This is one of those "Don't work harder, work smarter" things. I'm still learning about render management in college, but if things are done in passes (ambient occlusion in one pass, reflections in another, basic shading in one) then put everything together in compositing / post, it could essentially make the workload a bit less. Another thing people can do is sacrifice some rendering effects or effect qualities like Sub-surface scattering or even find ways to keep polygon count as low as possible. I can't speak for any of the companies out there, but it's completely possible that the renders aren't as optimized as they could be.
[QUOTE=Daemon White;46522334]This is one of those "Don't work harder, work smarter" things. I'm still learning about render management in college, but if things are done in passes (ambient occlusion in one pass, reflections in another, basic shading in one) then put everything together in compositing / post, it could essentially make the workload a bit less. Another thing people can do is sacrifice some rendering effects or effect qualities like Sub-surface scattering or even find ways to keep polygon count as low as possible. I can't speak for any of the companies out there, but it's completely possible that the renders aren't as optimized as they could be.[/QUOTE]
Yeah I was taught the same. Rendering out one beautiful image or movie takes much longer and if there's something wrong with a particular aspect, you have to re-render the whole thing all over again.
With passes, you can throw all your files into Photoshop or After Effects or whatever and edit shadows, colors, highlights, AO on the fly.
[quote]This grew substantially by the time Illumination made Despicable Me 2 when peak memory use hit [B]680 terabytes[/B][/quote]
Jesus fucking christ
I'm assuming that's all RAM type memory?
No, they're talking about storage space.
I'd be interested to see a CGI movie in HFR, I imagine it wouldn't suffer the 'soap opera effect' and could look pretty good
[QUOTE=ijyt;46522687]No, they're talking about storage space.[/QUOTE]
Damn, that would take 114 6 TB hard drives. Double or triple or quadruple that if you want some RAID mode for redundancy.
But they're talking about flash memory? Christ do they want to put 680 TB on all SSDs? That's nuts.
Well I'm just glad that that they are taking up the challenge, should end up making animated films look a lot better than they currently do. Its going to be interesting to see that if it catches on whether older animated CGI films will be rereleased not just in 4K but in a higher framerate too. That's something I'd really love to see.
Maybe they'll go for more stylized stuff so they can get away with cutting certain corners.
[QUOTE=Daemon White;46522334]This is one of those "Don't work harder, work smarter" things. I'm still learning about render management in college, but if things are done in passes (ambient occlusion in one pass, reflections in another, basic shading in one) then put everything together in compositing / post, it could essentially make the workload a bit less. Another thing people can do is sacrifice some rendering effects or effect qualities like Sub-surface scattering or even find ways to keep polygon count as low as possible. I can't speak for any of the companies out there, but it's completely possible that the renders aren't as optimized as they could be.[/QUOTE]
It really depends on how the renderer is structured, for something like path tracing AO/reflection/shading are literally the same thing, to work out one of them is to calculate the others. Ray tracing on the other can have them all as separate passes (And then you can do interesting things like calculating AO at a lower resolution)
Most renderers are fairly optimised, it's just that they need to do a huge amount of work for a single pixel. Even for a simple ray tracer that only does simple lighting, it's multiple ray intersections per sample (A ray per light, and each ray needs to be tested for intersections), and you might have 4 samples per pixel for simple AA, etc. Add reflections to the mix and then for each ray fired into the scene, you might end up with another 5 bounces (Each one performing their own lighting tests, etc.), and that's not counting translucent objects (Which have scattering, absorption, etc.)
Which is where GPUs come into play, 3D rendering loves parallelism, and GPUs provide that in spades.
Another thing that can help is DX12 (Or a new iteration of OpenGL that will compete with DX12). If you can put off certain things like particles or ambient occlusion and your main shading into a game engine, then in theory you can save more time.
[QUOTE=Daemon White;46522334]This is one of those "Don't work harder, work smarter" things. I'm still learning about render management in college, but if things are done in passes (ambient occlusion in one pass, reflections in another, basic shading in one) then put everything together in compositing / post, it could essentially make the workload a bit less. Another thing people can do is sacrifice some rendering effects or effect qualities like Sub-surface scattering or even find ways to keep polygon count as low as possible. I can't speak for any of the companies out there, but it's completely possible that the renders aren't as optimized as they could be.[/QUOTE]
You should know that all major studios already make use of render passes for compositing and maximizing efficiency (in case of a pass needing to be re-rendered) this is not a new technique and it is already in use even at the student level. In fact, render passes are part of the reason memory issues are as big a problem as they are. Instead of passing around 1 images, you're passing around 5 - 10.
