With polygon counts rising are you going to update your portfolio with 100k meshes?
polycounter lvl 5
Order 1886 has 100k characters, infamous second son has 120k, Ryse 80K. Are you going to update your portfolio with higher-res meshes? Do studios look at that sort of thing?
Replies
What counts is what looks good and shows competence. If your characer is movie grade and rendered over 8 hours in arnold, whatever. The result counts.
Show what you can pull off, and let the companies give you constraints later.
If you effectively use those tris, yea nice. If your character looks like last gen, and you write 100k, then its a downgrade of course.
It depends on your overall presentation. If you have killer art everywhere, people assume you know your stuff
and don't demand a lot of info. If you are borderline, people will care more about details to check your competence.
Dont forget people care mainly about the outcome, this is stronger,
the better you are, as the rest is shifting to details assumed known.
"If he used 4 textures instead of 6, that would be impressive" - no one ever
Also those 100k are for close up cinematic shots in Lod0.
Lods actually trivialize the whole portfolio polycount thing anways. Work reasonable but dont let anything constraint you too much.
120k character would have more verts on screen than pixels on your TV at the distances they are seen.
Killzone Shadowfall for example had only slightly more verts drawn on screen than Killzone 3 did. Actually it might have been slightly less verts on screen due to no longer having the 8 SPU's of the PS3 Cell.
This generation is more about physically based shaders, procedural textures/substance textures, increased particle spawning, more havok/physics based objects than actual high polycounts.
The PS3's cell had to pick up the slack for its GPU which was weaker than even the Xbox 360. PS4's GPU is rated 10x better and has a peak of 1.6 billion triangles per second (Xbox 360 was 500 million triangles).
Killzone was also a launch game made for constantly changing dev kits. Like every generation, we don't see the final results of a console till the final years of it.
The Order 1886 cinematics transition instantly with the gameplay. Those are the same models.
Ryse actually has no LODS. That's why he [Marius] dropped from 150k to 80k polygons and again, the cinematics transition instantly to the gameplay.
When it comes to polygons, and portfolios, the key isn't how many polygons i can make it with. The key is how many polygons does it take to hit the quality bar I'm try to achieve.
Tessellation is a performance killer. Very few games actually pull it off very well.
I think 100k is perfectly capable for real time. Remember the Dark Sorcerer demo was shown it had characters made of one million polygons. It was controllable and it was running on old dev kits of the PS4 so they could do even better.
But I do agree it's topology that's important.
I know a lot of Art Directors dont like seeing straight edges caused by low poly characters. 35-50k should be enought tho..
It might help if you are looking for a job!
Wouldn't it make more sense for recruiters to look for people who can work within their range of polycount? Like knowing how to use a game engine is better for a game artist... I actually started wondering this because I see many artists putting polycount on their portfolio pieces. Is that a pointless exercise? If I wanted to make my art look the best why not just render highpoly?
Topology will keep getting more refined per number of polygons increased.. Wouldn't it?
What kind of standard do you set yourself polycount-wise for this gen then? Do you look at how high it goes or should you just see what looks best?
If you're trying to get your first job, you need to show that you understand topology.
I think it depends on 2 things:
1. Budget
2. Scope
If you're making an AAA game, obviously you'll want your game to look the best. If you're making a $20 indie game, avoid doing something that will take months.
Scope is next. Does the game take place in a corridor or is it open world? Open world games tend to sacrifice some detail so they can still run. A corridor game can afford to make everything look high poly.
Sorry, it was a misunderstanding. I thought you meant tessellating in a video game. Turbosmooth/sub divison during modeling makes more sense.
I said nothing about the performance of said meshes. I don't care if they run there is no reason to spend all of that time generating a mesh of that density by hand. It is not worth it. I don't know for certain, but I am guessing there is some kind of subdivision process going on for the meshes to get to that tri count. Remeshing a 30k tri mesh takes a while as it is. Now try tripling that and tell me that is usable in a game production pipeline.
Though in my experience playing games I tend to just see those higher polycounts in dense areas on characters rather than the environments but that's another can of worms when it comes to making things run well with instancing and tilable textures and whatnot.
Proper edge loops shouldn't change all that much between something 10k tris and 100k tris normally. It's a little hard for me to explain, but eventually you get to a point where the edgeflow is all there and any more polygons will just go to essentially subdividing the mesh for smoothness rather than organized topology.
If you're considering a portfolio piece the #1 thing to do is settle on the camera distance for final presentation. If you want to show how awesome you can model all the macro details - then sure - go for it. But if you're making an asset that is a part of an environment that I'm only going to see as 20 square pixels on the screens, then spending 10k tris on it might be overkill and might raise some legitimate concerns about optimization in your work.
As long as you make it looks awesome though I think you should be good. It's easier to trim some verts then to teach someone how to make it awesome
The SPU's gave us a lot of power to push a lot of geo that we now cant do since we dont have the SPU's. This is because we knew how to really develop for the Cell being first party and not ever worrying about Xbox or PC versions.
The added power of the PS4 equals out to the power we were able to achieve due to the Cell. Cell is far more powerful than most people think, it was just extremely difficult to code for as it was a new language that most didnt know so most didnt know how to harness that power. The way we built our games/assets was far different than the Xbox that we could never make the same game on the Xbox because it didnt have the Cell.
Of course this is the start of the generation and things will get better as they always do. But the jump of polygons is nothing compared to the PS2/PS3 jump.
There's no way PS3 is equal to PS4. The new MLB The Show for both PS3/PS4 prove that. PS4 characters have more polygons in their faces than the PS3 counterpart.
I would never suggest to arbitrarily target a high tri count always have a purpose.
The same goes for texture sizes. Slapping a 4k texture on a pistol can actually have the opposite effect that is intended, it shows ignorance.
I was watching 2 games today. One of them had a higher polycount but no normal maps, and the other had a lower polycount but used normal maps and better lighting.
I actually liked the lower poly one better.
My advice is to just focus on making good normal maps. You can go crazy with the geometry and no one will care because it's going to get baked anyway.
These 2 meshes use the same amount of geometry but as you can see, the normal maps help give it a boost in wrinkles,cracks, bandages etc.
Maybe one day we'll get computers/consoles that can handle any zbrush or turbosmooth model with ease. Maybe one day....
This...
In some cases I do this with far lower limits in 3DCoat, I'll retopo something with the lowest amount I can get away with and then subdivide it with little clean-up afterwards.
Geometry is also very cheap, and for character close-ups I think we need it. Yep, games can look great with far less then 20k tri's in a face, but no amount of fancy shaders will negate a horrid faceted edge.
If your portfolio work is close to what the studio does, it will make it easy for them to see that you can do the job.
Whether you wanna work on games with 100k characters or 1k characters is up to you, though.
Normalmaps have always been used to convay internal detail. Higher polycounts means we can have more detailed silhouettes, and *potentially* better looking animations due to having smoother bends etc. but normalmaps will always be important. you could have a 500k character and i'd still say to use a normalmap, because just like the image you showed, you will still need to convay subtle surface detail and internal things that just don't show up in the silhouette.