Loading...
> >

Render Quality

exstarsisexstarsis Posts: 2,128
in The Commons

When doing research, I saw tests that demonstrated that Render Quality did not have very much of an impact on a picture, while drastically increasing the time required for convergence. They were pretty old (in the Iray scheme of things) and I don't know if this is something that changed with the most recent release of Daz Studio, but I've found that in extremely dark scenes with only a small amount of light, upping the quality can be the difference between a candleflame in darkness reflecting off a tabletop and a bunch of fireflies and rage.

I don't have saved versions of the Quality 1 version of the attached render, but I think I hiked it up to 6 or so to get a good candle. I'm doing another render now that could illustrate the difference between 3 and 6 but I need to 1.) finish the render and 2.) make it Daz-safe first.




 

against the cold test4 candleflame.png
1920 x 1080 - 524K
against the cold test4 candleflame-cropped.png
585 x 573 - 132K

Comments

  • exstarsisexstarsis Posts: 2,128

    OK. I realize not many people care (and many are probably wondering.... why?) but... here's two pictures, one showing Render Quality 1 and one showing RQ 6. The RQ 1 render gave up before 100 iterations. The RQ 6 1 took around 900 more. When you're screen blending light layers, fireflies add up. It is, admittedly, hard to see in the smaller versions of the pictures, but if you open up the attachments in tabs you can see it.

    RQ 1:


    RQ 6

    Fountain-sidespot-bad-small.png
    1194 x 1261 - 1M
    Fountain-sidespot-cropped.png
    1203 x 1356 - 1M
  • AnotherUserNameAnotherUserName Posts: 2,727
    dreamfarmer said:

    OK. I realize not many people care (and many are probably wondering.... why?) but... here's two pictures, one showing Render Quality 1 and one showing RQ 6. The RQ 1 render gave up before 100 iterations. The RQ 6 1 took around 900 more. When you're screen blending light layers, fireflies add up. It is, admittedly, hard to see in the smaller versions of the pictures, but if you open up the attachments in tabs you can see it.

    Just out of curiosity, what was the time difference between the two rendrers? Ive always understood that render quality and noise filtering were supposed to be big factors in how well a render turns out. But looking at the two renders, I would have thought that a render quality six would have produced a much higher quality version. Yes, I see a difference in the noise, etc., but based on my own experience, the time difference it would have taken between the two renders would not justify the results. I really like iray, but I swear that it can be super twichy at times.

  • Leonides02Leonides02 Posts: 1,379
    AnotherUserName said:
    dreamfarmer said:

    OK. I realize not many people care (and many are probably wondering.... why?) but... here's two pictures, one showing Render Quality 1 and one showing RQ 6. The RQ 1 render gave up before 100 iterations. The RQ 6 1 took around 900 more. When you're screen blending light layers, fireflies add up. It is, admittedly, hard to see in the smaller versions of the pictures, but if you open up the attachments in tabs you can see it.

    Just out of curiosity, what was the time difference between the two rendrers? Ive always understood that render quality and noise filtering were supposed to be big factors in how well a render turns out. But looking at the two renders, I would have thought that a render quality six would have produced a much higher quality version. Yes, I see a difference in the noise, etc., but based on my own experience, the time difference it would have taken between the two renders would not justify the results. I really like iray, but I swear that it can be super twichy at times.

    True, but I've not encountered a render engine that doesn't share this "twitchy" quality!

  • exstarsisexstarsis Posts: 2,128

    The time difference was pretty extreme too-- 4 minutes vs 1 hour. Honestly I ran it at 6 originally and THAT completed in about 8-10 minutes the first time (around 300 iterations) and I was dissatisified so I dragged it up to 10. Then, once it was at about an hour and I was satisfied (bored), I started nudging the RQ slider down, because the way it works when you change the slider during a render is that the convergence adjusts upward or downward. I wanted to see what RQ would push the 98.7 to 100%. 7 wasn't enough; 6 finished immediately. So probably somewhere between the two. It was clear that the time diff between 6-7 was probably substantial though.

    Prior to starting to work with light layers I never did see any use in RQ. I know that post-4.9.x release, 95% convergence no longer quite hits my standards so I do wonder if RQ was made more useful as well. As for now-- I spent some time doing light layers with RQ 1 mostly-dark layers and it was just... those little pinpricks show through like crazy. My original approach to beating the 'wham bam thank you ma'am' 60 iterations and DONE render was to turn off quality and just let it render for hours until I wandered by and decided it was done. But I like this way better.

  • AnotherUserNameAnotherUserName Posts: 2,727
    Leonides02 said:
    AnotherUserName said:
    dreamfarmer said:

    OK. I realize not many people care (and many are probably wondering.... why?) but... here's two pictures, one showing Render Quality 1 and one showing RQ 6. The RQ 1 render gave up before 100 iterations. The RQ 6 1 took around 900 more. When you're screen blending light layers, fireflies add up. It is, admittedly, hard to see in the smaller versions of the pictures, but if you open up the attachments in tabs you can see it.

