View unanswered posts
View active topics
It is currently Thu Nov 23, 2017 10:07 pm


Author Message
cgpower2008
Post  Post subject: about features and Game engine!  |  Posted: Tue May 10, 2016 8:19 am

Joined: Mon Nov 30, 2015 11:32 am
Posts: 67

Offline
Dear Sir/Madam,

I am a TV producer using your product, Furryabll.

I was wandering if the Furryball can increase the rendering speed as fast as the game engine. Like the Unreal 4, Cinebox,film engine, they have a faster real-time preview speed, and they can be used in the 3D animation production as well.

I am using the 980Ti graphics card, however the Furryball was still very slow. Comparing with the RedShift, the Furryball appears weakness in rendering speed, picture quality, efficiency and price.

Friendly recommending upgrade the rendering speed.

Recommendations:

1. Seperating the RT version into 2 versions, TV version and film version. In TV version, the speed is the priority.
Reducing a little bit the quality, and makeing sure no clear noises appears. That will be enoungh for TV production.

2. Developing the system that supporting dual graphics cards, in order to boost the rendering efficiency

3. In the RT version, optimising the texture mapping system, in order to reduce the GPU usage, the requirment of the GPU, and the costs.

4. Reducing the price of the Furryball, in order to gain customers.


Top
Jan_Tomanek
Post  Post subject: Re: about features and Game engine!  |  Posted: Tue May 10, 2016 10:13 am
User avatar

Joined: Tue Oct 27, 2009 11:03 pm
Posts: 1567
Location: Prague - Czech republic

Offline
Response of FurryBall is the fastest possible and RENDER can be NEVER same speed like game engine.

Render is versatile agains game engine who fake effect etc. Also effect in game engines are wrriten for concrete engine and not like pluggin for Maya.
FurryBall is speed enough, but it has to wait for Maya responses (same as all other renderers)

1) above
2) ? FurryBall supports unlimited number of GPU.
3) Sorry we don't understand. Do you mean compress textures?
4) FurryBall is the cheapest from ANY GPU renderers and also offers very fair rental and credit versions.

_________________
Thank you for contacting us.
If you have any further questions, please do not hesitate to contact us.

All the Best
Jan


Top
cgpower2008
Post  Post subject: Re: about features and Game engine!  |  Posted: Wed May 11, 2016 7:54 am

Joined: Mon Nov 30, 2015 11:32 am
Posts: 67

Offline
thank you for response

the 3rd question is for the Texture Optimization,will there a tool for Optimization the tex like Arnold?


Top
Jan_Tomanek
Post  Post subject: Re: about features and Game engine!  |  Posted: Wed May 11, 2016 9:57 am
User avatar

Joined: Tue Oct 27, 2009 11:03 pm
Posts: 1567
Location: Prague - Czech republic

Offline
Do you mean you will anlayze the scene and if is the texture very far, it will load lower resolution and save memory for example?

_________________
Thank you for contacting us.
If you have any further questions, please do not hesitate to contact us.

All the Best
Jan


Top
ymangolds
Post  Post subject: Re: about features and Game engine!  |  Posted: Sun May 22, 2016 10:28 pm
User avatar

Joined: Fri Aug 17, 2012 8:23 am
Posts: 89

Offline
cgpower2008 wrote:
I was wandering if the Furryball can increase the rendering speed as fast as the game engine. Like the Unreal 4, Cinebox,film engine, they have a faster real-time preview speed, and they can be used in the 3D animation production as well.

FB RT is a ray-tracing engine not a real-time (game) engine. So even though it can never be as fast as a game engine, it's capable of doing things easily (and very quickly) that a game engine can never do (like real reflections and indirect lighting which can only be 'faked' using HDR images (and a ton of setup work) in game engines). Also, it utilizes CUDA (GPU) cores for processing so while it'll never be as accurate/high quality as the highend CPU based engines (like Arnold), it can do it waaaaaay faster using only artist machines (instead of needing a million CPU/cloud render farm which still is slower).

cgpower2008 wrote:
Friendly recommending upgrade the rendering speed.
For all of FB's quirks and limitations, one thing it does not lack is render speeds. If your scenes are rendering slow then something's wrong in your settings. There was also a recent update that makes it render stupid fast for the type of calculations it does. Both of these frames were rendered in under 45 seconds each :ugeek: :
Attachment:
test.jpg
test.jpg [ 101.19 KiB | Viewed 4552 times ]


Attachment:
test2.jpg
test2.jpg [ 116.74 KiB | Viewed 4552 times ]



cgpower2008 wrote:
1. Seperating the RT version into 2 versions, TV version and film version. In TV version, the speed is the priority.
Reducing a little bit the quality, and makeing sure no clear noises appears. That will be enoungh for TV production.
FB is the best solution for TV Production already. FB is all about speed so it'll never be an ideal solution for Film where the timeline is long enough and budget high enough to prefer accuracy/quality over anything else (FB attempting to compete with things like Arnold in the 'Extreme Quality' realm of 3D production would be a terrible idea and would fail miserably).

