Vray Vs Maxwell Render Comparison Essay

This article is my personal opinion, this question comes up a lot, so I’m going to try break it down here. This article is for people without any allegiances. I will attempt to break down why i made the choice that i made and depending on your situation which render engine would be good for your needs.

[update]: cycles4d has been released, and while i haven’t had much time to test it yet, it looks like my new number one choice. partly because its developers are the same as xparticles and its all integrated, but mostly because they do a great job of documenting features, which makes the whole process a lot easier. So stay tuned for updates.

 

[update 2]: redshift is also now out of beta for cinema4D, and seems to be a very soild competitor. ( too much stuff to test) :)))

Contenders

Vray – the grand daddy of 3rd party render engines for c4d.

  • Big user base. Like REAL big
  • lots of material presets
  • many/most features of c4d supported

cons:

  • too many settings ( this is promised to be fixed in next release)
  • no gpu support, so not as speedy as others (also should be fixed next release

Arnold – The new big gun.

  • Can handle astronomical sizes of scenes without problems
  • Supports lots of features straight off, including hair, xparticles, openvdb volumes from houdini.. etc..
  • if you are planning on working in production houses this is the way to go.

cons:

  • aquired by autodesk…  ( chance of being discontinued development for other applications)
  • you’re really going to need a render farm for everything. its very slow to clean up noise.
  • also no render node licenses, but really only go for this render if you are relatively free with your budgets.

Octane – The first GPU render, with super fast growing popularity

  • lots of features supported. especially turbulenceFD
  • very large user base, and growing rapidly

cons:

  • otoy don’t sell render nodes, so you will never find farms that support it. they promised their own farm solution, but its not out yet. And even then there will be one farm, so if something happens to it, if its down, you have no alternatives.
  • only gpu render, so if you run out of memory on your video card, there is no work around, no fall back.
  • aggresive online activation system:  for example this quote from an online chat: “what’s with octane today, said i have to deactivate, and i did, said i had to wait 60 minutes, and i did, then i try again and says deactivate and i did, NOW 60 FUCKING MINUTES AGAIN!!!!!! AHHHHH I’m losing my shit”

[edit]: Octane just announced availability of new version: https://home.otoy.com/octanerender-3-and-roadmap-update/ This may be a pretty big shift. If i find the time i’ll go experiment with it. I like a lot of the stuff they brought up there. eg. $10/month for up to 2 gpus. and then just pay for their render farm is an approach i kinda like. But still the aggressive licencing might prevent me from using it on jobs.. will experiment see how it goes. 

[edit2]: ORC octanes render cloud/farm thing is live. It renders right from c4d. Awesome! BUT you need to upload orbx files. which is pretty much alembic.. which means all the animation is baked.. which means a 10mb file can become a 22gb file which you have to upload… fun…

Thea render – First and only render engine to combine CPU and GPU.

  • uses all available cpu’s and gpu’s together. If GPU runs out of memory, can just keep rendering on CPU.
  • render nodes available. results in farms being available.
  • substance material converter.
  • material repository which is cross platform. so thea users that make materials for sketchup, max etc, can share those materials and they can be used in c4d.

cons:

  • relatively small userbase at the moment.
  • still behind octane in feature support, for example turbulenceFD, motionblur and some other small issues. ( read my thea review for more details)
  • doesn’t support xparticles colors like arnold does.
  • lacks node based material editor

Here are some more renders, that ive heard very little about, but i felt like i should mention them: Corona, some say its as fast as gpu renders, but on cpu. I tried, i disagree, but then again im using a 4790k, maybe on a 12 core its actually blazing fast. but so far, haven’t heard anything great about it. Indigo render, only seems to have an exporter. so its not a native implementation, also plugin doesn’t seem to have been updated since july 2015. Redshift in an biased GPU engine, that is rumoured to be heading for c4d. but hasn’t yet been released. Maxwell is basically Octane/thea but only on CPU, so it makes really pretty things, but its kinda slow. Im not sure what is market position is right now.