    Just out of curiosity, what was the time difference between the two rendrers? Ive always understood that render quality and noise filtering were supposed to be big factors in how well a render turns out. But looking at the two renders, I would have thought that a render quality six would have produced a much higher quality version. Yes, I see a difference in the noise, etc., but based on my own experience, the time difference it would have taken between the two renders would not justify the results. I really like iray, but I swear that it can be super twichy at times.

    True, but I've not encountered a render engine that doesn't share this "twitchy" quality!

    Lol! Good to know... Im glad its just not me. wink

  • AnotherUserNameAnotherUserName Posts: 2,727
    dreamfarmer said:

    The time difference was pretty extreme too-- 4 minutes vs 1 hour. Honestly I ran it at 6 originally and THAT completed in about 8-10 minutes the first time (around 300 iterations) and I was dissatisified so I dragged it up to 10. Then, once it was at about an hour and I was satisfied (bored), I started nudging the RQ slider down, because the way it works when you change the slider during a render is that the convergence adjusts upward or downward. I wanted to see what RQ would push the 98.7 to 100%. 7 wasn't enough; 6 finished immediately. So probably somewhere between the two. It was clear that the time diff between 6-7 was probably substantial though.

    Prior to starting to work with light layers I never did see any use in RQ. I know that post-4.9.x release, 95% convergence no longer quite hits my standards so I do wonder if RQ was made more useful as well. As for now-- I spent some time doing light layers with RQ 1 mostly-dark layers and it was just... those little pinpricks show through like crazy. My original approach to beating the 'wham bam thank you ma'am' 60 iterations and DONE render was to turn off quality and just let it render for hours until I wandered by and decided it was done. But I like this way better.

    I typically render at rq 1. Thats usually enough quality for when im just playing around with stuff. Ill go to two if im rendering something that I want to post but thats about it. Also, since im not very knowledgable about computer hardware and how well it tolerates heat before meltdown, I try to keep my render times at no more than 30 minutes to an hour. RQ 6 would have me nervously chewing on my nails!

  • fred9803fred9803 Posts: 1,565

    I'll quote sickleyield because I think this explains the issue quite well-

    "Rendering Quality Enable:  If this is set to OFF, Iray will just use the Max Samples and Max Time to determine when render stops.  If you want render time to go just based on those, turn this to OFF.  If it is off, the Quality and Convergence sliders don't matter.

    Rendering Quality: This linearly increases render time each time you raise the number: 2 is double the value of 1, 3 is triple the value of 1, etc.  It overrides the time and samples counts.  Use this only if you really want to try the "render forever and tell it to stop when I like how it looks" method.  Otherwise just leave it at 1."

  • exstarsisexstarsis Posts: 2,128

    I'm pretty Sickleyield's analysis is from the earlier release of IRay though. It DOES stop now and it's not a linear progression. Thus my OP.

    AnotherUserName: I don't know what kind of card you have, but graphics cards usually shut themselves down well before they melt. And they can run for many hours before that happens if they're sufficiently cooled. You can get (or may already have) software that will tell you how hot it runs.

  • AnotherUserNameAnotherUserName Posts: 2,727
    dreamfarmer said:

    I'm pretty Sickleyield's analysis is from the earlier release of IRay though. It DOES stop now and it's not a linear progression. Thus my OP.

    AnotherUserName: I don't know what kind of card you have, but graphics cards usually shut themselves down well before they melt. And they can run for many hours before that happens if they're sufficiently cooled. You can get (or may already have) software that will tell you how hot it runs.

    I have a gtx 770. My pc has sensors that will display the temp for me, I just have no idea what is considered normal and what is over-heating. Good to know theyll shut down before melting down though.

  • fred9803fred9803 Posts: 1,565
    dreamfarmer said:

    I'm pretty Sickleyield's analysis is from the earlier release of IRay though. It DOES stop now and it's not a linear progression. Thus my OP.

    I stand corrected. But I think many if not most people would be using pre 4.9 DS. So that advise might be good for them.

  • exstarsisexstarsis Posts: 2,128

    That might be true right now, but anybody who wants to use the 10xx series of NVidia cards ongoing has to upgrade to the version I'm talking about (4.9.3.166), and now this thread will (hopefully) be here when they try to search the forums (via Google, of course) for 'render quality', trying to determine what it's good for.

  • fastbike1fastbike1 Posts: 4,078

    Rendering quality changes the internal convergence criteria. However, it isn't magic. There is only so much that Iray can do with minimal light.

Sign In or Register to comment.