Chaos Corona Forum
Chaos Corona for 3ds Max => [Max] General Discussion => Topic started by: dfcorona on 2019-04-28, 00:14:18
-
Wanted to get you guys opinions on Vray Next since most of you came from Vray or still use it but also use Corona. We own Vray 3x version and Corona among other renderers. We are going to keep using Corona but might use Vray Next for animations. Had some observations and questions.
We just used Corona 4 nightly for an animation, render quality, user interface & user friendliness is fantastic. Render times get to be quite large compared to Vray Next that we tried for a few small tests, and that can really add up for animation. Vray Next IPR seems much more responsive and final rendering seems to be much faster. But my question is about quality, many Corona renders look very realistic compared to Vray. Not sure if this changed with Vray Next, but I can't place my finger on if it's Coronas lighting or shaders. Seems to have more depth and reflective properties seem more natural. Like I said we had only limited time with Vray Next trial, so for those who use both would love to hear your opinions or even see you comparisons.
-
We just ran through a scene with Next to see how some new GPUs performed. It did render verrrrry fast and in my opinion the viewport didn't lag nearly as much during IR with Forest Pack. For us Corona has really gotten slow with FP distributions during IR.
That's the good news. Bad news was we found the reflection\refraction in glass to lack a certain 'depth' and realism. Multiple people commented on it and we just couldn't get it where we wanted. Hard to put your finger on but it just didn't seem right. Corona always gets glass 'right.' Also the denoiser leaves something to be desired. If you have fireflies it creates these huge white squares. I understand this is going to be fixed soon. And if you're using Next GPU the denoiser seems to really screw up glass as well. It's a bummer, because since the denoiser has issues, even if you render for hundreds of passes to avoid having to denoise, there still will be some grain and it won't be even across the image so it doesn't look like film grain. It'll be totally clean in one spot and then (usually behind glass) noisy as heck. Perhaps this has to do with adaptive dome light sampling, it just seemed like we kept hitting walls when it came to final quality issues. I was also disappointed to discover 2D Displacement isn't supported on the GPU, one of VRay's best features.
We think we might end up using Next GPU for certain types of animations. We'll see. FStorm is really nice too and, aside from lacking a denoiser, not supporting becrontiles, etc, is a pretty good GPU option for certain types of scenes.
Just our\my impressions. Jumping back into Corona was kind of a relief. Hope that's helpful and I'm sure others will chime in. Cheers!
Daniel
-
That was great danio1011, thanks for your experience. This is what I'm kinda finding also. I am now putting together a scene with Vray Next and then will remake it in Corona. Vray Next GPU is very fast, which is nice but it also is so slow in development. Missing many features you can get with say Redshift. IPR if it's Vray CPU or Gpu seems much snappier than Corona and I've noticed that issue too with Forest Pack. But the scene I'm putting together with Vray the render is just getting that plastic feeling. Before I rendered with materials I did both scenes in vray and Corona with override shader and the lighting between both where quite different, this is an exterior scene. I know what your saying about depth, Corona feels like the scenes have actual depth where Vray Next feels flat. I also would like to use Next GPU for certain types of animations but the quality just leaves me kinda disappointed so far. I'm not done with my own internal testing but will see what comes of it.
I've tried Fstorm while nice I always hit Vram issues, so it's only ever good for small scenes. Octane does better but is limited in shaders and ALWAYS crashes over all the years we've owned it we were never able to use it for production once except one still that had it crashing all the time also, We've owned from the beginning and own the latest versions also. For GPU nothing really beats Redshift. Would love to hear more from others.
-
dfcorona - Are you using Redshift for arch viz scenes? I've never tried it but maybe now I will. Interesting you run out of VRam with Fstorm, I've found it to be super efficient in that regard. I think it's getting NVlink soon, so that should help.
Cheers,
Daniel
-
Yes Redshift for Archviz, It's super fast and never had a issue with vram. It is missing some key features though but that is being resolved a lot in version 3.0 that's coming soon. It can handle massive scenes. I always run into vram issues with Fstorm, especially with exterior vegetation, and it has no OOC. Nvlink will help but i still see it having Vram issues even then with very large scenes.
-
I have to agree with the opinion of most users here. In Next the IPR is much more responsive, it's faster then ever.
What I can not understand are the comments about the palstic look. Attached are some renders straight out of the VFB, with applied LUT. No further post work whatsoever. Looks good in my opinion...
