PS2 vs Dreamcast Graphics

[Cloofoofoo]

it certainly impresses, I just didn't mention it because I was more focused on humanoid models even though some car polygon numbers were mentioned in my previous post

Sometimes is hard because each game is so different. Even from platforms you think it would be faster. Check these polygon counts and bone counts on this humanoids . These are ps2, psp, dc. You would be surprised that dc is doing compared to the competition bone wise but i feel its best to just to stick to polycounts. Btw you want human polycounts heres one. The main character model seventh cross (dc)usually is about 4.5k triangles. Regular enemies range from 700 tris to 1.3k. Bosses around 3.5k to 4k. Stages range from 4k to around 16k. So about on par if not a little more than with twin snakes i guess but at 30 fps. Another one would be the gameplay model for julian on tee off golf dc, he is 6,000 around, strange enough at the point tally screen if everyone is him there will be 4 of them displayed, so 6k x4, the engine definitely can do better than what it was used for. Animal wise suzar vasar the elephants replay model is about 7.5k triangles, not counting weapons or slay.

[Cloofoofoo]

Virtua Tennis 2/ Tennis 2k2 - Dreamcast version

Tedious to extract, reminds of resident evil 4 or the mercenaries 3d in that it uses a ton of morph target for hands. Uses a lod on heads for upclose or gameplay. Small number of stages had less detailed version meant for doubles( 2vs 2) that removes some details from the arena. Some of the shadows are modeled in, smart. These are pretty high numbers considering they pan out to show the whole stage before and after the match.

Serena williams ( hi lod): 2,183 triangles


Haas(gameplay lod): 1,758 triangles


USA: 30,156 triangles


Secret stage: 28,846 triangles


Rio( doubles ): 17,225 triangles

[Cloofoofoo]

18 wheeler - Dreamcast version

I knew the game looked pretty , I guess because it uses a decent amount of polygons. The trucks are about up close to 2,000 triangles. The trailers seem to range from 3,000 to 4,000 after you get rid of duplitcate meshes. Thats alot for just the truck + cargo. Single player arcade features very long tracks that are very detailed but does has noticable distance culling. traffic cars range from about 100 to 500 triangles.While parking mode tracks are much lower and dont distance cull. Theres also screenplay tracks that around the same in detail compared to parking mode. This is a naomilib game so I used that to extract it.

Asphault cowboy - 1,790 triangles


gas trailer - 2,872 triangles


Arcade mode stage 1 - 97,046 triangles


Arcade mode stage 4 - 48,287 triangles


splitscreen mode stage 4 - 14, 359 triangles


parking mode challange stage 3 - 13,345 triangles

[SegaAMD]

Sometimes is hard because each game is so different. Even from platforms you think it would be faster. Check these polygon counts and bone counts on this humanoids . These are ps2, psp, dc. You would be surprised that dc is doing compared to the competition bone wise but i feel its best to just to stick to polycounts. The main character model seventh cross (dc)usually is about 4.5k triangles. Regular enemies range from 700 tris to 1.3k. Bosses around 3.5k to 4k. Stages range from 4k to around 16k. So about on par if not a little more than with twin snakes i guess but at 30 fps.

I also did maken x and it was interesting, reminds me of twin snakes with how its designed and its a 60 fps game. Arm models are around 1.5k and cutscene models are 2.5 to 3k. Stage segments are from 4k to 8.5k. Bosses are around 2.5k to 3k with final boss being 6.5k. Regular enemies range from 1.2k to almost 2k. Imagine had they gone for 30 fps instead

.

I noticed you compare polygon counts as proof that such games would be possible on the Dreamcast (Maybe they could who knows) but something got me thinking, as you may know MGS2 was developed for the PS2 using all the strengths of the ps2 including physics, particles and frame buffer effects all this aiming for a 60fps target (which is solid for most of the game) see, not even the Xbox did well with this type of game, being in this specific game worse in textures, buffering effects and frame rate, being better only at 480 resolution. Explanations range from Konami's accusations of poor programming to the fact that the ps2 have a higher fill rate, there are many doubts as to why the game was not at least equal, but I don't want to focus on this version. Twin Snake is a MGS1 remake using features present in MGS2, since when I saw it I didn't know the polygon count but I knew it was a graphically inferior game to MGS2 even though it was exclusive to a more powerful console, a 2004 game. I mean there is a reason for these games are so demanding,you know these consoles have more powerful cpu/gpu/memory/media . I have a hard time imagining a 30fps Dreamcast version of Twin Snake, as the GC game itself despite being 60fps clearly spends most of its time closer to 30fps than 60fps. I think this type of game cannot be underestimated, MGS: TT is one of the few GC exclusive games to run in 512x448 instead of the traditional 640x480. these are some details that need to be considered in addition to polygon counts.

