Announcing Redshift - Biased GPU Renderer

Plugins linking to this thread: (hide)

Redshift Renderer 2.5Author: Redshift Rendering Inc.
Version 2.5.32 updated Oct 3rd 2017 / Redshift continues to support the Softimage plugin, posting updates almost weekly. Version 2 added many new features including volumetrics (OpenVDB), nested dielectrics, a new BRDFS response, new SSS models, light baking and a new PBR based redshift material.

From the company's website: Redshift is a powerful GPU-accelerated renderer, built to meet the specific demands of contemporary high-end production rendering. Tailored to support creative individuals and studios of every size, Redshift offers a suite of powerful features and integrates with industry standard CG applications. Biased Rendering: Redshift has the features and uncompromising quality of a CPU renderer, but at GPU rendering speeds. Unlike other GPU renderers out there, Redshift is a biased renderer that allows the user to adjust the quality of individual techniques in order to get the best performance/quality balance for their production. Out-of-Core Architecture: Redshift's efficient memory management allows rendering of scenes containing hundreds of millions of polygons and TBs of texture data. Proxies and Instances: The user can export groups of objects and lights to Redshift Proxy files which can be easily referenced by other scenes. Proxies allow for powerful shader, matte and visibility flag overrides as often required in production. [..] Follow the product link to continue reading.

Pricing information: Price is $500 for a node-locked or $600 for a floating license (5 lic. minimum for floating). Both options include 1 year of maintenance. For other options, see the purchase page.

New plugins, tools etc.
Post Reply
User avatar
Cacoman
Posts: 81
Joined: 02 Aug 2010, 23:11

Re: Announcing Redshift - Biased GPU Renderer

Post by Cacoman » 02 Jul 2013, 01:38

Well, the advantages of the Titan over the GTX 780 are the following

200 More CUDAS
6 gigs against 3 gigs RAM

But the memory bandwith is the same 288.4 GB/sec

So the Titan will render a little bit faster than the 780, and is able to allocate more memory for textures, rays, etc. But frankly, before the Titan I had a GTX470 with 2 gigs in RAM and never got in trouble.

The reports from a guy having both the Titan and the 780 is that the speed is the same, due to the same memory bandwith.

I went to the nvidia site to check on this, and I just found the GTX 690. I haven't seen that before. I guess is new. But from what I just read from the specs sheet, seems to me that actually this one is the winner.

3072 CUDAS vs 2688 of the Titan
Both 6 Gigs
and the memory bandwith is higher than the Titan.
384 gigs/sec vs 288.4 of the Titan.

Damned!!! This was not around when I bought the Titan....
To create something that does not exist, to exist.

Falam

Re: Announcing Redshift - Biased GPU Renderer

Post by Falam » 02 Jul 2013, 02:17

The 690 came before the Titan, it is more similar to a dual card setup. As for regular, model, texturing, ICE operations, a 5xx,6xx,7xx series cards are more then adequate. If you want the extra RAM, you could spend the $1K (US) on a Titan, I don't really think it is worth it unless you are doing major GPU rendering, as with RedShift as you do.

scaron
Posts: 119
Joined: 08 Jul 2009, 05:16

Re: Announcing Redshift - Biased GPU Renderer

Post by scaron » 03 Jul 2013, 02:55

@cacoman, @draise... thanks for the feedback, i will just need to try it myself.

my questions are because of how arnold and vray are integrated into softimage, its not their fault sometimes it is softimage's fault because their sdk doesn't provide the necessary granularity to match the mental ray integration. the scenarios i outline are difficult areas of the sdk in regards to rendering plugins, i wanted to see if redshift has the same issues or if they have workaround these issues :)

s

User avatar
ActionArt
Posts: 853
Joined: 25 Nov 2010, 18:23
Location: Canada

Re: Announcing Redshift - Biased GPU Renderer

Post by ActionArt » 03 Jul 2013, 04:15

If I understand you correctly, no it does not have the same issues. I'll check further to see if this is the case. Integration is much like MR (better in many ways) and if you modify something on a big scene it does not seem to have to re-transfer the entire scene.