You can be damn sure that the people who work at these companies (pixar, dreamworks, blue sky, real fx, illumination, etc), already do everything the can to optimize render times, both on an artistic and technical front. A lot of the problem IS in fact a "work harder" mentality because models need to have higher polycounts, higher texture resolutions, higher EVERYTHING in order to still look good at these new massive resolutions.
However I absolutely agree that with the advent of high fidelity real time graphics, a lot of this stuff is going to start being worked on from a totally different angle, you hit the nail on the head there.
I wonder if the affects live action films too. Like consider Transformers, Edge of Tomorrow, and Avengers. All films with huge digital effects departments. So you huge, when you see the credits, all the names of the digital effects artists makes a paragraph wall that has to be scrolled across the screen.
Can you imagine how much technology and manpower you would need to make just one scene of Optimus Prime? I mean, he has to have a higher texture resolution, higher polycount, better lighting, and better everything else because he's portrayed in a real world setting. Whereas in a film like Despicable Me, the textures and polycount are great, but they don't have to be that great cause it's cartoony style, a cartoon for kids to young too care.
[QUOTE=cqbcat;46524507]Whereas in a film like Despicable Me, the textures and polycount are great, but they don't have to be that great cause it's cartoony style, a cartoon for kids to young too care.[/QUOTE]
That's a really naive assumption, and downright untrue in many many cases. Cartoon Stylizing != simple or easier. Different amounts of work have to go into different areas.
Cinematic renders have always been massive fucking hogs, regardless of the technology.
[QUOTE=pentium;46524552]Cinematic renders have always been massive fucking hogs, regardless of the technology.[/QUOTE]
I think that goes without saying. The point is the leap from 24 fps to 48 fps, combined with the going from 2k to 4k is going to cause a massively exponential increase in render times and required storage space for these films, and it is all happening so abruptly that the studios are unprepared.
[QUOTE=Socram;46524582]I think that goes without saying. The point is the leap from 24 fps to 48 fps, combined with the going from 2k to 4k is going to cause a massively exponential increase in render times and required storage space for these films, and it is all happening so abruptly that the studios are unprepared.[/QUOTE]
That and a lot of technologies recently have been hitting R&D walls in terms of thermal limits, transistor density and bandwidth but what you said doesn't stop them from hitting the ceiling any slower. It's the 90's again in terms of cinema graphics.
[QUOTE=lonefirewarrior;46522488]Yeah I was taught the same. Rendering out one beautiful image or movie takes much longer and if there's something wrong with a particular aspect, you have to re-render the whole thing all over again.
With passes, you can throw all your files into Photoshop or After Effects or whatever and edit shadows, colors, highlights, AO on the fly.[/QUOTE]
Anyone with a half brain won't render just a single image per frame. Rendering in multiple passes has been pretty much the standard for a while now.
Programs like nuke can do crazy shit in post such as relighting and volume fog, if you have the required passes for it.
[quote]Rendering animation has [B]become [/B]a very computer-intensive task[/quote]
Are they trying to imply there was a time when it [I]wasn't[/I]?
[QUOTE=Daemon White;46522334]This is one of those "Don't work harder, work smarter" things. I'm still learning about render management in college, but if things are done in passes (ambient occlusion in one pass, reflections in another, basic shading in one) then put everything together in compositing / post, it could essentially make the workload a bit less. [/QUOTE]
A: Except film-quality 3D art is leading away from that direction with the advent of (or more over, re-popularization of) [URL="http://madebyevan.com/webgl-path-tracing/"]Monte Carlo path tracers[/URL] which have far less intrinsic need for render pass compositing.
There's no intelligent way around this, this purely comes down to brute force hardware strength. In the shortrun using GPUs instead of CPUs would allivate the issues as GPUs are much better at doing parallel processing, however GPUs have stability and memory issues that inhibit them from being completely viable for large productions.