[TrekkiesUnite118]

From what I remember, MGS2 and Twin Snakes aren't really doing anything too crazy in the realm of polygon counts, frame rate, resolution, etc. if you're talking about a theoretical port of them to the Dreamcast. Where things might get tricky would be certain lighting and transparency effects, depth of field effects, reflections, bloom, bump mapping, etc. While not impossible on Dreamcast, they can be a bit more challenging and demand more resources. But, you may be able to cut some of those effects or replace them with something similar.

[Leynos]

It was just using the MGS2 engine no?

[Cloofoofoo]

.

I noticed you compare polygon counts as proof that such games would be possible on the Dreamcast (Maybe they could who knows) but something got me thinking, as you may know MGS2 was developed for the PS2 using all the strengths of the ps2 including physics, particles and frame buffer effects all this aiming for a 60fps target (which is solid for most of the game) see, not even the Xbox did well with this type of game, being in this specific game worse in textures, buffering effects and frame rate, being better only at 480 resolution. Explanations range from Konami's accusations of poor programming to the fact that the ps2 have a higher fill rate, there are many doubts as to why the game was not at least equal, but I don't want to focus on this version. Twin Snake is a MGS1 remake using features present in MGS2, since when I saw it I didn't know the polygon count but I knew it was a graphically inferior game to MGS2 even though it was exclusive to a more powerful console, a 2004 game. I mean there is a reason for these games are so demanding,you know these consoles have more powerful cpu/gpu/memory/media . I have a hard time imagining a 30fps Dreamcast version of Twin Snake, as the GC game itself despite being 60fps clearly spends most of its time closer to 30fps than 60fps. I think this type of game cannot be underestimated, MGS: TT is one of the few GC exclusive games to run in 512x448 instead of the traditional 640x480. these are some details that need to be considered in addition to polygon counts.

And I noticed when you dont like what I said you try to manipuiate in very stupid ways by ignoring what I actually said. I brought up more than just polygons by letting you know i found that dreamcast actually uses tremendous amounts of bones, more than you would think for a console with "weak" cpu as people like to believe( i seen games that if you think about it, it can have like 500+ bones for a scene in game.). These are very math / cpu intensive when animated and yet it still has room to run physics , TNL , ai and a decent rate at that. Look atthe bottom pic of the elephant from zusar vasar on dc, this ran real time with physics and all that on with 5 other character models on a track, and included the bone count .

Theres a reason i Feel twin snakes is a better fit for Dc than mgs2 would have been. Not only are the stage assets lower but it doesnt bother with high detail cutscene models like the 2 did. Also the top down is actually really good for DC since you can cull anything out of the screen( and the dc seems to this most efficiently). so lets say the highest stage i could was 9k but only 50% of that stage is visible then only 4.5k will be rendered, this not taking into account if its worth it cpu wise to use backface culling to further cut that number down. Lets say snake + weapon is 3.5k tops and lets downgrade the grunt enemies to around 1700 triangles x4(4 enemies) which is mgs 2 numbers( 1.3k to 1.5k but iam adding a few hundred more for weapons). which is what around 15k for the entire scene and lets say run it at 30 fps its 500K a second. While this is a simplification you can see how this leaves alot of room for everything else like physics ,ai and so on. Leave post process effect for cutscenes and stick to blob shadows, 1 or 2 directional lights. Also twin snakes doesnt seem to be doing any outrageous physics , Mostly IK on snake and hitboxes on enemy to load the animation per they were struck. Ive seen snakes collision model , its just a extremely simple box models. The point isnt to match the gamecube but to bring the game to dc standard which in this case this game isnt the challenge you make it out to be, if anything it might be one of the best case for dc.

