Fstorm render

[PIXELBOX_SRO]

i have to say (someone woudl sooner or later anyway most likely) that i am surprised that Vlado and the crew arent already tired explaining stuff over and over once another post like this appears....sure it usefull to a certain extent but i mean come on guys....this is VRAY forum.

Just wanted to say this...now you can start throwing stones at me

[vlado]

Well, it is actually quite useful for me, even if it is unpleasant. Discussions like these give me directions on what we need to improve in V-Ray.

Best regards,
Vlado

[PIXELBOX_SRO]

cool i understand the need for feedback and i am sure everyone has their list of wishes ready to throw at you
How about a poll though?
Lets say TOP ten most wanted/needed?

[Ivan1982]

You shouldn't feel discomfort, nothing is perfect in life. It is not BMW vs Mercedes talk as they are not even near giving automotive industry what chaos group gave to CGI industry.
People out there are counting on you guys to lead this transition, if it wasn't for you we would have completely different situation today. It is quite normal that anyone in that position have different priorities.

You have implemented lens effects some times ago and that was great then, but they are not top notch as they used to be, people mostly rely on ArionFX while you are providing same stuff(that is not satisfactory). No one is saying you are not able to do better job, they just want to see it working better.
People try Octane or Fstorm and say "Glare/bloom is working out of the box" maybe it isn't working precisely as Vray unclamped EXR + ArionFx + camera RAW processing workflow, but in 99% of cases it gets job done just as good, of course they will say something about that. Same goes for other examples.

You've already have this conversation countless times, there is difference between someone saying "that render engine is better as i like it more" and giving you realistic unbiased examples why it isn't better.
If you were in position of user of VRAY it would be different, you would like to continue using something you used for 10 years, as you are feeling comfortable about it, you have everything setup for that engine, you have knowledge and in worst case scenario you will need to adjust to few new options every couple of months.
As i said Vray became "standard" in this industry, even if licences are free, if you change engine there is period of adjustment you need to rely on converters, new workflow... People can't afford such changes every year.

[joconnell]

Well, it is actually quite useful for me, even if it is unpleasant. Discussions like these give me directions on what we need to improve in V-Ray.

Best regards,
Vlado

Exactly. There's a really good book called "The obstacle is the way" by Ryan holiday which uses a lot of stories from the the old stoic thinkers. The title comes from the idea that if you come up against a problem, you'll be shown what you're weak at currently and where you need to improve to be able to solve the problem. If you treat each problem as an opportunity to become better then hard situations becme a huge advantage!

[joconnell]

You have implemented lens effects some times ago and that was great then, but they are not top notch as they used to be, people mostly rely on ArionFX while you are providing same stuff(that is not satisfactory). No one is saying you are not able to do better job, they just want to see it working better.

If we take this as a small example, the issue is that lots of different people working in different areas want different features in vray and they'd like the built in stuff to be better than a separate program who's only designed to perform that one feature. The film guys really don't care about lens effects or arion because it doesn't suit their purposes and they'll do all the stuff in nuke, likewise they don't care about gpu yet because the cards don't have enough memory to fit one of their scenes. They do care about things like light being technically accurate because they have to do 3d renders that match the way light worked in the footage that they're comping into. Arch vis or fstorm users on the other hand don't care about the technicality of the light, they just want to get a nice looking image out in the quickest time possible that looks about right.

Chaos isn't that big a company in terms of staff numbers and while I'd love it to have the best raytracing speed (which it might already), better shaders than whatever is out there, better volumetrics than whatever, a better atmosphere / sky system than terragen or vue and be a faster gpu renderer than whatever the fastest thing is. They still have a finite number of staff to make all this stuff with so can't concentrate on everything to the same level at once.