-
Thank you kosso_olli for your info and pictures. I agree that Vray can render cars realistically. But find some interiors or exteriors with different materials and vegetation. They seem to always show a plasticy look from Vray to me, now that we are use to the realism of say Corona and Fstorm. There might be a handful of them that look so good I can't tell there Vray. Every time I see a render from Vray now I can tell it is Vray, it has a flat plastic feel. Characters and cars always seem to come out good. I think one major issue with Vray is how it handles Vegetation also, It always seems to feel less quality than other renderers. I had to take a break because of work but I want to get back to my testing between the two.
-
Well, the old Alfa from the side has vegatation in it, the grass, wall and the hedge. Same for the Mazda: Racetrack, Grass, Hills etc. are all rendered.
The Dodge in the fog also, it is full CGI. Droplets the same. The only car on photo background is the BMW. So I think V-Ray can render any environment in a realistic way...
But here are some more I did with Forstpack and some assets from Megascans.
-
Those are hard to tell anything because Photo-scanned objects can look just as real in a game engine. I'm talking full scenes interior and exterior, the majority of them don't look as realistic as corona. Like I said I cannot pin it down to Lighting, Shaders, or possibly tonemapping. I know that I've heard a lot of people also complaining how vray handles Bump mapping, so maybe it is a shader issue. But I have to say again it's IPR and render speed is pretty amazing.
-
Erm, as I said, besides the BMW, all images are full cgi exteriors.
-
Nice images! I had good luck with vegetation, particularly megascans stuff. And I was very, very pleased with the interactivity with Forest Pack and IPR. For us the issue really was the glass, particularly indirectly lit glass. Aside from the denoiser making a bit of a mess of it, even when not denoised the refraction\reflections in glass didn't seem to have the look and depth we'd expect from Corona or FStorm. I showed a couple of colleagues the scene we were working on without mentioning we were using VRay and they both said, independently, something along the lines of 'Nice! But the glass doesn't have a lot of depth...?' A short time after that I noticed on the FStorm facebook page Grant W. commenting on how he was much more pleased with FStorm's refraction than VRay's. He seems to like VRay a lot from what little I've seen of his recent stuff, so I thought that was significant and somewhat relevant to our independent experiences. Not that his word is Gospel or anything, I just thought it was worth mentioning.
It's kind of an aside, but we also encountered 4+ serious bugs in one week (VRay Camera randomly loses vertical correction, Max had to be restarted. Bucket rendering randomly skips half the rendering, had to switch to progressive. Denoiser generates large white squares on fireflies. Render regions occasionally not being respected, not serious but annoying. Switching to material override generated an error, Max crashed. Running out of VRam often makes Max crash.) I reported most of these to CGroup and they were able to reproduce a couple of them so that's good! I just never really see that level of bugginess with Corona or FStorm. So while the renderings went faster and interactivity went up, production actually slowed due to some of these issues.
All that said, we're still considering using VRay Next on certain types of jobs. I think there is potential there for us, this is just an unpolished representation of our experiences. Heck...I spent a healthy chunk of change on a couple of 'test' GPUs, I would like to use them occasionally :)
-
Erm, as I said, besides the BMW, all images are full cgi exteriors.
I guess I'm looking at this from more of a Archviz point of view. I understand that the motion blurred exterior of the mazda is cgi, but hard to tell anything form well all the blurring. The cream side car is only 3 elements in the exterior where the rest is a photo. They look good, but when you have all elements of a full cgi archviz project, things seems to start looking well CGI. I think the best is to have the same archviz project rendered in both, I'm trying that now and hopefully be pleasantly surprised by Vray, but so far not exactly convinced.
-
Regarding the glass issues: There recently are discussions to increase the default Max depth parameter, as well as the reflect on backside option for the VrayMtl. These two options are well known to long-time V-Ray users, but new guys have no clue what they are. Both are neccessary to get good looking glass in any renderer.
The bugs you mentioned are unknown to me. Never had the bucket mode skip half of the image, nor does it crash when assigning an Override Material. Guess some stuff is down to the user.
But I can tell you something else: In my 12 years of V-Ray, I have seen a lot of renderers come and go. I tried all the promising ones (yes, including Corona), but so far no competitor ever matched the speed and the set of supported features. There is literally nothing I could not do with V-Ray.
-
Yes vrays supported features are amazing. Wish they would develop faster with gpu though, development on that is so slow.
-
Regarding the glass issues: There recently are discussions to increase the default Max depth parameter, as well as the reflect on backside option for the VrayMtl. These two options are well known to long-time V-Ray users, but new guys have no clue what they are. Both are neccessary to get good looking glass in any renderer.
