen lines and render the screen in 2, 4 or 8 "parts" respectively, in-sync with the video out, allowing you to conserve lots of VRAM. Was it too troublesome to implement?

There was a rumor back in the day that Soul Calibur used Strip Buffers (but it was, probably, an urban legend)

I think we did not use it. I have no memory of such. What games did use it? If I am not wrong DoA2 do not use double buffer or maybe it was strip bufffer to be able to fit as much as possible in memory and from what I rememebr they are still bordering exploding the DC
 
I am not sure, I think it actually just was a top down projected light map mesh hull but not sure.Yeah I think the main difference was that we baked a lot of lighting into textures. While others just used vertex color light baking. Of course our apporached used a lot more texture memory as it meant we could not reuse textures in several places but many meshes had their dedicated meshes with the baked lighting. But it did exactly what you said. Play with light and dark.
A few months ago I was inspecting the Katana SDK demos. One of them was showcasing dynamic searchlights using Modifier Volumes and my brain instantly went "oh, the Headhunter team must have looked at this demo". Interesting that you did it with texture projection, One of many techniques that were underused on DC.
Exactly. I mean that is basically the same as used for any type of reflections. Either fake lightmap on top. or if you want to control what part of a texture that is specular you can use it as a mirror box under the road. And its excactly what I plan to use in the racer. If you handle this with code you can also do it a bit smarter and optimized. This tech was underused. It can be used for so many things. And i fillrate is an issue sometimes it even works with Punch through tris on DC. Also lets not forget that DC supports offset/additive vertex color so you can use that also where the effects do not need to be so detailed.

FakeSpec (GoogleDriveClip)
Yes, I agree, you can do lots of cool stuff with this (e..g. I think fake refractions work pretty much the same way?).​

Needless to say I can’t wait for the kickstarter :p

I think we did not use it. I have no memory of such. What games did use it? If I am not wrong DoA2 do not use double buffer or maybe it was strip bufffer to be able to fit as much as possible in memory and from what I rememebr they are still bordering exploding the DC
I don’t know if any game actually bothered. I’m no game dev but, assuming strip buffers are practically usable, it seems to me that utilizing this effectively in a visually advanced game would require lots of extra effort and planning (which is why I asked if it was too troublesome to implement). Most DC games didn’t bother much with optimization to begin with.​

The only strip buffer-related thing I’m aware of, is that ancient rumor about Soul Calibur (or was it Dead or Alive? It was definitely a 3D fighting game, but it’s been way too long). However, as I said, I bet it was an urban legend, much like the mythical "bump mapping in Soul Calibur".
 
Back
Top