As vlado says "if you find a renderer that fits what you want to do better, then great!" makes sense too. If you were a carpenter would you try to find a single multi tool that had loads of options for plumbing and electrical stuff too or would you get a few saws and screwdrivers that do that one single job really well? A lot of the big animation companies have replaced renderman with their own in house path tracers (let's say animal logic on the lego film). They know exactly the features that they need to make their current project so can put all their resources into those specific areas only. If they keep doing exactly the same type of film, this'll keep working for them (They are with the lego batman film). If in the next film they need totally different features like volumetrics and fur, maybe their next film is the Jungle book 2 so rather than only doing solid, shiny plastic which their fast raytracer is great for, they now need a renderer than can handle a tonne of hair and texture maps and displacement with motion blur efficiently. This is a totally different direction from where they started so the renderer will change heavily. The ui will get loads of new controls and get more complex looking but they need that to support all the next things that they're doing. It's the same thing with vray and the general versus specific tool option. If you only ever do car renders and nothing else, it'd probably make sense to pick a renderer that does really good solid object rendering and maybe has good passes and matte options to integrate into live action stills. If you only do arch vis stuff, you probably want something that does really good GI and has lots of memory handling - if you don't care about post processing then having some glow / glare stuff to add some realism would be nice too. If you do cheap kids tv show stuff where quality isn't a concern then you probably want a fast gpu renderer so you don't need to invest in a render farm and you don't want too many options or controls so that any user can pick it up easily.

If you want to be able to do nearly everything well, you'll have to get a good general purpose renderer. It won't be as good as the specific render that's purpose built for one single task but then that's kind of the point - if you know exactly what you want and there's a thing that does it far better than anything else it makes total sense to use that thing. If you don't know exactly what you want or if what you need to do changes every time then you can either go with a general purpose tool that'll do a good job of each of them, or for each individual job keep buying the one thing that's perfect for that individual job.

Totally up to you as an artist and a business person to weigh up the good points and bad points of what's right for your needs and make a decision.

[Ivan1982]

If we take this as a small example, the issue is that lots of different people working in different areas want different features in vray and they'd like the built in stuff to be better than a separate program who's only designed to perform that one feature. The film guys really don't care about lens effects or arion because it doesn't suit their purposes and they'll do all the stuff in nuke, likewise they don't care about gpu yet because the cards don't have enough memory to fit one of their scenes. They do care about things like light being technically accurate because they have to do 3d renders that match the way light worked in the footage that they're comping into. Arch vis or fstorm users on the other hand don't care about the technicality of the light, they just want to get a nice looking image out in the quickest time possible that looks about right.

I agree, but that doesn't mean i dismiss there isn't need for some changes right now. There is no need to compare big studio vs other usage here, users are in the same boat.
Hypothetically speaking, let's imagine there is one click universal solution available right now, it would be same for big studio and "little" guy it will only matter in terms of resources available.
Every car have it's own wheel, it differ in performance greatly between high performance one and middle tear version, but again it serve same purpose, it is standardized and fits purpose until something better pops out.
I am taking this as example as it is now more than ever obvious that GPU will take over lead in upcoming years from CPU, why not being ready for it?

Chaos isn't that big a company in terms of staff numbers and while I'd love it to have the best raytracing speed (which it might already), better shaders than whatever is out there, better volumetrics than whatever, a better atmosphere / sky system than terragen or vue and be a faster gpu renderer than whatever the fastest thing is. They still have a finite number of staff to make all this stuff with so can't concentrate on everything to the same level at once.

I understand, i know their resources is not infinite, it can't be all done in the same time, this is not about pushing them and request some crazy things, at least i didn't got that feeling from this thread, we are just talking about new possibilities, if they find it useful that is great, if not they do not need to feel discomfort.
It is opinion from "light" users, i am sure studios have more voting power but that doesn't mean we can't express our perspective.

As vlado says "if you find a renderer that fits what you want to do better, then great!" makes sense too. If you were a carpenter would you try to find a single multi tool that had loads of options for plumbing and electrical stuff too or would you get a few saws and screwdrivers that do that one single job really well? A lot of the big animation companies have replaced renderman with their own in house path tracers (let's say animal logic on the lego film). They know exactly the features that they need to make their current project so can put all their resources into those specific areas only. If they keep doing exactly the same type of film, this'll keep working for them (They are with the lego batman film). If in the next film they need totally different features like volumetrics and fur, maybe their next film is the Jungle book 2 so rather than only doing solid, shiny plastic which their fast raytracer is great for, they now need a renderer than can handle a tonne of hair and texture maps and displacement with motion blur efficiently. This is a totally different direction from where they started so the renderer will change heavily. The ui will get loads of new controls and get more complex looking but they need that to support all the next things that they're doing. It's the same thing with vray and the general versus specific tool option. If you only ever do car renders and nothing else, it'd probably make sense to pick a renderer that does really good solid object rendering and maybe has good passes and matte options to integrate into live action stills. If you only do arch vis stuff, you probably want something that does really good GI and has lots of memory handling - if you don't care about post processing then having some glow / glare stuff to add some realism would be nice too. If you do cheap kids tv show stuff where quality isn't a concern then you probably want a fast gpu renderer so you don't need to invest in a render farm and you don't want too many options or controls so that any user can pick it up easily.

