Home General Discussion

Can you tell which render is using 4k textures?

FrankPolygon
grand marshal polycounter
Offline / Send Message
FrankPolygon grand marshal polycounter
Here's an image with two renders. Both were done in Toolbag 3 at 1920x1080, using the same (100x) render sampling values, lighting setup and model. The only difference between the two is the texture size. (One texture set was down sampled to 2k. Mip correction was enabled.) Which is which?


This is an informal poll (for fun) based on a recent discussion. Feel free to pixel peep and compare the renders before voting. After a week I'll post the answer. Thanks!

Replies

  • Alex_J
    Offline / Send Message
    Alex_J grand marshal polycounter
    4k texture is on the right.
    looks like a little more detail in the specular highlights on the right?

    but even if the texture is 4k, if the textures/images you used to create that textures weren't 4k (like the grunges and any images used), would it make a difference?
  • Ashervisalis
    Offline / Send Message
    Ashervisalis grand marshal polycounter
    @Alex Javor Nah, you can tile those lesser-than 4k textures to increase their resolution. I think the main issue in identifying the 4k image is this rendered image is 1800x900 pixels, and we're trying to identify which model is 4k textures. The render resolution would need to be higher to identify that. Unless I'm totally wrong, which I usually am :P
  • PixelMasher
    Offline / Send Message
    PixelMasher veteran polycounter
    4k texture is on the left.
    Im guessing the left is 4k, but given the fact its a 1080 render, chances are you would never be able to see the difference as it is pixel bound by the resolution anyways.  Im also looking at it on a phone screen. 

    Even on a 4k 75 inch tv, chances are even a fps game asset like this that the player interacted with right up to the camera would still be 2k max, as it would not take up 100% of the screen pixels.  Like 30% of the screen pixels max even up close to the camera like the player picking it up and putting it on like in a CoD game etc. 

    Even in the current example the difference is unnoticeable to the human eye to the point where the results are split 50/50 currently, so definitely not worth 4x the memory footprint than a 2k 😂 a technical artist would be at your desk bringing the fire and brimstone the second that was in the build haha. 
  • Blaizer
    Offline / Send Message
    Blaizer polycounter
    4k texture is on the left.
    I voted for the left, but in my 4K monitor, with such a small image (1798x915), it's very hard to see any difference. The mic looks a bit different, and some speculars aswell.

    Could you show the texture sheet for the model?

    Too many people avoid UV optimization and a correct use of UV space (it takes time and knowledge). At the end, they waste too much space and they usually need a 4k, 2k, texture when in fact, they only needed the half. I'm sure you have done a very clever use of uv space and mapping :+1: .
  • oglu
    Offline / Send Message
    oglu polycount lvl 666
    Could you upload a 4k PNG without compression?
    And im not sure if we see the real 4k textures. Does Marmoset mipmap? I fear we dont the the highest mipmap. 

    Edit: I reread your post. One image is downsampled to 2k.

  • Tekoppar
    Offline / Send Message
    Tekoppar polycounter lvl 10
    4k texture is on the right.
    I'm assuming that the right is the 4K one and the left is the downsampled texture due to the fact that the left mic is a lot more noisier. But it might as well be the other way around.
  • defragger
    Offline / Send Message
    defragger sublime tool
    4k texture is on the right.
    the right is 4k. No doubt.
  • Joopson
    Offline / Send Message
    Joopson quad damage
    4k texture is on the right.
    I feel as though the right one is 4k, but functionally in this use-case, there's no meaningful difference, so I wouldn't place bets personally.
  • Panupat
    Offline / Send Message
    Panupat polycounter lvl 17
    I don't even know what's real and what's rendered anymore these days.
  • Finnn
    Offline / Send Message
    Finnn greentooth
    4k texture is on the right.
    both 2k ?
  • EarthQuake
    oglu said:
    Could you upload a 4k PNG without compression?
    And im not sure if we see the real 4k textures. Does Marmoset mipmap? I fear we dont the the highest mipmap. 

    Edit: I reread your post. One image is downsampled to 2k.

    Toolbag uses mip-mapping in the viewport. For rendered images, mip-map biasing is applied depending on large the image is and the sample count. For high sample counts mip-mapping is effectively disabled.