 

Biased vs unbiased

You might here these terms regarding render engines thrown around. In short. Unbiased render engines are brute force light calculations. Generally most unbiased render engines will have a very similar look and feel to them. So Octane, Thea, maxwell.. etc. While biased means that the render engines cheats in order to increase speed. so it throws more power at the areas with more details, and less at smooth surfaces, and then averages stuff. This is the reason you get splotchy renders and flickering in animations. Modern days biased render engines have ways of overcoming these shortcomings, with relatively little effort, but nonetheless, its still an issue sometimes.

 

So lets wrap this up

Once again, this is targeted at someone who is just looking at 3rd party renderers. And mostly this is my opinion and specific to my situation. If money for farms is no obstacle, i would probably recommend arnold. If you can afford it, as it stands right now arnold + cash = awesome. It supports the most c4d features can handle any scene you can throw at it and is used by top studios so it’ll probably help if you are intending to get employed. For me this is not the case. Im not looking to get full time employment or work for hollywood movie productions and i need more speed. If arnold introduces GPU i will re-evaluate my options.

 

Now on to Octane  i personally like choice, so im not going to lock myself into a company that doesn’t let me use 3rd party farms. Right now they do not have a farm service, they are building their own one, but still it will be just one farm. So if its down, you’re up a creek. Also it means you can’t build your own mini farm at home without paying for a full license for every computer you use. ( octane has no render node licenses). Also if your gpu runs out of memory, you have no choice, you can’t do anything about it. that’s it. either optimize your scene or re-do the render in another engine.

 

Vray  im actually very excited to see what the new vray4c4d can do. It sounds REALLY promising. but i need something now, and well it’s not avaible right now. The current version is not GPU accelerated, so its not really on my radar for that reason.

 

And well this brings me to Thea. It runs on both CPU and GPU, it comes with 2 render nodes. If your gpu runs out of memory you can just render on c4d ( albeit a bit slower).  Its crazy fast, recently hired c4d specific developers which have been doing an awesome job, the speed improvements have been phenomenal with every release. Has 3rd party render farm support. Since it uses both CPU and GPU, it can clean up noise really fast, as it uses the strengths of both architectures, GPU is good for cleaning up areas of uniform light due to its 100-1000’s of cores and cpu for cleaning up smaller details.  And to boot it’s cheaper than the competition. So if your projects don’t need turbulence fd or complex xparticle color control its a really solid option. more info in my review.

So there you have it. If you have any amendments you think i should make please do tell me, im only human and have only so many hours in a day to keep up with all this.

Cheers,
Aleksey

P.S. I am a beta tester of thea render, but i receive no compensation for writing this article.

 

CGTalk > Main > General Discussion > Renderman vs. Vray vs. Arnold.


PDA

View Full Version : Renderman vs. Vray vs. Arnold.


DataMachina

08 August 2015, 10:45 PM

Hi, I am new to rendering software.

I'm thinking about learning either Renderman or Vray. Is there any clear differences between these software? I'm looking to do 3D animations and still images.

I guess if you could share your perspective on what your preferred rendering software is and why, that would be great :D


majstorovich

08 August 2015, 01:58 PM

I dont think you will get many answers here cause " vs threads " are not welcome.
But, you can look at something like this, it should get you started :

8dppdhUejwE


mister3d

08 August 2015, 08:07 PM

If you are an individual without programming background, forget about renderman. Also it's almost dead, that's why they give it for free afaik.
The difference is huge. Renderman was built in times of no raytracing, and its days are over.
Vray is a polished tool with all you may need. It doesn't have the upper quality of Maxwell, but it's quite a good tool. Corona is another good renderer, especially if you are just starting, and quality is a bit higher than Vray.
For everyday needs - Vray or Corona (I'd prefer Vray though as it's faster). For quality stills - Corona or Maxwell (easy to learn). I never tried Arnold, but I think it's an interesting renderer.