Have you even seen the final boss for twin snakes, takes place on top of a broken metal gear that is around 5k triangles for the stage which both snakes being 3,000 and is really zoomed in, perfect for culling. This should give you an idea how the rest of the game is, they really underachieved compared to part2 but thats kinda a given since it was programmed by silicon knights using mgs 2 engine. I think at worst it might just be the fact its 2 GC disc which would make it around 3 gdroms.

7.5k triangles:

[SegaAMD]

ok, I just made an impossibility assumption, you guys on the other hand accept that it's possible, so that's ok we agree to disagree regarding the possibility of a dreamcast version MGS:TTS.

[Cloofoofoo]

Daytona 2001 - Dreamcast - Amusement Vision / genki
Game has alot of car body duplicates for lod and damage modeling, internal view. Very competent game, one of the view games that uses environment map heavily.

Secret online car- 1,408 triangles
daytona2k1.jpg


Secret rocket car- 1,183 triangles
daytona2k2.jpg


Seaside Galaxy track - 48,352 triangles
daytona2k3.jpg



Mermaid lake track - 25,400 triangles
daytona2k4.jpg

[SegaAMD]

https://youtu.be/ng4a2AmDnOk?t=64


I've never heard of this game. maybe it's worth ''measuring'' the geometry of both
you know if this is possible

[TrekkiesUnite118]

https://youtu.be/ng4a2AmDnOk?t=64


I've never heard of this game. maybe it's worth ''measuring'' the geometry of both
you know if this is possible

Honestly I'm not sure the polygon counts are what's having the biggest impact on performance on Dreamcast. If you pay attention a lot of those drops happen when there's any kind of Transparency somewhere on screen or about to come on screen. Once that transparency is no longer on screen, the frame rate jumps back up.

[SegaAMD]

Honestly I'm not sure the polygon counts are what's having the biggest impact on performance on Dreamcast. If you pay attention a lot of those drops happen when there's any kind of Transparency somewhere on screen or about to come on screen. Once that transparency is no longer on screen, the frame rate jumps back up.

to me it's a processor bottleneck, you know 60fps requires a lot of cpu and in some parts the fps drops because of that, It is normal for many engines to have one map and another one that forces more cpu usage 1:56 2:42 2:58. in my experience in PC games, whenever I wanted to gain better cpu performance I turned off the shadows, the Dreamcast version is very smart, it has reflexes turned on before the race starts but turns off during the race, the sparks are also turned on only during the passage through the tunnel at 1:21. the geometry cuts are acceptable and do not influence the experience, the AI cars has no shadows or sparks on the Dreamcast, in general I liked the port I think they explored the Dreamcast well, not as well as Sega did on the F355 but without a doubt better than the other 60fp racing games on the platform.

Edit: I watched the video again, you may be correct, the frame rate drops a lot during scenes with transparencies, if I'm wrong, the dreamcast gpu has difficulty with transparencies.

[Cloofoofoo]

Vermilion Desert - Dreamcast
Interesting slow strategy game. you seem to be able to keep large groups of armored soldiers, i saw 11 of them. They other soldiers, tanks and even big mechs( multiple of them).

on of the player character( no armor) : 803 triangles


armored character : 998 triangles


another armored : 1,034 triangles


giant mech : 1,707 triangles


another giant mech : 2,165 triangles



stage : 9,668 triangles


another stage : 8,849 triangles

[Cloofoofoo]

Ranger Mission - Atomiswave/ Dreamcast
Light gun shooter with very detailed stages. 2004 atomiswave/dreamcast game

Player/squad soldier - 1,346 triangles


regular enemy - 528 triangles


helicopter enemy - 1,593 triangles


embassy exterior - 18,514 triangles


embassy interior - 23,037 triangles

[Cloofoofoo]

Demolish Fist - Atomiswave/Dreamcast - 2003
A nice looking beat them up , well modeled even with the polycount. By dimps. Can have many enemies on screen. Reminds me alot of psp stuff ive extracted.

Chimps( regular enemy) - 700 triangles


Berserker(regular enemy) : 931 triangles


Bloody Tumor Alkanide( mid boss) : 3,559 triangles


Senator Kain Murasame (final boss phase 1) : 1,278 triangles


8 ( final boss phase 2) - 3,806 triangles


Leoneed Burns(player, cutscene model ) : 2,704 triangles


Leoneed Burns (player, gameplay ): 1,200 triangles


Stage 1 Appollonia , part 2 : 9,468 triangles


Final boss phase 1 room : 11,522 triangles