Yes, every tool have good/bad side of dealing with different scenarios, but how realistic is to expect single person to buy 10 different licences for 10 different things? If things are working like that we would still have Amiga type computers, everything tends to evolve into one system that will overlap those 10 different things into lets say 3. No need for coprocessor type stuff...
I Agree that you need to pick your poison wisely, and that is why most of people are on vray(it simply offers most of those things in one package). But I do not agree archviz and tvkid shows are destined to look bad indefinitely.
Big studios can afford to develop their own render engine but that can't be expected from low level user, i do agree with most of people when they are expecting some new things to be changed or implemented, it will not mess up VFX, Archi, kids show industry it can only make better product at the end.

If you want to be able to do nearly everything well, you'll have to get a good general purpose renderer. It won't be as good as the specific render that's purpose built for one single task but then that's kind of the point - if you know exactly what you want and there's a thing that does it far better than anything else it makes total sense to use that thing. If you don't know exactly what you want or if what you need to do changes every time then you can either go with a general purpose tool that'll do a good job of each of them, or for each individual job keep buying the one thing that's perfect for that individual job.

Totally up to you as an artist and a business person to weigh up the good points and bad points of what's right for your needs and make a decision.

Again, everyone here realise that this is how things are, customers first talk with developer before they decide what to do next, if someone wants to change mobile/internet provider i am sure in majority of cases they speak with their current provider of services to see how thing are and how they will be before they make final decision. That is what happened here, people are interested about their future, and that is most important thing to them no matter how trivial it seems to someone else.
Big studios, can afford one machine rendering 12+hours one frame, develop render engine, buy nuke licences, hire 10 guys to do post processing... How can you translate that principles to smaller companies or individuals?

No matter how big something becomes it can crumble eventually, just remember what happened with all those projects from Autodesk(one example gMax), where is industry leader LightWave now, Softimage, Brazil?(actually Vray put a final nail in their coffin), there is so many examples.
On other hands we have something like zbrush, marvelous, even Vray, big companies did nothing about them. Usually all start as "little" guy fun, and before everyone blatantly disregard Fstorm or redshift as "kids tv show renderers" bare in mind they can end up being game changers eventually.

[joconnell]

I agree, but that doesn't mean i dismiss there isn't need for some changes right now. There is no need to compare big studio vs other usage here, users are in the same boat.

Again the tough bit is there's different users that want different outcomes - some users want a final render out of the box, other users want better data as most of their time is spent tweaking render elements afterwards so there's big and small studios, each big studio works differently from another big studio and each small studio works differently from each other small studio, the users in each of those places all want totally different things to achieve their tasks

Hypothetically speaking, let's imagine there is one click universal solution available right now, it would be same for big studio and "little" guy it will only matter in terms of resources available.

Yep indeed - if there was one "best" renderer that did everything then most people would just hop on that. When autodesk started buying all of the 3d apps you'd have hoped that they'd take the best modelling, animation, layers, function curves tjhat they had out of the three programs and just combined those into a single "best" application but unfortunately it didn't happen. It'd be great if it did though, it'd make it way easier to find freelancers

Big studios, can afford one machine rendering 12+hours one frame, develop render engine, buy nuke licences, hire 10 guys to do post processing... How can you translate that principles to smaller companies or individuals?

You kind of can't really. I normally work in a fairly small place doing vfx work on some decent tv shows but with quite small teams. You'd have maybe 4 or 5 3d guys and about 15 compers - we mainly do comp work so maybe only 2 or 3 of the compers are working on our shots. I need to turn around a lot of shots very quickly and be able to put out the passes and data that they need to do their work with short frame times. Last month I was working in a studio that only does high end vfx, there was 430 artists working on the show that I was on and their render times were sometimes 20 hours or more on houdini renders. The difference is my normal company get a budgets maybe in the tens or low hundreds of thousands for their tv work, the big studio got budgets of tens of millions, they're playing by totally different rules so the rendering solution for one place might not fit the other

When I mentioned gpu renderers with kids tv shows earlier it wasn't any kind of disregarding of them, it's just that one company I know who does a lot of kids tv shows has to get out a very large amount of renders very quickly and at the cheapest cost possible. Since the show is for 5 year olds they don't care about quality of lighting and materials at all, the animation is also quite rough. The budget for this type of show is low and it's all about volume. For the company to be able to make the show and not lose money, they have to work with cheaper animators and less rendering power so if a gpu renderer is able to get out shots quicker at the quality that is acceptible then it makes financial sense for that company - they won't need to invest in a render farm, I.T. people to manage the render farm or the building space and air conditioning to sit the render farm in

All jobs are totally different with different needs so it just depends on what's the appropriate tool for you, there's no perfect ones for everything at the minute that does everything so you just have to make a choice about which one is the right one for the project you're working on!

