Announcing Redshift - Biased GPU Renderer

Become a member of the CGSociety

Connect, Share, and Learn with our Large Growing CG Art Community. It's Free!

THREAD CLOSED
 
Thread Tools Search this Thread Display Modes
Old 03 March 2013   #31
looks very nice!

at jumamu, vray on c4d numbering is different than on maya, the 1.2.6 version that is available uses the vray 2.35 core, and implements already many of the 2.3 features. some yet missing version features are coming in the next update (see the c4d vray forum). also full mb despite the lack of mb in c4d sdk.

gratulations to the redshift team anyway, looks very interesting

cheers
stefan
__________________
V-Ray on FB:: http://www.facebook.com/vrayforc4d
http://VRAYforC4D.net
http://www.3dtools.info
http://rhino.io

Last edited by lllab : 03 March 2013 at 04:31 PM.
 
Old 03 March 2013   #32
Thanks for the words of encouragement guys, we really appreciate it!

Regarding Cinema4D support: point well taken!

Regarding noseman's question about 8GB texture scenes: Yeah, I believe your scenes should work fine. Redshift does a background conversion on all scene textures when it first encounters them (and caches the result to disk so you don't have to wait over and over again). This includes tiling and making mip-maps for them. From that point onwards, it loads only the tiles of the textures it needs and only at the resolutions it needs. This ensures that closeup detail is as high as possible (full-rez) while distant detail is properly (via elliptical filtering) softened and does not alias/flicker.

Regarding passes: support for these will be added in the next couple of months. We fully appreciate that this is a super-important feature for production people given the usual post work.

Finally regarding animations and more pictures: yep, we know we are extremely content-light at the moment. We'll probably create a movie or two showing the product in action later today. Also some people had questions on the quality of the irradiance cache on animations so might make some short movies on that too.

Thanks!

-Panos
 
Old 03 March 2013   #33
This is looking very interesting indeed!

How does it deal with rendering at very high resolution? Does it store the frame buffer in RAM instead of VRAM? And what about distributed rendering?
 
Old 03 March 2013   #34
Hi CaptainObvious,

The rendering is partitioned in tiles, similar to other biased renderers. So the entire frame buffer does not have to be resident in VRAM. We've tested Redshift with 8K renders and it worked fine.

Having said that, the space needed for the point-based techniques does have some relevance to the framebuffer size so *their* sizes could grow (talking about irradiance cache, SSS, etc). But, unless the scene is insanely highly detail (has extremely few flat areas), the growth is sub-linear. Meaning: if you need X amount of megabytes for the irradiance cache at some resolution and then you grow that resolution 4-fold, the irradiance cache size (and computation time) most likely won't grow 4-fold.

Also, with very high resolutions, the texture cache will be utilized more since everything will require more detail.

Finally, network rendering is not currently supported but will be added in a few months.

Hope this helps

-Panos
 
Old 03 March 2013   #35
This looks very interesting, are there plans for a standalone renderer?
 
Old 03 March 2013   #36
Originally Posted by elindell: This looks very interesting, are there plans for a standalone renderer?


We do have an (in-house) scene exporter and a standalone executable that can load the exported scene file and render it. Unfortunately, though, it's not fully featured yet and customized primarily for our own testing/profiling purposes.

We are currently prioritizing development effort on features. Once the list of missing features shrinks down a bit, we'll look into these kinds of tools/APIs.

Thanks

-Panos
 
Old 03 March 2013   #37
Originally Posted by PanosRS: Finally, network rendering is not currently supported but will be added in a few months.



With Network Rendering you mean Distributed, right? Is it possible to use several GPU from different Machienes?

Another Question: How Much does CPU play a role? Is CPU also used when Calculating Irradicane Cache for example? Or is only GPU used?
__________________
____
Cinema 4D Studio MSA, Zbrush, CS6, 3DCoat, Moi
Twitter: @holgerbiebrach
www.c4dStuff.com

HB_ModellingBundle
 
Old 03 March 2013   #38
Originally Posted by HolgerBiebrach: With Network Rendering you mean Distributed, right? Is it possible to use several GPU from different Machienes?

Another Question: How Much does CPU play a role? Is CPU also used when Calculating Irradicane Cache for example? Or is only GPU used?


Hello HolgerBiebrach,

Yep, by "network rendering" I meant "distributed rendering"! . Multi-GPU support is also something that will be coming up in the next few months.

The CPU plays a relatively small role in Redshift. We do use it for certain things like building the ray tracing acceleration structure. Also during irradiance caching and other point based techniques, it helps out with rebalancing some data structures on the videocard which, in turn, helps the GPU performance.

We haven't (so far) seen a situation where the CPU performance was dominating the final rendering time. That *could* happen if the scene contained many tens of millions of triangles but the scene used no GI, reflections/refractions and area lights. If that was the case (unlikely!) the GPU would be done almost instantly but would first have to wait for the CPU to 'feed it' the triangles.

