By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Nintendo Discussion - New Pokémon Legends Arceus trailer

sc94597 said:

As far as I can tell, mesh quality hasn't been changed between the 3DS and Switch versions of MHGU, so if we are looking at total geometry of any given scene, the Switch version should be representative. Textures of course are better, but that is a given since the reason why developers go with cel-shading in the first place is because it goes well with lower-resolution textures. 

I never intended to suggest that the cutscenes in Pokemon are pre-rendered. What my point was -- was that a cutscene is going to show the best asset quality and asset density of a game because 1. the camera is limited and therefore render-loads are reduced to the small-scoped visible scene and 2. the developer has control over asset density within this limited camera view. So a scene like this looks much better than the average scene in Sword and Shield precisely because it is scripted, is indoors, and if it didn't it would be very noticeable.  

I can do the same thing with MHGU. Here is a scripted in-game sequence. Notice that the asset density and quality is much better than usual for that game.  

When you make the bolded statement, what are you basing it on? For example, when Sword and Shield released the game was data mined and polygon counts for many Pokemon were shown to be the exact same as in Sun and Moon. This isn't as bad as it sounds because the asset density of any given scene in Sword and Shield is better than Sun and Moon (you don't have 2d backdrops in battles, the battle field isn't a circle with a small radius, etc) but when looking at the actual wireframes I am not seeing what you see here. 

There are about 5,000-10,000 polygons in any single Pokemon/character model in the 3DS era games and Pokemon Sword/Shield. 

Comparatively Monster Hunter monsters in Monster Hunter Tri (which should be similar to Generations Ultimate as assets are reused between Monster Hunter games, like they are in Pokemon games) are about 10,000 - 15,000 polygons with characters and minor enemies having about 6,000 - 8,000 polygons -- roughly similar to Pokemon Sword and Shields characters but with a higher maximum. Similar can be said of Monster Hunter Stories and other 3DS games. Compare that to say Daisy in Mario Tennis Aces who has 20,000 polygons or Morag from Xenoblade Chronicles 2 who has 17,000 polygons. This is what you should expect for a character in a Switch game. What then matters is asset density (how many polygons in a given render-load.) This is what I mean when I say Monster Hunter Generations Ultimate and Pokemon Sword are roughly comparable. 

Source for character polygon counts: Download the models from here and open them up in 3D Paint, switch to wireframe. https://www.models-resource.com/

Camera panning is important because limited camera perspectives allow one to incorporate many tricks to get away with much lower model density and to reduce the render-load. For example, in Sun and Moon you have pretty much just the character models and a small circular field being rendered. Not much else. In the Gamecube games there are more polygons beyond the Pokemon being rendered (as it is fully 3D.) This means that even though the Pokemon have fewer polygons in the Gamecube games (about 3-5 times fewer) the actual render has more assets in it, meaning total polygon count is probably roughly the same in any given scene. 

I didn't say the camera didn't move at all, my point was that the degrees of freedom in Sun and Moon are much fewer than in the GameCube games and this is for a reason. Compare the respective motions of the cameras in these two battle scenes. Consider that the total scene is almost always present in XD while Moon switches to blank back-drops as the Pokémon animate attacks. When that happens, pretty much the only thing being rendered is the Pokemon model. Also a good moment to look at how limited animations are in the newer Pokemon titles. 

In the characters I can say they are about par for the course for a 3DS game. Most 3DS games have characters in the 5000 - 10,000 polygon range. I don't agree if we are talking about total polygons rendered though. While characters have a normal number of polygons Game-Freak uses tricks to reduce the polygon load of environments and to limit character count in active scenes. These are tricks you can't do in an Action game like Monster Hunter. 

Bold 1: Texture work can mask polygonal flaws. I was not stating that geometry changed. 

Bold 2: You can go into that exact building in Pokemon Sword and Shield and none of the assets have changed. The only thing I could see changing here is resolution if there is dynamic resolution scaling (pretty sure there is) and it wasn't hitting the top end. There are not any unique Sonia models for cutscenes. It uses the exact same model. 

Bold 3: I actually don't see an overall change in asset quality there. Polygonal models look exactly the same as far as I can tell. 

Bold 4: I actually understand this. Many of the Pokemon (not all, but a substantive amount) have the exact same meshes as Sun and Moon. Part of the reason why Sword and Shield are disappointing graphically compared to other Switch games. However, the point in this here is that Sun and Moon's Pokemon assets are surprisingly high polygon, especially compared to other 3DS games (See: Bold 5 and Bold 7). So, when you take in the fact that Pokemon Sword and Shield have far more complex geometry for the environment and character models (even those battle scenes) you can see how it's not fair to suggest that it's comparable. It isn't. Sword and Shield could not run on a 3DS even if you lowered the resolution to 200p. Period.  

Bold 5: I just looked at most of MH Tri's character models and only the bosses and the player model have anything above 7,000 (10k seems to be about the highest). Minor enemies have around 1,000-3,000. Not only is this part disingenuous, but the idea that the 3DS can match the Wii's polygonal output is ridiculous. Super Mario Galaxy's Mario model is around 5,000 polygons where the 3DS's Super Mario Land is less than 2,000. There are many instances of this. Pretty much across the board for almost all like for like comparisons, the Wii wins out. Except when it comes to Sun and Moon's Pokemon models. In this sense, Monster Hunter Generations models are more likely than not much less dense. Even Rathalos in Monster Hunter Stories is less than half of Monster Hunter Tri, and Monster Hunter Stories is a turn based game where more resources can be spent on models. 

Bold 6: The stadiums simplistic geometry does not equate to ten thousand or more polygons in a frame. Not only can Pokemon character models be gargantuan in comparison polygonally (see: Bold 7), but even the characters themselves are incredibly more dense (Sun/Moon: almost 5,000. Colosseum: almost 1,000). This is complete exaggeration on your part. 