The bugs you mentioned are unknown to me. Never had the bucket mode skip half of the image, nor does it crash when assigning an Override Material. Guess some stuff is down to the user.
But I can tell you something else: In my 12 years of V-Ray, I have seen a lot of renderers come and go. I tried all the promising ones (yes, including Corona), but so far no competitor ever matched the speed and the set of supported features. There is literally nothing I could not do with V-Ray.
Reflect on backside was on (I always toggle this to evaluate results with glass). I usually think of max trace depth as needing to be bumped higher with more complex glass situations than doublepane archviz glass, but perhaps it's worth a try. With refraction I think of this as relating to thicker or more complex geometry, particularly when you start to see exit color show up (is that still a thing?) I would also imagine that Grant would be aware of those settings, too, when making his comments that so closely aligned with our impressions. I've been using VRay since 2009 at this point, so I'm not totally unfamiliar with the engine.
As for the bugs, I think it's just idiosyncrasies of user patterns and machine configurations, which are innumerable. The fact that two bugs (so far) have been reproduced by CG gives a bit of credence there, and I haven't seen that instability in other engines. But again, it could be pure bad luck. It also seems to be a known issue that in GPU the denoiser doesn't treat glass very well, so that is a bit of a downer for archviz.
I'm not saying VRay doesn't have features or potential, I'm just relaying our impressions. It sounds like it's treating you well. It certainly has lots of features and is fast!
-
Since you guys are avid Vray users, what is going on with Normal maps? Do they not work correctly or am I doing something wrong? Below is two renders one is Vray and the other is Corona, you can see how much more realistic Corona handles the Normal Maps. Especially in the Stucco.
Vray I'm using the Vraynormal map node.
EDIT: Okay so the stucco looked more correct when I imported the Normal map as gamma 2.2 or automatic, which makes no sense because all other renderers look correct when It's imported as Gamma 1.0
-
To me the normal issue seems to be related to the red and green channels, and/or gamma. There is a little tip here by Vlado, which should give you a hint. It seems the Corona version is the correct one, but I think it can be fixed.
If you could share the scene I could help you out.
Also, do you have any issues with the glass in this particular setup? Looks good to me!
-
To me the normal issue seems to be related to the red and green channels, and/or gamma. There is a little tip here by Vlado, which should give you a hint. It seems the Corona version is the correct one, but I think it can be fixed.
If you could share the scene I could help you out.
Also, do you have any issues with the glass in this particular setup? Looks good to me!
I had to flip the Green channel to get correct. Played with both gamma and channels, still get weird results. Let me package the scene it's about 500MB. Glass no, in Vray I always do reflect back side and turn up the reflection/refraction depth to 25. I do notice sometimes that the glass doesn't behave correctly still, but then all of a sudden it will render correctly, almost like a bug or something, but it's random.
Attached is something interesting, render comparing Corona and Redshift.
Corona render: 26min 36sec, was only maybe 1/4th done with rendering so expect 1-1.5 hours render. But Corona also ran out of Ram gave me info about needing more ram in pop-up message.
Redshift render: 5min 44sec, Render was complete. Had no message or issue with Ram let alone Vram.
-
You can tell that Redshift does heavy clamping by looking at the highlights in the glass. In the Corona version you get glare through the refraction, in Redshift you don't. It is one of the engines which cut many corners to gain render speed. You decide if that's the right way...
-
You can tell that Redshift does heavy clamping by looking at the highlights in the glass. In the Corona version you get glare through the refraction, in Redshift you don't. It is one of the engines which cut many corners to gain render speed. You decide if that's the right way...
That might have to do with the fact that Redshift transparency and reflection refraction depth is not high, nor can it be set higher. This is going to be fixed with 3.0 and suppose to be even faster doing it. Also Redshift uses 4.0 Max subsample intensity, I can try changing that to 25 like corona and see what happens.
Vray scene: https://drive.google.com/file/d/1PjdcY9oQFshSPGCEBuSRspLgie9Xbkd-/view?usp=sharing
-
You can tell that Redshift does heavy clamping by looking at the highlights in the glass. In the Corona version you get glare through the refraction, in Redshift you don't. It is one of the engines which cut many corners to gain render speed. You decide if that's the right way...
Okay here is a new render at 25 subsampling intensity. Looks good to me now and the weird thing is it actually rendered faster, 5min 24sec opposed to 5min 44sec. What do you think?