The requirements/criteria for TV Production is similar to the Military Con-Ops work that I do ( final quality is determined by almost entirely on production speed and a low budget). I've researched/tried out most of the other GPU RT engines as well as brief attempts at implementing a game engine rendering pipeline, and FB was the most optimal solution. All GPU renderers are fairly new and they all have their problems/quirks, and the FB staff has been quick to resolve any problems that arise as well as show steady progress with development (FB has also been around the longest for GPU renderers I believe). BTW since so much work is needed to 'fake' effects using game engines the trade-off is that even though render (machine) time is faster, it requires much more human time to create/simulate/fake the effects that FB can do (automatically and accurately) just by checking on the 'use indirect...' check boxes.

cgpower2008 wrote:
2. Developing the system that supporting dual graphics cards, in order to boost the rendering efficiency
It already does. When I switched to FB, my render farm went from a 48-core CPU network of machines to a single workstation that has 4 Titan-X GPU cards in it (I'm even able to render at higher resolution then I used to, and can even set FB to use 3 cards so I can work on the same machine without being impacted by the rendering).


Top
Jan_Tomanek
Post  Post subject: Re: about features and Game engine!  |  Posted: Mon May 23, 2016 10:26 am
User avatar

Joined: Tue Oct 27, 2009 11:03 pm
Posts: 1567
Location: Prague - Czech republic

Offline
ymangolds wrote:
FB is the best solution for TV Production already. FB is all about speed so it'll never be an ideal solution for Film where the timeline is long enough and budget high enough to prefer accuracy/quality over anything else (FB attempting to compete with things like Arnold in the 'Extreme Quality' realm of 3D production would be a terrible idea and would fail miserably).


Thanks - we are happy you like FurryBall such. But try to write us, what is the biggest weakness of FurryBall, when you compare it with Arnold.
Because when you don't count memory limitation, and limitations to writing own shaders, we are not aware any BIG limitation for movie quality you ask. We believe you can achieve same HIGH quality like in Arnold.

BTW, your images are great, but the pixilated edges on the truck are on purpose? Because it look you have so huge scene and huge distance from NEAR to FAR camera clip.
This make this issue for count precision on GPU.

_________________
Thank you for contacting us.
If you have any further questions, please do not hesitate to contact us.

All the Best
Jan


Top
ymangolds
Post  Post subject: Re: about features and Game engine!  |  Posted: Tue May 24, 2016 12:49 am
User avatar

Joined: Fri Aug 17, 2012 8:23 am
Posts: 89

Offline
Jan_Tomanek wrote:
Thanks - we are happy you like FurryBall such. But try to write us, what is the biggest weakness of FurryBall, when you compare it with Arnold.
Because when you don't count memory limitation, and limitations to writing own shaders, we are not aware any BIG limitation for movie quality you ask. We believe you can achieve same HIGH quality like in Arnold.

The biggest difference isn't necessarily a weakness, just that they both were designed with different goals in mind. FB is designed to utilize GPU technology to render things as fast and accurate as possible, which makes it ideal for things like TV production where speed and budget are critical limitations of a production (which also restricts FB to GPU limitations like the count precision you just mentioned). Arnold was designed with the purpose of rendering using the most physically-accurate real-world lighting as possible (with things like speed as an after-thought), which requires a massive amount of CPU processing to calculate (which isn't a problem for the big firms that have a crazy renderfarm, and/or projects that have a comfortable timeline that allows for slow rendering).

Even if the resulting images are identical between the engines, that difference in design philosophy will make the biggest studios choose Arnold (also anyone who has a crazy and expensive renderfarm wants to make use of it). While I would consider using FB for a CG only film, a live-action film with heavy CG elements is better suited to make use of Arnold's physically accurate lighting so that IES lights can be used to match the real-world lighting on set (note that IES lights drastically slow a render down and unless you really want to compete in the CG+live action realm the developement effort to add them would be better spent developing other areas).

Another difference that makes Arnold better suited for CG+live action (or any project that requires a massive amount of post-production) is the power and flexibility of its AOV (render pass) system. Although FB allows you to render almost all (still hoping for motion vector, objectID, coverage, and deep data) the AOVs needed for most projects, Arnold's AOV system allows you to add any custom AOV you can imagine (as well as all the standard passes). So you can have 100's of different AOVs if you need them and don't have to resort to rendering multiple render layers (which renders separately so increases render time). The AOV customization ability also allows rendering out very fancy passes when needed. For example in the scenes of my current project that are extremely volumetric heavy (I only use Arnold for volumetrics/deep data now) I have separate AOV/passes of the volume for different types of light sources effecting it (sun/environment, vehicle lights, and muzzle flashes are AOV groups I use), which allows me to control how the different light groups contribute to the final render in post.