    So it depends if this is a viewport screenshot or a rendered image.
  • wirrexx
    Offline / Send Message
    wirrexx quad damage
    4k texture is on the left.
    Left one is the 4k That's my two cents. I feel like the shadows are softer on the left one for some reason and the inner part of the microphone seems less jaggy. And there is a white spot on under Peltor text which seems more visible then on the right side.
  • throttlekitty
    Offline / Send Message
    throttlekitty ngon master
    4k texture is on the right.
    Oh no now I don't know what to think! didn't want to cheat for the vote, but took it into photoshop for comparing afterward. I could see the stark difference in the foam, and interpreted the deeper contrast as coming from less detail + mipmap sharpening, as well as the large scratch on the left ear cup.

    Anyway, the point remains that the difference is practically nil at this distance from the camera.
  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter
    4k (image size) renders using the same assets and render settings as the 1080p (image size) renders.



    4k (image size) closeup renders.



    Poll results as of 8am EST is 42 votes and an even 50/50 split.


    Answer: In all of the images (including the first 1080p render) the 2k textures are on the left and the 4k textures are on the right.

    Methodology:
    All renders use the same 3D assets, lighting setup, render settings and mip correction. One model uses 4k (4096x4096) texture files and the other uses texture files that were down sampled from 4k to 2k (2048x2048).

    All comparison images have a 1:1 pixel size. Renders were cropped, without resizing or sharpening, to remove empty space and place both renders side by side in one image. Uncropped renders were captured at 1920x1080 (1080p) and 3840x2160 (4k).

    Other thoughts:
    This was an informal survey (for fun) and the spirit of the initial poll was to get an off the cuff response based on the limited information provided by the post and the sample image. A 1080p render size was chosen for the initial poll because: it's the most common resolution reported by the STEAM hardware survey, Artstation (according to their documentation and unless you have a Pro subscription) converts images to normalized 1920xN JPGs and other social platforms have similar or smaller image size restrictions.

    So far there's been some great technical discussion but looking at it from a less technical standpoint: does this result have any broader implications for situations where non-artists and non-technical people are reviewing portfolios?

  • Eric Chadwick
    Did you answer which is which, or did I miss that somewhere?
  • DavidCruz
    Offline / Send Message
    DavidCruz interpolator
    The split even was a pretty neat ending to this, fun little discussion and hopefully an eye opener as well.
  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter
    Added an answer section below the poll results in my second post: 2k textures on the left and 4k textures on the right.
  • Alex_J
    Offline / Send Message
    Alex_J grand marshal polycounter
    4k texture is on the right.
    but what is the truth though. you're killing us
  • sacboi
    Offline / Send Message
    sacboi high dynamic range
    4k texture is on the right.
    FrankPolygon...but looking at it from a less technical standpoint: does this result have any broader implications for situations where non-artists and non-technical people are reviewing portfolios?

    Well, in my 2¢ opinion, just taking into account this poll's result alone. If for example lets say, a temp from HR was handling initial submissions via a scripted set of criteria for eventual final review by the interviewing panel, then probably not?

    Because at a glance there's no discernible difference, aesthetically, between the two images textured or otherwise. In other words 'it looks good'

    EDIT:
    But I get the point though. For the uninitiated, nuanced artistic/tech expertise portrayed by a particular piece, would be next to impossible too finesse/assess, amongst a horde of applications choking their bandwidth.         
  • rollin
    Offline / Send Message
    rollin polycounter
    4k texture is on the left.
    @FrankPolygon Why is the noise kind of upscaled on the 2k version?
  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter
    The interpolation method used to down scale the textures has an impact on how high frequency details in the normal map are handled. Here's what the various methods look like when the 4k normal map texture is down sampled to 2k.


    The default (automatic) sampling was used because that's just what mode Photoshop happened to be in when the test was run. After testing the other scaling methods it seems like Nearest Neighbor had the most appealing results. Here's what the textures look like after being down sampled using nearest neighbor interpolation. The 2k textures are on the left and the 4k textures are on the right.


Sign In or Register to comment.