-
kosso_olli, Did you figure out the Vray normal issue?
-
I tried to download your scene, but it seems I have to request permission for this?
-
I tried to download your scene, but it seems I have to request permission for this?
sorry, you should have permission now.
-
Yes, I have.
But sadly I can't open the file, because we are still on Max 2017. We do not use any newer version in production, because we do not update yearly for pipeline integrity.
Most recent version I have installed would be 2018. Can you save with backwards compatibility?
-
Also, do you have any issues with the glass in this particular setup? Looks good to me!
I actually see a lot more detail in the Corona glass, in the highlights but especially in the darker areas. The VRay glass just looks 'duller' to my eye. Maybe just a tonemapping thing?
-
Yes, I have.
But sadly I can't open the file, because we are still on Max 2017. We do not use any newer version in production, because we do not update yearly for pipeline integrity.
Most recent version I have installed would be 2018. Can you save with backwards compatibility?
Here is 2017 version:
https://drive.google.com/file/d/1aeA5Z8rfh0U2_iBp6PuOYU15N3UrT8AZ/view?usp=sharing
-
Yes, I have.
But sadly I can't open the file, because we are still on Max 2017. We do not use any newer version in production, because we do not update yearly for pipeline integrity.
Most recent version I have installed would be 2018. Can you save with backwards compatibility?
Here is 2017 version:
https://drive.google.com/file/d/1aeA5Z8rfh0U2_iBp6PuOYU15N3UrT8AZ/view?usp=sharing
Erm, I still can't open that one up. The last file from your download, vray material tests 2017.
Did you make sure you saved with backwards compatibility to Max 2017?
-
Yes, I have.
But sadly I can't open the file, because we are still on Max 2017. We do not use any newer version in production, because we do not update yearly for pipeline integrity.
Most recent version I have installed would be 2018. Can you save with backwards compatibility?
Here is 2017 version:
https://drive.google.com/file/d/1aeA5Z8rfh0U2_iBp6PuOYU15N3UrT8AZ/view?usp=sharing
Dammit, I saved as 2017 then I archived it to zip. Unless 3dsmax resaves as 2020 when it archives.
Erm, I still can't open that one up. The last file from your download, vray material tests 2017.
Did you make sure you saved with backwards compatibility to Max 2017?
-
Reflect on backside was on (I always toggle this to evaluate results with glass).
You also need fog to be active to be able to get (total) internal reflections.
I would also imagine that Grant would be aware of those settings, too, when making his comments that so closely aligned with our impressions
I wouldn't assume.
What you describe in the above is pretty much the definition of "confirmation bias".
-
Reflect on backside was on (I always toggle this to evaluate results with glass).
You also need fog to be active to be able to get (total) internal reflections.
I would also imagine that Grant would be aware of those settings, too, when making his comments that so closely aligned with our impressions
I wouldn't assume.
What you describe in the above is pretty much the definition of "confirmation bias".
Fog was active because I was emulating non-low iron glass. And you might be right about interpreting Grant’s comments, I could see why you’d say that. I don’t see that it changes the results we saw though.
For us it was the combo of the ‘look’ of the glass and also the denoiser not treating glass very well. The combo of those two left us going ‘ergh...that’s a bit sticky’ since most of our scenes have glass. I did open a support ticket a few days back, hoping to get it sorted since we have a few underutilized GPUs right now :)
-
Okay kosso_olli, this one should work or I give up.
https://drive.google.com/file/d/15vdOrfUGX1XbT3gRC0oEllHV3JcKatsn/view?usp=sharing
-
Well the Vray GPU seems to render the Normal map correctly, but not the CPU Version.
-
Reflect on backside was on (I always toggle this to evaluate results with glass).
You also need fog to be active to be able to get (total) internal reflections.
Never new you need to have fog enabled for accurate internal reflections, is that so even for Corona?
kosso_olli, Here is a render side by side to show how Vray advance doesn't render Normal Maps correctly, but Vray RT does. This is with Normal map loaded with gamma 1.0
-
I tried your file.
I can not get it to work correctly, either. I have to agree the GPU/Corona version is looking much more detailed and crisp, and also the lighting on the right side is looking much nicer. More correct for the direction the lighting is coming from.
-
@dfcorona: Can you send you scene to the developers over in your thread at the Chaosroup-Forums? I am curious what comes out of this, because this seems to be a bug.