Laa-Yosh

08 August 2015, 08:20 PM

I don't think Renderman is dead at all... It's just existing in a more competitive space now.


pingking

08 August 2015, 08:38 PM

If you are an individual without programming background, forget about renderman. Also it's almost dead, that's why they give it for free afaik.
The difference is huge. Renderman was built in times of no raytracing, and its days are over.
Vray is a polished tool with all you may need. It doesn't have the upper quality of Maxwell, but it's quite a good tool. Corona is another good renderer, especially if you are just starting, and quality is a bit higher than Vray.
For everyday needs - Vray or Corona (I'd prefer Vray though as it's faster). For quality stills - Corona or Maxwell (easy to learn). I never tried Arnold, but I think it's an interesting renderer.

ohh, you missed some years on renderman development. since some years renderman is strong in raytracing and since renderman 19 they got a new default pathtracing engine (RIS) in it. now renderman 20 is new and got some more improvements in RIS.
you can still switch to REYES as engine but the default is now RIS. RIS shaders are now build as C++ code, which means you still need a proper TD to build new once. but they have everything you need in the package.
everything in RIS is build for PBR (lights and shaders).


mr Bob

08 August 2015, 09:09 PM

Hi, I am new to rendering software.

I'm thinking about learning either Renderman or Vray. Is there any clear differences between these software? I'm looking to do 3D animations and still images.

I guess if you could share your perspective on what your preferred rendering software is and why, that would be great :D

Really it depends on what your doing and more importantly the software framework your using to talk to the standalone render engine.


adisan

08 August 2015, 09:53 PM

Hi, I am new to rendering software. I'm thinking about learning either Renderman or Vray. Is there any clear differences between these software? I'm looking to do 3D animations and still images.
1. At your level the only difference is the amount of tutorials you can find on the internet. With vray you have tons of tutorials, for renderman not so much.
2. At higher level there is nothing that can beat renderman's flexibility, nothing!
3. Pricewise Renderman is half the vray price now.
4. It's much easier to find artists that know vray properly than artists that know prman properly so there are bigger chances that you know/meet someone which uses vray in production.
5. Stability wise, renderman rocks, everything is supported and you can go as deep as you want with it, with vray you're pretty much limited to the amount of shaders shipped with vray.
6. Since Prman is on the front end of shading development and everybody is following them you might get a better tech inside Renderman than with Vray.
7. Speedwise, if you're not raytracing it's blazing fast in REYES mode.
8. VRay is easier to use but it's not as flexible as renderman.
9. Renderman can handle a ton of geometry with no hiccups.

So I would say to give Renderman a try, especially since it's free for non commercial purposes. You might find it intimidating at the beginning but it's a very good investment.


luma91

08 August 2015, 10:55 PM

I've had a bit of a go at Renderman and Vray. I was able to try out Renderman since it's free for non-commercial use (big plus) and I use Vray at work.

Overall I'd say I prefer Vray, maybe that's because I've been using it more, so I could be pretty biased in my opinion. I wish they offered a free non-commercial version of Vray so I could use it for my own personal projects at home, though. Other than that, I think Vray ticks all the boxes I could ask for (much more than mentalray), so I don't really need to look at a different render engine anymore. That said I am curious about Maxwell, I've heard it a few times recently in CG circles. It would have to be godly to get me to move from Vray though.


tomevertonfc

08 August 2015, 11:34 PM

I've used Arnold, V-Ray and Renderman RIS in production and they are all excellent renderers.

In my opinion, Arnold is the easiest to set up and use of the three. You can basically just plug a good quality HDR into a dome light and immediately get nice results in interactive mode. Crank up the samples for the final output. Definitely very nice for a beginner to play around with.

When people say that renderman/prman is more flexible, I actually think it is more the case that renderman has been around for so long that plenty of tools and shading techniques have grown and developed out of it. People who have experience using renderman for years can do some awesome tricks and workflows that currently aren't possible with other renderers out the box.


mister3d

08 August 2015, 12:00 AM

Maybe I was a bit harsh, but I want the OP to understand, that choosing the first renderer is important. It might pull you back in studying rendering and lighting\materials setup for several years, as some renderers require much more experience and understanding than others. So yes, though as people say Vray lacks some shader flexibility, you must be a really advanced user to need anything beyond what it has. Also you can download shaders on the net if needed. Flexibility might be important if you render thousands of frames per month for a production with lots of effects. Are you in this situation? Most probably not yet. So choose a renderer, which gives you the knowledge of lighting and materials, without unnecessary diving into technical side. It might be totally redundant for you personally at all. Bear in mind. that some people here are serious professionals, working in top studios with a dedicated programming background, so what's good for them might be totally impossible for you from a technical point. Later you might come back to more advanced solutions, but by that time who knows, maybe we will use realtime game engines for rendering.