[Ivan1982]

We are back at the start

I agree some big studios would laugh their ass off if you mention them GPU in production. They do not care, they probably have much important stuff on their mind.
On other hand you have individuals and small studios that simply can't afford all those options big studios can and they would be screwed if there isn't alternatives.

You have Nvidia on one hand battling to stay alive, they don't have x86 licence like AMD(ATI) and it is much harder for them in mobile market as they are under pressure from giants(Intel, Samsung, Qualcomm...) it is just fraction of a fraction of resources for those giants to overwhelm Nvidia in that particular market.
That is why they can't afford 5-10% increments every 4 years like Intel. Only market where they are financially viable to survive in the long run is this "processing power for masses" promoting VR, utilization of all sorts of different game engines, small rendering engines that rely on cuda... all kind of gimmickry on their side to stay afloat. They are not some fairy tail ideal company but they deliver hope that we do not have to have 10 computers to render animation.

Maybe Intel have some mega cool processors they are not talking about, waiting to bury competition... but if i am to learn from history, big systems are not some mystery machines that know every aspect of their business, they are often hollow from the inside and usually head doesn't know what hand is doing, where petty jealousy and "experts" rule, billion dollar tax money deals, they are making so many products that they even do not need to sell desktop CPU's, how to expect they will care about anything else except cash, this situation is favoring that model.
I am not the only one feeling like this about current state of CPU market, top of the line CPU(4k euro xeon) can't even come close to compare to one middle tier GPU(400euro), if we disregard current memory limits(those low end GPU have) and undeveloped render engines it is clear where perf/$ value is in upcoming years.
It is much cheaper to build GPU node, much easier to upgrade, maintenance comes with less hustle. There is already new developments on the way to deal with memory issues like NvLink.
I really can't say i would bet my money on CPU in a long run. I know big studios spent loads of cash to upgrade their farms but i am looking at that like passive obstruction in development and transition to GPU.
It is same stuff like in automotive industry now, technology is already developed enough so we could start switching from internal combustion to electric but those big companies willingly or unwillingly always postpone that transition, what would be oncore to their current development if they do not develop electric? They probably act same as anyone else in that position would, making some ridiculous small baby steps, instead giant leaps like someone who is starting from scratch in that business, maybe newcomer doesn't have that extra bagade on his back and it is more free to do what it want's, to experiment and genuinely be more open and embrace new ideas. Comparing self driving car, that is cheap to run, doesn't require that much maintenance and it's faster to similar priced internal combustion car is not even fair... They are not perfect but sure there is enough room for development, same as GPU except big difference is that GPU technology is much more developed than current electric car technology.

Tapping GPU market is maybe number one priority to Chaos but they are keeping it to themselves, who knows. I can conclude that making Blagovest Taskov as lead developer(jan 2016) already showing results, there is more changes in past 6-7 months regarding GPU(he mentioned supported features few posts ago) than in previous 3-4 years, just don't want to see them stop and start calculating risks too much , this is probably most awaited service pack regarding GPU development ever.

[glorybound]

The Edsel, an American car in the 50's, was built for the people and by the people. The US government asked the people what they wanted in a car, and they built it. Trying to make everyone happy, the car was a monstrosity and failed miserably.

http://classiccars.about.com/od/clas...az/a/Edsel.htm

Moral of the story is, you can't make everyone happy.

[Ivan1982]

Really bad example, we are not talking about creating monstrosity, no one is here asking to implement upright radiator grill on Vray we are talking about bringing gap closer between CPU and GPU versions of Vray itself, well at least i do.
That is why i mentioned some features that need to be modified or supported, they are coincidentally already supported in some other rendering engines that are GPU based, CPU version of vray supports that already and they are working fine.
Bringing GPU power to Vray, it is far from edsel approach to things.