User avatar
ActionArt
Posts: 853
Joined: 25 Nov 2010, 18:23
Location: Canada

Re: Announcing Redshift - Biased GPU Renderer

Post by ActionArt » 03 Jul 2013, 04:20

scaron wrote:-if you have a large scene with many objects and you do a modeling change on one object, does the entire scene need to be reprocessed/exported?
-is instancing supported? through ICE?
-what happens when you do modeling changes to an object which is instanced?
-is there a fly through mode? ie. do you have to re export the scene every time the frame changes even if it is static?
-what happens if you import an softimage model from disk? does the entire scene need to be reprocessed/exported?
1) No, it does not seem to. I'll check further. It exports very fast with minor changes so I think not.
2) Yes, instancing is well supported. For ICE, I think so, will check.
3) Yes there is a flythrough mode. Even without flythrough, only changes to the scene are exported. It's very fast. (less than a second or two in most cases).
4) I don't think so. Will check.

Only partially helpful but I'll look into it further.

Kzin
Posts: 432
Joined: 09 Jun 2009, 11:36

Re: Announcing Redshift - Biased GPU Renderer

Post by Kzin » 03 Jul 2013, 12:04

without writing to much, the best thing about redshift compared to mr, all is working like it is intended. and you dont have to care about render features which the renderer can, but no one implemented it. you get whats on the renderers list because the renderer is under support which makes bugfixing alot faster.

User avatar
ActionArt
Posts: 853
Joined: 25 Nov 2010, 18:23
Location: Canada

Re: Announcing Redshift - Biased GPU Renderer

Post by ActionArt » 03 Jul 2013, 16:30

I've had a chance to double check on some of this:
scaron wrote:-if you have a large scene with many objects and you do a modeling change on one object, does the entire scene need to be reprocessed/exported?
-is instancing supported? through ICE?
-what happens when you do modeling changes to an object which is instanced?
-is there a fly through mode? ie. do you have to re export the scene every time the frame changes even if it is static?
-what happens if you import an softimage model from disk? does the entire scene need to be reprocessed/exported?
1) No. RS is smart enough to know what has changed and what hasn't.
2) Instances and ICE instancing are well supported.
3) Changes to instances seem to happen immediately.
4) There is a flythrough mode for lighting purposes but if the scene has animated parts only those parts are updated between frames which is very fast.
5) As with everything else, RS seems smart enough to know what has changed and what hasn't.

Hope that helps :)

scaron
Posts: 119
Joined: 08 Jul 2009, 05:16

Re: Announcing Redshift - Biased GPU Renderer

Post by scaron » 11 Jul 2013, 21:28

ActionArt wrote:I've had a chance to double check on some of this:

1) No. RS is smart enough to know what has changed and what hasn't.
2) Instances and ICE instancing are well supported.
3) Changes to instances seem to happen immediately.
4) There is a flythrough mode for lighting purposes but if the scene has animated parts only those parts are updated between frames which is very fast.
5) As with everything else, RS seems smart enough to know what has changed and what hasn't.

Hope that helps :)
yes, this does! thanks...

El Burritoh
Posts: 151
Joined: 30 Nov 2010, 19:56
Location: Tennessee
Contact:

Re: Announcing Redshift - Biased GPU Renderer

Post by El Burritoh » 02 Sep 2013, 17:41

Sorry I'm late here... I've just been so busy using Redshift on any project I possibly can... ;-)

Steve, I'm surprised you're not testing this yourself. I think your input would be valuable. But I know there's the relationship with Solid Angle.