ThE_JacO

08 August 2015, 01:02 AM

I don't think Renderman is dead at all... It's just existing in a more competitive space now.
This is possibly the best and most succinct answer to both those categories of people that think it's dead, as well as those who claim it's more powerful of flexible than competitors, neither of which holds true.

Test drive them all, pick what you like best. Demo or free for individual use versions are available anyway.
Renderman and 3Delight are both completely free for individuals anyway, so I can't see why someone who has more time at hand to learn than time to spend selling the skills they haven't matured yet wouldn't give at least one of the two a shot.


tonytrout

08 August 2015, 02:24 AM

Just as a note if you watched the video above, RIS now renders volume fluids
Btw, you can use renderman straight out of the box without programming skills.


Tad

08 August 2015, 03:06 AM

i would just learn to be the best artist you can be.
Just pick whatever is most widely used in the industry that way getting a job
will be easier in the future.

any other advice about picking a renderer is just for bigwig studios.

They all spit out your pictures for you ;)


mihai112

08 August 2015, 08:32 AM

I'm probably the one with the least amount of experience working with multiple renderers here, used mental ray for years because it came with Maya and then switched to Vray and never looked back. For me rendering all sorts of objects in my free time and how easy it is there hasn't been a reason to try anything else, it got crazy fast with each release and in the last version it has a slider for setting up all the quality levels for whatever type of object u are rendering. Just to give an example it's as easy as adding a light, applying a shader turning on GI and hitting render, of course it can get very complex as u get more advanced with it.

https://www.youtube.com/watch?v=28BvbUj2dag


SebKaine

08 August 2015, 09:19 AM

For a mix of still image / 3d anim , i would tend to give the edge to v-ray.
Basically Ris and Arnold are very similar in their approach. the strength is that their way of doing thing will avoid flickering , and this is great for 3d animation. but to achieve that they choose a brute force approach and this doens't give you many flexibility. Using reyes and ris could be a potential way of gaining some flexibility but having to redo all your shading twice is not a clever approach imo and that's the main cons of Renderman they have ditch 20 years of dev from the equation by creating 2 separate world. Vray on the other hand have a lot of way to use tricks and cheats. Those technics tends to create flicking but for still image they are ace. And if you need to render 3d anim you can switch to a brute force approach. You can use the same shaders with any methods. one other argument for v-Ray is that the community is huge. And price wise a render threads will cost you 1000e for arnold 500e for renderman and 250e for vray.

I am not a fan of vray myself , i find it a little indigest and messy but i really think it's an extremely well done allroad engine for what you want to do : stills + anim

Cheers

E


sierra62

08 August 2015, 03:43 PM

I am assuming you are working from home or a student. Everyone has their favorite renderer (Maxwell and Mantra for me), but all the ones mentioned are CPU renderer's that are designed to utilize renderfarms. If you are learning lighting and rendering, look at a GPU renderer like Octane. This will let you render really fast and make changes to see what each of the variables and lighting setups do without waiting a long time to see any results.

If you aren't rendering volumes or really large geometry, use Octane, other wise check out a CPU renderer.


SebKaine

08 August 2015, 05:06 PM

(Maxwell and Mantra for me).

We should build a club of zealote together Cody ! :)


mihai112

08 August 2015, 06:42 PM

I am assuming you are working from home or a student. Everyone has their favorite renderer (Maxwell and Mantra for me), but all the ones mentioned are CPU renderer's that are designed to utilize renderfarms. If you are learning lighting and rendering, look at a GPU renderer like Octane. This will let you render really fast and make changes to see what each of the variables and lighting setups do without waiting a long time to see any results.

If you aren't rendering volumes or really large geometry, use Octane, other wise check out a CPU renderer.

Umm Vray also has GPU rendering with RT and also has a fast Progressive image sampler that gives you a fast preview except it isn't interactive as RT.


sierra62

08 August 2015, 06:45 PM

Umm Vray also has GPU rendering with RT and also has a fast Progressive image sampler that gives you a fast preview except it isn't interactive as RT.

I didn't know that Vray added a GPU renderer, when did they add that?


vlad

08 August 2015, 07:41 PM

I didn't know that Vray added a GPU renderer, when did they add that?