On a similar note, extracting several million triangles from XSI and Maya can take a bit of CPU time. For this reason we use caches so that only geometry that was modified by you is re-extracted. This way you don't have to wait seconds each time you move the camera on a high-poly scene. Our Redshift proxies also help a lot with that (they eliminate that extraction time).

-Panos
 
Old 03 March 2013   #39
Thank you for all that good and detailed explanations. Really appreciate it.
This sounds all very promising....really hope for OSX and CINEMA 4D support.
__________________
____
Cinema 4D Studio MSA, Zbrush, CS6, 3DCoat, Moi
Twitter: @holgerbiebrach
www.c4dStuff.com

HB_ModellingBundle
 
Old 03 March 2013   #40
So how about advanced texturing? Is it possible to blend together multiple different textures in a single material channel and such? A major limitation with most of the crop of GPU renderers right now is that each material channel only accepts one image map. My texturing workflow usually involves loads of different textures blended together with other textures as masks.
 
Old 03 March 2013   #41
Since so many people here are voicing in favor or C4D (no problem with that!) let me be a voice in favor of 3dsMax. I know a version of this render engine is in your plans (or already in the works?) but I just want to show support to this project.

This renderer is what I expected VRayRT and iRay to have been - a fast, useful GPU renderer that can be used in a real production environment for animations. Sure, for packshots these mentioned render engines can do the job, but start adding geometry and/or high res textures to see your graphics card choke and pass the job to the CPU. I'm sick of this "unbiased" path tracing wave - I just want images to look beautiful, not necessarily 100% physic accurate.

I believe you have your proprietary set of materials type, correct? Would be nice if it could convert previous MR/Vray material to whatever format you use.

And don't forget - hurry up with the 3dsMax version! How can I sign up for a alpha/beta for this version?

Cheers!
 
Old 03 March 2013   #42
Originally Posted by CaptainObvious: So how about advanced texturing? Is it possible to blend together multiple different textures in a single material channel and such? A major limitation with most of the crop of GPU renderers right now is that each material channel only accepts one image map. My texturing workflow usually involves loads of different textures blended together with other textures as masks.


Hi CaptainObvious,

Redshift supports a wide variety of shading nodes so its possible to combine colors or scalars together (add/multiply/subtract/blend/etc) irrespective of whether these colors came from textures, a color picker or other nodes. The output of any node can then be connected to any material input. So blending two textures should work fine.

You can even do slightly more exotic things like grab the green channel of the shading normal (the 'up' component) and use it to blend a 'dust map' over your base texture so that only faces pointing up get the dust. Just a small example.

So short answer: yes we support that!

-Panos
 
Old 03 March 2013   #43
Originally Posted by davius: Since so many people here are voicing in favor or C4D (no problem with that!) let me be a voice in favor of 3dsMax. I know a version of this render engine is in your plans (or already in the works?) but I just want to show support to this project.

This renderer is what I expected VRayRT and iRay to have been - a fast, useful GPU renderer that can be used in a real production environment for animations. Sure, for packshots these mentioned render engines can do the job, but start adding geometry and/or high res textures to see your graphics card choke and pass the job to the CPU. I'm sick of this "unbiased" path tracing wave - I just want images to look beautiful, not necessarily 100% physic accurate.

I believe you have your proprietary set of materials type, correct? Would be nice if it could convert previous MR/Vray material to whatever format you use.

And don't forget - hurry up with the 3dsMax version! How can I sign up for a alpha/beta for this version?

Cheers!


Hi Davius,

Regarding your question on materials: we do have our own list but we also support several of the default XSI/Maya nodes. They are all listed in the documentation. We do try to make it as painless as possible to transition from other renderers but sometimes some tweaking will be needed as we do things differently here and there. For example we have found specular and glossiness looking different between different renderers.

As for the 3dsMax version: yeah, on the tools side of things this is on the very top of the list!

-Panos
 
Old 03 March 2013   #44
Originally Posted by PanosRS: So short answer: yes we support that!

Great!

I suppose what you're doing is replicating the functionality of the built-in nodes inside RedShift, and then basically translate, right?
 
Old 03 March 2013   #45
Originally Posted by CaptainObvious: Great!

I suppose what you're doing is replicating the functionality of the built-in nodes inside RedShift, and then basically translate, right?


That's exactly right! Check out our docs for a complete list of what is translated.
 
Thread Closed share thread



Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
CGSociety
Society of Digital Artists
www.cgsociety.org

Powered by vBulletin
Copyright 2000 - 2006,
Jelsoft Enterprises Ltd.
Minimize Ads
Forum Jump
Miscellaneous

All times are GMT. The time now is 06:40 PM.


Powered by vBulletin
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.