[glorybound]

Unless you think that you are the only one requesting things get implemented. The Choas team would run ragged if they did everything everyone asked. I think the Choas team listens and they try to implement things that make sense and personally, I trust that they'll make good decisions to stay on top of the render engine heap. Over the years there has always been a flavor of the month and that's probably healthy, but you need to take it with a grain of salt.

[joconnell]

We are back at the start

I agree some big studios would laugh their ass off if you mention them GPU in production. They do not care, they probably have much important stuff on their mind.
On other hand you have individuals and small studios that simply can't afford all those options big studios can and they would be screwed if there isn't alternatives.

You have Nvidia on one hand battling to stay alive, they don't have x86 licence like AMD(ATI) and it is much harder for them in mobile market as they are under pressure from giants(Intel, Samsung, Qualcomm...) it is just fraction of a fraction of resources for those giants to overwhelm Nvidia in that particular market.
That is why they can't afford 5-10% increments every 4 years like Intel. Only market where they are financially viable to survive in the long run is this "processing power for masses" promoting VR, utilization of all sorts of different game engines, small rendering engines that rely on cuda... all kind of gimmickry on their side to stay afloat. They are not some fairy tail ideal company but they deliver hope that we do not have to have 10 computers to render animation.

Maybe Intel have some mega cool processors they are not talking about, waiting to bury competition... but if i am to learn from history, big systems are not some mystery machines that know every aspect of their business, they are often hollow from the inside and usually head doesn't know what hand is doing, where petty jealousy and "experts" rule, billion dollar tax money deals, they are making so many products that they even do not need to sell desktop CPU's, how to expect they will care about anything else except cash, this situation is favoring that model.
I am not the only one feeling like this about current state of CPU market, top of the line CPU(4k euro xeon) can't even come close to compare to one middle tier GPU(400euro), if we disregard current memory limits(those low end GPU have) and undeveloped render engines it is clear where perf/$ value is in upcoming years.
It is much cheaper to build GPU node, much easier to upgrade, maintenance comes with less hustle. There is already new developments on the way to deal with memory issues like NvLink.
I really can't say i would bet my money on CPU in a long run. I know big studios spent loads of cash to upgrade their farms but i am looking at that like passive obstruction in development and transition to GPU.
It is same stuff like in automotive industry now, technology is already developed enough so we could start switching from internal combustion to electric but those big companies willingly or unwillingly always postpone that transition, what would be oncore to their current development if they do not develop electric? They probably act same as anyone else in that position would, making some ridiculous small baby steps, instead giant leaps like someone who is starting from scratch in that business, maybe newcomer doesn't have that extra bagade on his back and it is more free to do what it want's, to experiment and genuinely be more open and embrace new ideas. Comparing self driving car, that is cheap to run, doesn't require that much maintenance and it's faster to similar priced internal combustion car is not even fair... They are not perfect but sure there is enough room for development, same as GPU except big difference is that GPU technology is much more developed than current electric car technology.

Tapping GPU market is maybe number one priority to Chaos but they are keeping it to themselves, who knows. I can conclude that making Blagovest Taskov as lead developer(jan 2016) already showing results, there is more changes in past 6-7 months regarding GPU(he mentioned supported features few posts ago) than in previous 3-4 years, just don't want to see them stop and start calculating risks too much , this is probably most awaited service pack regarding GPU development ever.

Yep I get you. At the minute I've got less interest in GPU as it won't do all of my final renders but again I work in a different area to you. As you say though gpu's are pushing ahead far faster than cpus can so all of the things you're pushing for with your current work will be the things I'm asking for when GPU memory doubles in a few years time

Great to have the discussion on the forum and your work is gorgeous!

[Ivan1982]

Ah ok, i didn't get that impression, that others are requesting odd things.
Maybe you are thinking requests in general on the forums, i must agree to some extent, but people have right to do so.
You said "I think the Choas team listens and they try to implement things that make sense and personally, I trust that they'll make good decisions to stay on top of the render engine heap" probably that is why people are asking/requesting.
Anyhow i don't think Edsel was good comparison either way.
Of course, unless you are like Edsel that much and you just wanted to express that by posting that link

[Ivan1982]

Well i got a dream few years ago about gpu ready environment, unfortunately that come to abrupt stop then, it's different now. Actually some of the scenes i did couldn't even fit in gpu. There is failback situation as cpu's are always there but they are so expensive, to maintain render farm able to perform what few 1080 can do is crazy(talking exclusively about vray performance). Just waiting those few crucial changes.
You are welcome, likewise!