What does this mean?
>>321483983
Read the picture, it's clear what it means
Polycount is very noticable and that image that gets copy-pasted everywhere is retarded because the model itself is not high-rest in the first place
It means that artistry comes from competent devs who put a lot of time into detailing their models, not from retards who think more polygons = better looking.
More polygons = POTENTIALLY more detail. It all depends on the developer.
>>321483983
That the importance of triangles is due largely to screen coverage. However, reduced tesselation reduces detail under close scrutiny.
There's already a tentative solution to this; go look up "sparse voxel octrees", where an octree voxelization of a mesh is raycast to determine what needs to be drawn at a given pixel.
>>321484818
Should also note that fragment shading allows for multiple fragments to contribute to a final pixel, so the issue of "diminishing returns" is also an issue of triangle and shading/shadowing overdraw. It's kind of difficult to cull micropolys that all fall in the same range of the depth buffer under a single output pixel.
>>321483983
The law of diminishing returns anon, jesus fuck.
>>321484745
This poster is retarded and trying to paint the picture to have the same meaning as the first one.
Ignore artistry or whatever because that's obvious, everyone wants goo artdirections and this isn't what the image is about, simple if you want a certain level of detail you need a certain polycount, simple as that.
Graphics are all about tricks and trying to get the best visuals without taxing the hardware much, devs will try to use all sorts of maps and low polycount and deliver beautiful looking shit, but it goes without saying that you can't make Crysis with 3 polygon models.
Anyone and bitching about something as stupid as polycount (I don't think these sort of people exist, it's just stealth HURR GRAPHICS DUN MATTER faggots) is a retard.
>>321484818
it looks like shit though
>>321485282
but that's not the "point"
>>321483983
It basically shows the amount of triangles before it doesn't matter anymore. Though good devs bake high poly on low models instead of hogging resources
>>321485110
yeah the funny thing is that polygons are much cheaper than materials nowadays and devs still didnt scale to that. instead of doing higher polycounts and wasting less memory on various texture maps they still think that putting a material of four different maps in 1024x1024 resolution is a good idea.
>>321485030
>The law of diminishing returns anon, jesus fuck.
Except for the picture says the opposite, original picture is about law of diminishing returns, but this one states that law of diminishing returns doesn't exist here.
Argue all you want about importance of visuals but this image originally was used as a retarded stealth argument to "prove" that you don't need to improve graphics after a certain point and that it doesn't change that much after a certain point, which is lies and bullshit that only mind of the most deluded consolefag could believe.
>>321483983
Diminishing returns for model polycounts
Detail in textures, normal/bumpmapping, and lighting is where the money is at
>>321485428
I like the asthetic of high poly and large texture maps, but I also think it's cool what's possible nowadays with lighting in games. Textures may not be perfect and antialiasing is lacking but the lighting done in video games now either is or is near physically accurate. Games even use databases of properties for real-world materials which is pretty cool. Go check out how BRDF materials are done; they basically take a camera rig and photograph a surfce under many lighting conditions.
>>321485586
the op image itself is DISPROVING the idea that diminishing returns applies to increased polycounts you fucking moron.
>>321485654
yes PBR is pretty much the future for next 10 years if not more, thankfully this workflow isnt hard, but it still pisses me off when people decide to put high resolution maps on anything and then wonder why framerate goes to shit on everything thats not at least 4gb card
then again, you gotta rely more on textures when your engine cant do proper lighting.
>>321485654
Except for the fact that DX11 can handle at most what 4 light sources or something.
We'll only see lightning really shine in the Vulcan DX12 days.
>>321485924
>Except for the fact that DX11 can handle at most what 4 light sources
NO
that argument is retarded as lighting depends on engine
>>321485717
Well it does a pretty bad job at doing so.
The 20000 poly mesh looks just as good as the 40000 poly in the first line.
>>321483983
Someone is arguing that the image (top line) that was meant to "prove" diminishing returns exists in 3D rendering is false because the mesh was based on the 2k model. He's absolutely right, but he's wrong if he thinks it disproves that diminishing returns exists. There's a point where you reach life-like photo realism, and from that point no amount of extra polygons would enhance the detail. All of the models used in the example are a far way from that, so it's a bad example to prove diminishing returns to begin with.
The usual counterargument is that "well, you could add more stuff to the scene" but this is only true to a certain extent, since it will always beg the question "And then what?" once you've added everything you would have added and still have left over processing power