For being an Alpha product, with a long roadmap yet ahead, I'm downright impressed with the variety of projects I'm able to render with Redshift already. And the beauty of it is it's seamless integration with Softimage. From my experience, it scales quite nicely. That said, I've never done anything like those crazy Whiskey Tree shots with trillions of polygons, and although RS suits our purposes at the commercial/broadcast/cinematic level, I'm simply not in a position to comment on it in terms of big money shots like the wide vistas of Elysium. I've just not had to render anything of that magnitude yet, so I can't say for sure. That's why I wish someone from studios like Whiskey Tree could join the alpha. Users in your position would have some valuable insight into Redshift's use at the heavier end of the scale.

Redshift is not even in beta yet. Considering how young it is, it's remarkably mature. Nic, Panos, and Rob have something to truly be proud of. I can't wait to see where it goes, and we continue to use it on any production we can (so far my last 5-6 jobs have all been in RS).
-Tim Crowson
High Nerd, of the Order of Magnetic Nerds
Personal Website

adrencg
Posts: 196
Joined: 05 May 2012, 00:50

Re: Announcing Redshift - Biased GPU Renderer

Post by adrencg » 02 Nov 2013, 00:31

I just finished a medical animation job with Redshift (2 minutes) and it was the rendering portion was the easiest time I've ever had -- in terms of scene setup and especially render times. At 720p, the average time per frame was 8 minutes. That average is kind of skewed because there was one trouble scene that was 30 minutes per frame, but most of the shots were around 3 minutes per frame. That scene now renders pretty well because they released a fix that solved some issues. They're great about listening to the users.

User avatar
McNistor
Posts: 605
Joined: 06 Aug 2009, 17:26

Re: Announcing Redshift - Biased GPU Renderer

Post by McNistor » 02 Nov 2013, 05:06

Any support for OpenCL coming in the future?
Radeon cards are cheaper and stronger than nvidia's which I ditched a while ago.
The society that separates its scholars from its warriors will have its thinking done by cowards and its fighting done by fools.
-Thucydides

BenR
Posts: 92
Joined: 11 Jun 2009, 20:52

Re: Announcing Redshift - Biased GPU Renderer

Post by BenR » 02 Nov 2013, 14:02

McNistor wrote:Any support for OpenCL coming in the future?
Radeon cards are cheaper and stronger than nvidia's which I ditched a while ago.
Well here's the "coming soon" page: https://www.redshift3d.com/products/coming-soon
No idea about timeline.

benoit81
Posts: 73
Joined: 03 Oct 2011, 20:51

Re: Announcing Redshift - Biased GPU Renderer

Post by benoit81 » 12 Nov 2013, 10:21

interview of Redshift, Nicolas Burtnyk (CEO) :
http://www.3d-test.com/contenu.php?id=2747

mkr
Posts: 11
Joined: 28 Sep 2011, 13:59

Re: Announcing Redshift - Biased GPU Renderer

Post by mkr » 12 Nov 2013, 12:59

So open beta coming soon... nice.

User avatar
owei
Administrator
Posts: 840
Joined: 03 Jun 2009, 17:25
Location: Siegen/Germany
Contact:

Re: Announcing Redshift - Biased GPU Renderer

Post by owei » 12 Nov 2013, 13:06

..and you may already be able to apply to the beta here:
https://www.redshift3d.com/get-redshift

cheers,
oli

BenR
Posts: 92
Joined: 11 Jun 2009, 20:52

Re: Announcing Redshift - Biased GPU Renderer

Post by BenR » 12 Nov 2013, 17:14

benoit81 wrote:interview of Redshift, Nicolas Burtnyk (CEO) :
http://www.3d-test.com/contenu.php?id=2747
Thanks for the link!
Interesting interview. One message though is that OpenCL is on the "long-term roadmap" and they haven't done much work on it yet. Perhaps it should not be on Redshift's "coming soon" page.

As others have said, I have really enjoyed testing Redshift and am really impressed by the work they are doing on integration.

Post Reply

Who is online

Users browsing this forum: No registered users and 32 guests