About 4 years ago...


SebKaine

08 August 2015, 01:34 PM

On top of the GPU render. V-ray unlike ris and arnold offer 2 shading language. You can build OSL shaders. But and this is i think ? a unique feature you can render GLSL shaders in V-ray.
Vray RT / texture baking / glsl translator are other proof of the great versatility of the tool.

http://docs.chaosgroup.com/display/VRAY3/GLSL+support+in+V-Ray


leif3d

08 August 2015, 07:32 PM

There's a lot of misinformation on this thread. Starting from that video, to many comments on RenderMan being dead, which is ironically the opposite.
As a user, let me clear things up a bit.

-RenderMan has a new pathtracer renderer now called RIS, which is taking all the Pixar development focus.
-There is a lot of Hyperion (Disney renderer) technology in RIS, including the new denoiser used in Big Hero 6 and Disney shaders. The development teams share progress.
-Yes, RenderMan REYES is very much being deprecated, so is RSL language, which means RIS is just being born and will be used in all their future films, starting with Finding Dory.
-Katana comes bundled with RenderMan RIS.
-It has many integrators, including the first commercial bidirectional pathtracer (rays from camera and lights).
-Web-based Tractor render manager included.
-Shading and Lighting is C++ and supports OSL.
-Houdini 15 native plugin.
-Blender native plugin.
-C4D support is being developed.
-Google Zync support.
-Free for non-commercial use, so you can give it a try instead of reading misinformation online...

I hope that helped clear things up, at least on the RenderMan side.

If you guys need to learn RenderMan, the new renderman community (https://community.renderman.pixar.com/index.html) is a great place to start, as well as the renderman university (http://renderman.pixar.com/view/how-to)


mr Bob

08 August 2015, 10:21 PM

There's a lot of misinformation on this thread. Starting from that video, to many comments on RenderMan being dead, which is ironically the opposite.

I disagree, when you cannot even render a decent volume using RIS after 7 hrs and there is not one shipping example nor can support provide a set up that's out of the box to be used with Katanna. As a CG sup prman is the last choice in rendering for me, its a combination of arnold and mantra on any show I set up.


tonytrout

08 August 2015, 12:36 AM

There's been quite a few improvements in volume rendering with the last two updates in the last month or so. I'm happy with the speed Pixar are developing renderman RIS I think it augers well for the future


leif3d

08 August 2015, 12:50 AM

I disagree, when you cannot even render a decent volume using RIS after 7 hrs and there is not one shipping example nor can support provide a set up that's out of the box to be used with Katanna. As a CG sup prman is the last choice in rendering for me, its a combination of arnold and mantra on any show I set up.
The Developments on volumes and fluids in version 20 are great and VDB support has been great for over a year now. It is very fast, especially with the denoiser. I'm not sure why you'd need to render a frame for 7 hours...might be user error?
Of course any renderer has its shortcomings, but development has been incredibly fast, including a restructuring of the development team and tight collaboration with Disney's r&d department, so renderman is more alive than ever.
The fact that you choose a different renderer doesn't mean the ones you haven't chosen will somehow disappear.


catche

08 August 2015, 09:56 AM

Does the denoiser interpolate? Does it produce flicker and/or remove small details? Exactly how does it work?


JWRodegher

08 August 2015, 10:12 AM

Here, I think there's some interesting info in this section:
http://renderman.pixar.com/resources/current/RenderMan/risDenoise.html
Hope that clears some doubts.
Cheers.


Samo

08 August 2015, 10:59 AM

Basic understanding of raytracing internals & color pipelines will give you enough skills to produce good results with any engine and to accomodate them to your needs. If and how raytracing engines will give you access to that kind of information is another matter.


bjoern

08 August 2015, 11:27 AM

I worked with all 3 renderers in the past years (prman only before RIS) and I'm evaluating at the moment all Arnold, Vray and Prman20 for a personal project and for teaching reasons.