A minor difference is that AOV Driver (output file type, (FB can set differently file types per pass as well so ignore that for this comparison)), and Filter (anti-aliasing) nodes can be created and assigned to AOV's which makes them as customize-able as you want. FB lets you set the aliasing filter globally, and while I can't image why anyone would want to use different filters on different passes, the ability to disable filtering on certain data passes that shouldn't have it (like depth, world position, UVs, normals, etc) is important for the big budget productions where "every pixel must be perfect." Note that if you add the ability to disable filtering per pass (which you should if it's possible, though I wouldn't bother adding the ability to change the filter type per pass), make sure to add the ability to render out a 'coverage' pass (basically a pass that isolates edges of objects, or think of it as a mask pass that indicates where filtering gets applied in the image), so that the unfiltered passes can be selectively filtered in post>post as needed so that they get utilized accurately. Hmmm... I wonder if using no filtering in the renders and applying filtering in post using a coverage pass would resolve or at least minimize the problems caused by the count precision of GPUs...

The other (and biggest) reason not to try competing for the 'film/extreme accuracy' industry is the size of the developing companies and the resources available to them (Arnold is co-developed with Sony and is now owned by Autodesk...). I guess saying that it 'would fail miserably' isn't fair and AAA-Studio might be able to hold their own in the competition. But given the size and resources of the competition, GPU renderers can only be in shadows of dead-accurate CPU engines in that industry. So I should of said that it makes more sense (and will be more successful) focusing on the 'speed + quality' realm of 3d where FB can totally dominate.

Jan_Tomanek wrote:
BTW, your images are great, but the pixilated edges on the truck are on purpose? Because it look you have so huge scene and huge distance from NEAR to FAR camera clip.
This make this issue for count precision on GPU.
That is not intentional and is a result of count precision limitation of GPUs. The shot that those frames are from is long (15 seconds) and uses elaborate camera animation and I didn't even notice the pixilated edges until now. Luckily the convoy vehicles are rendered in a seperate render layer that uses FB's motion blur (at one point the head vehicle almost runs over the camera so needed 3d motion blur for the rotational blur of the tires, and motion blur on everything else as well as DOF is added in post), so I'll rerender the vehicle layer with an optimized near/far camera clip.


Top
Jan_Tomanek
Post  Post subject: Re: about features and Game engine!  |  Posted: Tue May 24, 2016 8:15 am
User avatar

Joined: Tue Oct 27, 2009 11:03 pm
Posts: 1567
Location: Prague - Czech republic

Offline
Thanks for comprehensive compare - great.

So you are absolutely right - we also want to go by path of quality and speed.
But I believe that the future of rendering is GPU for movie and HUGE scenes in the future too. Same as huge datacenter and Learning machines, who change to GPU too.

Now I just arrange with NVIDIA some test for Siggraph for HUGE scenes on their DGX-1 technology over NVLINK. It will be able to render scenes up to 50 GB!!

BTW, please can you write us just LIST of your highlighted features with order of your importance we have to add to FurryBall?

_________________
Thank you for contacting us.
If you have any further questions, please do not hesitate to contact us.

All the Best
Jan


Top
Jan_Smejkal
Post  Post subject: Re: about features and Game engine!  |  Posted: Tue May 24, 2016 9:58 am
User avatar

Joined: Fri Aug 13, 2010 10:30 am
Posts: 794

Offline
ymangolds wrote:
FB lets you set the aliasing filter globally, and while I can't image why anyone would want to use different filters on different passes, the ability to disable filtering on certain data passes that shouldn't have it (like depth, world position, UVs, normals, etc) is important for the big budget productions where "every pixel must be perfect.
You can set different render settings node per render pass (so different multisampling/filtering/features/renderables if you want). However it may cause slowdown.


Top
ymangolds
Post  Post subject: Re: about features and Game engine!  |  Posted: Wed May 25, 2016 2:01 am
User avatar

Joined: Fri Aug 17, 2012 8:23 am
Posts: 89

Offline
Jan_Smejkal wrote:
ymangolds wrote:
FB lets you set the aliasing filter globally, and while I can't image why anyone would want to use different filters on different passes, the ability to disable filtering on certain data passes that shouldn't have it (like depth, world position, UVs, normals, etc) is important for the big budget productions where "every pixel must be perfect.
You can set different render settings node per render pass (so different multisampling/filtering/features/renderables if you want). However it may cause slowdown.
I also can't imagine why someone would want entirely different filters on different AOVs since it would make accurate post a nightmare. I wouldn't bother adding in that feature, but the ability to disable filtering entirely on certain AOVs is important if a production calls for heavy post (also require a coverage pass), especially when doing post-3D things like using point clouds made using world position pass (edge pixels that are filtered go crazy).

Jan_Smejkal wrote:
BTW, please can you write us just LIST of your highlighted features with order of your importance we have to add to FurryBall?
I'll make another thread for that.


Top
Display posts from previous:  Sort by  
Print view

Who is online
Users browsing this forum: No registered users and 1 guest
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum
Jump to:   
cron

Delete all board cookies | The team | All times are UTC + 1 hour [ DST ]

Powered by phpBB® Forum Software © phpBB Group
DAJ Glass 2 template created by Dustin Baccetti