-
@dfcorona: Can you send you scene to the developers over in your thread at the Chaosroup-Forums? I am curious what comes out of this, because this seems to be a bug.
I did post the scene yesterday in the forum, maybe I should also send to there support email.
-
Of one of the guys came up with the strangest solution to this problem: Apply UVW map in box mode and it works. I tried it, and it does... Weird!
-
Of one of the guys came up with the strangest solution to this problem: Apply UVW map in box mode and it works. I tried it, and it does... Weird!
yeah just saw that, lol. Weird solution, but quite the issue for a leading renderer. Hopefully they fix it.
-
What exactly does Corona do with it's shading? You can see here I used a diffuse shader on Corona vs Redshift and for some reason the corona one looks more detailed, but it also looks like it has specular on even though reflections are off. Vray renders just like the Redshift one also.
-
Isn't this something about tone mapping? It looks like in Corona you can see the highlights and in the other example they are clamped.
-
Isn't this something about tone mapping? It looks like in Corona you can see the highlights and in the other example they are clamped.
Maybe, It just seems like Corona has highlights in the shader even though it's diffuse. Here is another example.
-
Hi. What exactly do you mean when you say reflections is off? Is reflection gloss set to 0, or is reflection level set to 0?
If reflection gloss is set to 0 while reflection level is set to 1, then in Corona there'll still be some slight reflection when IOR is higher than 1. The only way to turn reflection off effectively in Corona, is to set reflection level to 0.
-
Hi. What exactly do you mean when you say reflections is off? Is reflection gloss set to 0, or is reflection level set to 0?
If reflection gloss is set to 0 while reflection level is set to 1, then in Corona there'll still be some slight reflection when IOR is higher than 1. The only way to turn reflection off effectively in Corona, is to set reflection level to 0.
I mean reflection level is 0, so no reflection. But diffuse still seems to render with what looks like specular highlights in Corona.
-
Any chance of sending the scene? Could be something odd about normals, or some sort of material override going on. Does it happen too with, say, a simple sphere?
-
Isn't this something about tone mapping? It looks like in Corona you can see the highlights and in the other example they are clamped.
Maybe, It just seems like Corona has highlights in the shader even though it's diffuse. Here is another example.
This definitely looks like the material has glossy reflections enabled. The other possible difference could be down to the way corona handles diffuse materials, there was some talk about how other render engines handle rough diffuse surfaces (Lambertian shading).
-
Any chance of sending the scene? Could be something odd about normals, or some sort of material override going on. Does it happen too with, say, a simple sphere?
Simple sphere looks primarily the same, little difference on shading falloff. Where do you want me to post the scene, don't think I can post publicly because the tree is a paid asset. If it was the normals wouldn't all other renderers render the same.
-
Any chance of sending the scene? Could be something odd about normals, or some sort of material override going on. Does it happen too with, say, a simple sphere?
Simple sphere looks primarily the same, little difference on shading falloff. Where do you want me to post the scene, don't think I can post publicly because the tree is a paid asset. If it was the normals wouldn't all other renderers render the same.
The private uploader sends it just to us, just for testing purposes - https://corona-renderer.com/upload
TY in advance for sending in the scene :)
-
Any chance of sending the scene? Could be something odd about normals, or some sort of material override going on. Does it happen too with, say, a simple sphere?
Simple sphere looks primarily the same, little difference on shading falloff. Where do you want me to post the scene, don't think I can post publicly because the tree is a paid asset. If it was the normals wouldn't all other renderers render the same.
The private uploader sends it just to us, just for testing purposes - https://corona-renderer.com/upload
TY in advance for sending in the scene :)
Okay I sent the scene. Thank you.
-
Hi guys, this is a really interesting thread. I have noted your opinions on both Vray and Corona and I will soon post a little poll about both software.
Meanwhile, can you express in short answer why do you prefer Corona or Vray? And if you should describe each of them in one sentence, what would you say? Good or bad, any answer is valuable.
-
Hey everyone, we made a little survey. here it is https://bit.ly/CoronaVraySurvey and here is the relevant forum thread. https://forum.corona-renderer.com/index.php?topic=24858.0
-
Hi guys, this is a really interesting thread. I have noted your opinions on both Vray and Corona and I will soon post a little poll about both software.
Meanwhile, can you express in short answer why do you prefer Corona or Vray? And if you should describe each of them in one sentence, what would you say? Good or bad, any answer is valuable.
For me it mainly like this :
Corona : lightmixer, vfb
VRay : speed, ram consumption, GPU support.