All I can say right now. Renderman 20 is very strong! very strong! And I really hate to use it and thought the same as "mister3d" when they gave it away for free.
But it is pretty much now like another arnold if not better. Anyway when I make my conclusion I will post again. Maybe it is of help.


catche

08 August 2015, 02:06 PM

That denoiser seems great, but I'm skeptical to how it works when there is detailed geometry in the background, can anyone confirm how good it is exactly? Pros and cons? Also why don't other renderers implement something similar?


leif3d

08 August 2015, 04:26 PM

That denoiser seems great, but I'm skeptical to how it works when there is detailed geometry in the background, can anyone confirm how good it is exactly? Pros and cons? Also why don't other renderers implement something similar?

-It's not a simple post process. It essentially makes a bunch of AOV's as part of your EXR beauty pass, including albedo, z, vector, normal, variance, etc... and then using proprietary code (developed for Big Hero 6) it denoises your image.
-It has no flicker, because it uses cross-frame interpolation.
-iRay also has developed a denoiser and I suspect other renderers will follow suit.
I've used it successfully in production and If you're skeptical, watch big hero 6...or try it on your own, since RenderMan is now free for non-commercial use.


leif3d

08 August 2015, 04:36 PM

Basic understanding of raytracing internals & color pipelines will give you enough skills to produce good results with any engine and to accomodate them to your needs. If and how raytracing engines will give you access to that kind of information is another matter.
It's not that simple for production. These are some simple examples I can think of...
-If a renderer doesn't support VDB, it might break a Maya/Houdini pipe
-if it doesn't have a denoiser, you might be looking at very high render times.
-not having deep data output might break a stereoscopic pipe or compositing pipe.
-if a renderer is not stable, you will have issues meeting deadlines and staying within budget. This is a huge point, because you're getting the same development budget used for film production at Pixar and Disney. They can buffer massive costs, because they make it, test it and use it, then it trickles down to the user in a very polished manner. Same thing goes for the render manager Tractor, which you can get for $100 bucks...
-if a renderer is not extensible, it might not fit a big pipeline.

Not to mention every renderer has workflow differences, which in itself is a huge point to consider.


catche

08 August 2015, 05:20 PM

-It's not a simple post process. It essentially makes a bunch of AOV's as part of your EXR beauty pass, including albedo, z, vector, normal, variance, etc... and then using proprietary code (developed for Big Hero 6) it denoises your image.
-It has no flicker, because it uses cross-frame interpolation.
-iRay also has developed a denoiser and I suspect other renderers will follow suit.
I've used it successfully in production and If you're skeptical, watch big hero 6...or try it on your own, since RenderMan is now free for non-commercial use.

I see, thanks. Also found some more information on it and it seems nice. What I wonder still, is do you have to use .exr for it to work? I'm usually stuck with rendering to jpegs. Also my work uses max so I'm just curious.. But again thanks for the information nonetheless.


leif3d

08 August 2015, 05:33 PM

I see, thanks. Also found some more information on it and it seems nice. What I wonder still, is do you have to use .exr for it to work? I'm usually stuck with rendering to jpegs. Also my work uses max so I'm just curious.. But again thanks for the information nonetheless.

Yes, you need to use EXR.
As far as I know, there is no development planned for 3dsmax "yet".


catche

08 August 2015, 06:39 PM

As far as I know, there is no development planned for 3dsmax "yet".
I know, although there is a guy doing something of a conversion, but I'm not technical enough to know whether or not he'll be able to convert it all properly without re-inventing stuff.

https://vimeo.com/133612957


leif3d

08 August 2015, 07:21 PM

I know, although there is a guy doing something of a conversion, but I'm not technical enough to know whether or not he'll be able to convert it all properly without re-inventing stuff.

https://vimeo.com/133612957

Yep, saw that a couple of weeks ago, it looks like a really great job so far.

As far as the volume questions...this is an HD frame that cooked 1 hour.
-All VDB files from the openVDB site, including the dragon.
-multiscatter volume shader.
-Emissive volume.
-Bidirectional pathtracing.

This is on a very old dual xeon X5550 with 24gb ram. On a modern workstation, this should be fully converged in under 30 minutes.