What's liable to be the big fix in the end will be the advent of a processor that has both CPU and GPU cores, and rendering engines that can use out-of-core memory to reduce memory overhead.
B: Management of Render passes does not necessarily improve render times, it simply allows you to go back and make modification to elements without needing to re-render everything. Movie studios already do this and have been for decades so it's not going to help the 4K gap.
[editline]19th November 2014[/editline]
[QUOTE=Coyoteze;46524726]Are they trying to imply there was a time when it [I]wasn't[/I]?[/QUOTE]
It's a lot worse. Render times increase parallel with render size, if pixel count increases by 4x, render times increase by 4x (generally speaking)
Going from 1080p to 2160p is essentially a 4x increase in render times. Think about the problem that presents if you're a giant studio with time constraints or a film producer with a budget
At our studio, we've just started producing all our content at 4096x2160 @ 60fps, up from 1080@60. Our poor little 16TB NAS is now 80% full, when before we made the switch it was 20%. I can confirm that storage space is getting out of hand with the new resolution.
If the video game industry faces the same challenge. the size issue will be a concern, but not big as the concern about rendering all the graphics on-screen a 4K. It will impact GPU and RAM depending on driver.
Who's gonna be the idiot who suggests they just tween everything to increase framerate?
I know you're out there.
[QUOTE=ScottyWired;46524888]Who's gonna be the idiot who suggests they just tween everything to increase framerate?
I know you're out there.[/QUOTE]
Anything with fast motion will ghost and skip. Eeewww.
[QUOTE=Socram;46524550]That's a really naive assumption, and downright untrue in many many cases. Cartoon Stylizing != simple or easier. Different amounts of work have to go into different areas.[/QUOTE]
When you introduce a style, you have to remain in that style when creating new props or characters. Proportions and such along with colour scheme need to be constant.
The age old example of TF2, people always whine how the style was broken, blah blah blah. The style remained the same, and was never broken. The theme of certain items and such got much more flexible, but the style being broken wasn't a matter of, say, "why is a unicorn head in TF2," but rather "what would a unicorn head look like in this world?"
[QUOTE=Socram;46524550]That's a really naive assumption, and downright untrue in many many cases. Cartoon Stylizing != simple or easier. Different amounts of work have to go into different areas.[/QUOTE]
adding to this, you get those goofy requests from clients in every creative field. Making it 'simple' is often exponentially more difficult to make it [i]good[/i]. Clients asking for things to be simple are expecting it to be cheap and fast. They don't understand that you're not just whipping up a final draft and asking for payment for the two minutes it takes to re-draw it.
You spend a lot of time developing something that is functional, readable, or convincing in its simplified state.
and with rendering, it can actually increase your render time far more than just the 4x resolution increase, because of the overall tax on the rendering hardware/software. Firms that can afford the ridiculous render farms won't feel the hurt so much in process time to that effect, but smaller companies may get burned hard.
I bet they're hoping 48 fps doesn't catch on too. It would be doubling the computing power needed again!
[QUOTE=ScottyWired;46524888]Who's gonna be the idiot who suggests they just tween everything to increase framerate?
I know you're out there.[/QUOTE]
Do you mean blending the frames together, or interpolating between the keyframes? Because "tween" seems to refer to both according to Wikipedia.
[QUOTE=LegndNikko;46525638]When you introduce a style, you have to remain in that style when creating new props or characters. Proportions and such along with colour scheme need to be constant.
The age old example of TF2, people always whine how the style was broken, blah blah blah. The style remained the same, and was never broken. The theme of certain items and such got much more flexible, but the style being broken wasn't a matter of, say, "why is a unicorn head in TF2," but rather "what would a unicorn head look like in this world?"[/QUOTE]
there's a difference between artistic direction and thematic direction
[QUOTE=Daemon White;46523978]Another thing that can help is DX12 (Or a new iteration of OpenGL that will compete with DX12). If you can put off certain things like particles or ambient occlusion and your main shading into a game engine, then in theory you can save more time.[/QUOTE]
You could theoretically dump movie stuff into DX9/10/11 or OpenGL already, no?
Sorry, you need to Log In to post a reply to this thread.