Bold 7: I'm not sure where you are getting this information. Pit in Kid Icarus has less than 3,000 polys. Mario in Mario Tennis has less than 3,000 polys. Incinaroar alone has 16k polys. Even Pokemon like Caterpie have over 4,000 polys. MOST 3DS games do not do this. See: Bold 5.

Last edited by Doctor_MG - on 25 August 2021

Around the Network
Doctor_MG said:

 

Bold 5: I just looked at most of MH Tri's character models and only the bosses and the player model have anything above 7,000 (10k seems to be about the highest). Minor enemies have around 1,000-3,000. Not only is this part disingenuous, but the idea that the 3DS can match the Wii's polygonal output is ridiculous. Super Mario Galaxy's Mario model is around 5,000 polygons where the 3DS's Super Mario Land is less than 2,000. There are many instances of this. Pretty much across the board for almost all like for like comparisons, the Wii wins out. Except when it comes to Sun and Moon's Pokemon models. In this sense, Monster Hunter Generations models are more likely than not much less dense. Even Rathalos in Monster Hunter Stories is less than half of Monster Hunter Tri, and Monster Hunter Stories is a turn based game where more resources can be spent on models. 

Bold 6: The stadiums simplistic geometry does not equate to ten thousand or more polygons in a frame. Not only can Pokemon character models be gargantuan in comparison polygonally (see: Bold 7), but even the characters themselves are incredibly more dense (Sun/Moon: almost 5,000. Colosseum: almost 1,000). This is complete exaggeration on your part. 

Bold 7: I'm not sure where you are getting this information. Pit in Kid Icarus has less than 3,000 polys. Mario in Mario Tennis has less than 3,000 polys. Incinaroar alone has 16k polys. Even Pokemon like Caterpie have over 4,000 polys. MOST 3DS games do not do this. See: Bold 5.

I am going to address 5-7 right now, since this is merely a matter of fact-checking. 

I've exported the models from a Monster Hunter XX 3DS Rom and was able to load them into 3D Max with the use of a plugin that allows one to import MT Framework .mod files. 

Here are the counts of some enemies.  

6,406 triangles, 8,801 vertices (note the model is in two parts the tail + the body.) I believe this one is Rathian. I am unsure which enemy is which because they're numbered. 

6,526 polys, 5,785 verts

Polys 7,485, Verts 7,406

4,521 Polys, 4,156 verts

Polys 12,167, Vert 10,903

7,555 polys, 8,882 verts This is obviously Barroth. 

Poly - 10,728, Vert 8,235

Poly - 11,152, Vert 11,017

Polys 10,420, Verts 8,332 This is Atoraku Ka

Polys - 11,332 , Verts - 10,475

4,284 polys, 4,737 verts

Also note that the following are the counts for MH Stories monsters. You can confirm them with the built in 3D modeler in windows. 

Brachydios: 11,306 polys, 33,918 verts

Kushala Daora: 6,118 polys, 18,354 verts

Nargacuga: 4,223 polys, 12,669 verts

Rathalos: 4,800 polys, 14,400 verts

Blue Yian Kut-ku: 6,864 polys, 20,592 verts

Another game with comparable models is Super Smash Bros for 3DS

Zelda: 5,462 polys

Bowser 5,927 polys

Bayonetta - 6,704 polys

I can go on all day with more examples. 

5,000 - 10,000 polygons is typical for late-generation 3DS games. I think I provided enough evidence of that. 


Last edited by sc94597 - on 25 August 2021

Would people be interested in a thread just for posting these wireframes of models from Nintendo games? I've wondered for 7 years what the poly counts were for the models in Smash for 3DS and it turns out that they are higher than I thought, looks like about the same as Brawl.



I mean, if we're gonna compare apples to apples this is a high end open world game on 3DS, one that can only run on the New model:

And this is Legends Arceus:



mZuzek said:
curl-6 said:

I mean, if we're gonna compare apples to apples this is a high end open world game on 3DS, one that can only run on the New model:

And this is Legends Arceus:

I feel like the "apples to apples" comparison here is the one where you show two games from the same platform... honestly even the New 3DS version of Xenoblade looks better than Legends Arceus apart from the crappy resolution. Thankfully there's a Switch version, and uhhhh

Then you think about how one of these is a niche franchise that barely sells one or two million per entry, and the other one is the highest grossing multimedia franchise in the world and constantly puts out 8-digit sales.

Yeah, no. The 3DS version of Xenoblade falls well behind Legends not just in resolution but in lighting, models, textures, details, the works.

My point was simply that comparisons to 3DS graphics are unfounded.



Around the Network
mZuzek said:
curl-6 said:

Yeah, no. The 3DS version of Xenoblade falls well behind Legends not just in resolution but in lighting, models, textures, details, the works.

My point was simply that comparisons to 3DS graphics are unfounded.

Maybe, but in my opinion the art direction easily makes up for it. The only jarring thing for me is 240p. Meanwhile in Pokémon everything looks bad, crappy graphics, crappy art direction.

Well, I mean art direction is always going to be a matter of personal taste. Could Legends look better? Absolutely. I just think it's going a bit far to say it looks like a 3DS game or even as I've seen elsewhere to compare its graphics to PS2 or even N64.



Another one



curl-6 said:

Another one

I think it actually looks better now, dare I say, it looks pretty good. Especially the water looks good now.



Still on the fence on whether to pick this up or wait for an inevitable sequel on better hardware. This is one of the few 1st party Switch games which makes me crave a Pro.



I'm surprised at the backlash. I think the game looks good. I will be grabbing it.