FULL RES IMAGE (http://payload399.cargocollective.com/1/5/164968/10287589/VDBTest_1920.jpg)

http://payload399.cargocollective.com/1/5/164968/10287589/VDBTest_1920.jpg


leif3d

08 August 2015, 02:00 AM

A new article just went online from fxguide. It's about day 3 at siggraph and has a huge focus on rendering. Really great read.

http://www.fxguide.com/featured/siggraph-day-3-and-out/


Samo

08 August 2015, 08:46 AM

-if it doesn't have a denoiser, you might be looking at very high render


In my opinion, denoise filters kill detail in areas which get only indirect or little lighting which is very very bad, since those areas are expected to have more diffuse detail to show and richer diffuse gradients. Personally I would never use any denoise filter in post pro. A good biased interpolator algorithm is better than any denoiser. I see denoise filters used in the Blender community quite often with Cycles and the results are catastrophic in my opinion.


Not to mention every renderer has workflow differences, which in itself is a huge point to consider.

Every raytracing engine and raytracing algorithm work more or less the same under the hood.


leif3d

08 August 2015, 05:27 PM

In my opinion, denoise filters kill detail in areas which get only indirect or little lighting which is very very bad, since those areas are expected to have more diffuse detail to show and richer diffuse gradients. Personally I would never use any denoise filter in post pro. A good biased interpolator algorithm is better than any denoiser. I see denoise filters used in the Blender community quite often with Cycles and the results are catastrophic in my opinion.



Every raytracing engine and raytracing algorithm work more or less the same under the hood.
The details of the technology are there for you to read in the renderman docs and I believe Pixar released the papers.
Comparing it to the blender solution is very strange since it's not the same solution at all. It has been successfully used in production for years now at Disney and you're getting the same tech used in big hero 6. I didn't see any "catastrophic" results there...
It's also free to use with blender.

Since the main algorithms are similar in all engines, the differences are in the details, especially development focus and workflows, which makes the decision of choosing one even more important.


SebKaine

08 August 2015, 02:11 PM

thanks for all the info Leif ! Just to be picky there is one point that is not 100% true :)


-It has many integrators, including the first commercial bidirectional pathtracer (rays from camera and lights).

The first commercial BDPT is Maxwell. And it is the only way he knows. If Pixar use VCM method for their BDPT it first compute a photon map before path sampling. Maxwell in this regards is more brute force, and i guess a little more accurate. But i have no idea on what is their exact method.

My biggest complain to pixar is that they ditch RSL. Because you ended up with an new render engine that is sell as Pixar Renderman, but that has nothing to do with Renderman. ris is pixar answer to arnold, it's a beautiful answer no doubt about that. but i would have love to see a binding of their new function into Rsl in order to unify and not isolate the reyes and ris world.

cheers

E


leif3d

08 August 2015, 04:25 PM

thanks for all the info Leif ! Just to be picky there is one point that is not 100% true :)



The first commercial BDPT is Maxwell. And it is the only way he knows. If Pixar use VCM method for their BDPT it first compute a photon map before path sampling. Maxwell in this regards is more brute force, and i guess a little more accurate. But i have no idea on what is their exact method.

My biggest complain to pixar is that they ditch RSL. Because you ended up with an new render engine that is sell as Pixar Renderman, but that has nothing to do with Renderman. ris is pixar answer to arnold, it's a beautiful answer no doubt about that. but i would have love to see a binding of their new function into Rsl in order to unify and not isolate the reyes and ris world.

cheers

E

Oh, you're right, thanks for the correction. Since this was a comparison thread for the main 3 film renderers, I forgot about the rest of the market. Maxwell had been at the tech forefront of the photoreal market for a while, especially archviz stills.
I've seen some posts on the renderman forums from people that are trying to use vcm for archviz, but the lack of caching is an issue for those eternal camera animations in archviz.

The issue with rsl is speed. There is no way to compete against a precompiled c++ shader. Eventually the entire engine will run on c++. In this version they transferred the lighting to c++ and all they have left are displacements, which are still rsl and dicing.


Samo

08 August 2015, 08:13 PM

The details of the technology are there for you to read in the renderman docs and I believe Pixar released the papers.
Comparing it to the blender solution is very strange since it's not the same solution at all. It has been successfully used in production for years now at Disney and you're getting the same tech used in big hero 6. I didn't see any "catastrophic" results there...
It's also free to use with blender.


Hi, thanks for the information. Could you provide a link about the denoise thing you are talking about? I wonder whether you are talking about a denoise algorithm in render time or a denoise filter in post production. In Blender/Cycles (a GPU path tracer) denoise filters are used in post production and the result are quite bad, they usually kill lot of diffuse detail in the lower part of the histogram, where more lighting information and detail could be allocated.

Since the main algorithms are similar in all engines, the differences are in the details, especially development focus and workflows, which makes the decision of choosing one even more important.

In general, there is not an universal engine for all rendering cases, this is one of the nice things about this industry. Between faster, more accurate and easy to use, users can usually pick only two.

Raytracers will be always comparatively more efficient that renderman engines at raytracing tasks but then brute path tracers have its own share of limitations when rendering certain scenes, though they provide the best consistency.

I would never recommend a brute path tracer to a person who is rendering mostly still interiors archivis, no matter how much hardware you have got to throw at them. Also I would never render a giants robots movie with a path tracer unless you don't care about caustics effects. But then, they have made several ones that way.

On the other hand, production ready path tracers can render all kind of sophisticated Global Illumination stuff (except from caustics) with lots of consistency, which makes then good for vfx. Besides, they are easy to use since cheatings are not allowed in order to keep consistency. And they are the parallelization kings.

The answer is that there is not a simple answer to the question.


Panupat

08 August 2015, 02:58 AM

I'd love to play with Renderman at home. It's really weird how it's named None Commercial yet you need full version of Maya to install it (I'm not a student obviously). I have no time for play with it at my studio and buying 1 for home use is just absurd... reseller here charges almost 40% more than US and will only sell bundle of 5 licenses. What a bummer.

Waiting for PRMan for Blender to mature here.

@Samo I see your point. In that case Vray sounds like an excellent choice since it offers both. 3.1 is looking real sweet as well.


beaker

08 August 2015, 06:08 AM

I'm pretty sure the Denoiser was based on this Siggraph paper:

http://www.cgg.unibe.ch/publications/robust-denoising-using-feature-and-color-information
http://cgg.unibe.ch/publications/2013/robust-denoising-using-feature-and-color-information


beaker

08 August 2015, 06:09 AM

I'd love to play with Renderman at home. It's really weird how it's named None Commercial yet you need full version of Maya to install it (I'm not a student obviously).You can install just the standalone renderman studio with the NC version and just use it via commandline.


Buexe

08 August 2015, 06:27 AM

I have had some time to play around with the new renderman and I was very impressed. I tested the denoiser and from my POV it does kill detail, but nothing that you would really miss in an animated sequence and that`s what it is for. It has two modes either for single frame or cross-frame denoising. I had some animations with a furry creature ( xgen ) but whenI found out that the rendertimes were pretty low I didn`t bother using the denoiser. It is a standalone app, so can be used with any exr sequence btw from my understanding.

If I were to use it in a professional context I would wonder which online renderfarms support it and while vray, which I have very limited experience with, is certainly a dominant player in the rendering realm, I find it hard to justify it`s cost in an academic context given what the competition is offering.


Panupat

08 August 2015, 10:49 AM

And what am I going to render without DCC to generate the content for it? :/ No that's not how I want to play with it.


leif3d

08 August 2015, 03:07 PM

I'd love to play with Renderman at home. It's really weird how it's named None Commercial yet you need full version of Maya to install it (I'm not a student obviously). I have no time for play with it at my studio and buying 1 for home use is just absurd... reseller here charges almost 40% more than US and will only sell bundle of 5 licenses. What a bummer.

Waiting for PRMan for Blender to mature here.

@Samo I see your point. In that case Vray sounds like an excellent choice since it offers both. 3.1 is looking real sweet as well.
That's a Maya limitation, not a RenderMan one. Maya LT doesn't support plugins.


JayBBK

08 August 2015, 01:16 PM

I am using arnold for sometimes now, as a brute force engine the price to get a noise free picture is very high.

People use to say that arnold is fast but in the fact it seem to be a once upon a time story...
Does we use the same arnold version than ILM? i think no. Arnold still don't offer a good denoiser solution, so how you guys who use it in production do to get a clean render ?



0 Thoughts to “Vray Vs Maxwell Render Comparison Essay

Leave a comment

L'indirizzo email non verrà pubblicato. I campi obbligatori sono contrassegnati *