Announcing Redshift - Biased GPU Renderer

[Nicolas Burtnyk]

Hello folks,

In March of last year, 2 colleagues and I left our jobs as software developers in the games industry to form our own company - Redshift.

Our goal was to apply our experience with graphics hardware to the problem of "offline" rendering.

Artists friends had been asking us for years why Mental Ray and other renderers were not taking advantage of the GPU.

As the ideas bounced around in our heads, we figured we'd take a crack at it.  As it turns out, it's really freakin' hard, but not impossible!

Today, we're very excited to announce the official launch of Redshift v0.1 alpha, to our knowledge, the world's first fully GPU-accelerated biased renderer.

Redshift supports multiple GI solutions: Brute-Force GI, Irradiance Caching (aka Final Gather), Irradiance Point Cloud (aka Light Cache) and Photon Mapping (GI and Caustics).

All are fully GPU-accelerated and perform many times faster than similar CPU-based offerings.

A problem that plagues many GPU renderers on the market is that they are limited by the available VRAM on the graphics card (and most systems have significantly less VRAM than main memory).  Redshift addresses this by using an out-of-core architecture for geometry and textures allowing you to render scenes with tens of millions of polygons and gigabytes of textures with off-the-shelf, inexpensive hardware.

Redshift currently integrates directly with Softimage 2011 through 2013 and Maya 2011 through 2013 on Windows XP or higher.  3ds Max support is in development.  To run Redshift, you'll need an NVidia graphics card supporting compute 1.2 or higher with 1GB VRAM or more.

You can check out our website http://www.redshift3d.com for more info.

We're currently in closed alpha, and looking to find some interested alpha testers on this list.

Our goals for alpha are to shake out bugs and gather feedback from users to help focus our development efforts.

If you're interested in taking Redshift for a spin, drop me an email at nic...@redshift3d.com.

Some sample renders:

http://s3.redshift3d.com/jp_studio_icp_1280.png

http://s3.redshift3d.com/mazda_1024.png

http://s3.redshift3d.com/AI_V8_S10_1200.png

And the classroom scene from the "Octane Render" thread from a couple weeks ago. 2 minutes on a single GeForce GTX 470.

http://s3.redshift3d.com/classroom.png

I'd also like to thank all of those on this list who've answered my various questions about the XSI SDK.  Your help has made our integration of Redshift with Softimage very strong.

Cheers!

Nicolas

[Steven Caron]

congrats to you and your team! i was wondering when we would see/hear about your work.

it would be great to see a video demonstration of redshift in softimage.

[Nicolas Burtnyk]

Thanks Steven!

We'll try to whip something together.  Our team consists of 3 software developers and we've been plugging away at the code side of things pretty hardcore for the last year.  We're admittedly very light on content, something we hope to rectify during this alpha period - in some part at least by having our alpha testers share some renders with us.

Speaking of videos - does anyone have any pointers on taking video screen captures of XSI?  Any software in particular that we should look at, or any software to avoid?

-Nicolas

[olivier jeannel]

The classroom is really 2min render ?

Congrats to you, sending a request :)

[Nicolas Burtnyk]

Yes - I'll try to make a video of that if I can get set up correctly for it.

Note that this is 2 mins on a GTX 470 which is nothing special in terms of GPUs.  You can expect significantly better times with a GTX 580 for example.  I don't have official times for that card, but I'd guess under 1.5 minutes.

These kinds of times really underscore the power of biased rendering.  When you need to reduce noise, you have a lot more options than "let's just throw a ton more samples at the whole thing".

[Alok]

Congrats on your achievement, looking forward to see more of it.

Just a quick question. For most of the render farm setups, usually the machines do not have high-end GPU's on them for the obvious reason that farm nodes never get user interaction and hence no need for realtime graphic intensive processing operations. Do you see that as a problem ? If yes then how would you address it. Does Red Shift turn to CPU for rendering in absence of the required GPU.

ALOK

GANDHI

/ chef directeur technique - lead technical director

alok....@modusfx.com

T:

450 430-0010 x225

F:

450 430-0009
www.modusfx.com

---------------------------------------------------------------------------------------------

MODUS

FX

120 Rue Turgeon,

Sainte-Therese (Quebec) CANADA J7E 3J1

Follow us on

Facebook

&

Twitter

No virus found in this message.
Checked by AVG - www.avg.com
Version: 2012.0.2238 / Virus Database: 2641/5646 - Release Date: 03/03/13
Internal Virus Database is out of date.

[Tim Crowson]

That's looking quite interesting, Nicolas! Can't wait to know more!
-Tim C.
Magnetic Dreams

[Len Krenzler]

Exactly!  This is VERY interesting.  Hope I can test (sent request already) :)

-- 
_________________________________________________

Len Krenzler - Creative Control Media Productions

Phone: 780.463.3126

www.creativecontrol.ca - l...@creativecontrol.ca

[Nicolas Burtnyk]

Thanks Alok!

That'a an excellent question.

We don't currently have any official support for farm rendering.  That being said, with a suitable render manager or some hand rolled scripts, I don't see why you couldn't have multiple machines working on different frames of an animation.  And of course, we do plan on having more complete render farm support in the future.

Your concern regarding the hardware configuration of render nodes is very valid.  Redshift does require a GPU and we do not fall back on CPU in the absence of a suitable GPU.  The performance gap is so large that it didn't really make sense for us at this stage to invest the time in making this kind of thing work (though we could entertain this possibility).  Unfortunately, this does mean that some farms will not be good candidates for Redshift, particularly those in which adding a GPU to some or all of the nodes is not possible (e.g blades with no space in them).

Keep in mind though that with the performance improvement factor provided by Redshift, you can get the performance of a farm with a dozen nodes from a single GPU-equipped node.  We don't want to make any hard performance claims yet, but our internal tests are often showing 10x to 20x the performance of MR and VRay, sometimes more (on suitably complex scenes - super simple 10 seconds to render in MR stuff doesn't get 10x faster).  So maybe you keep your existing farm for CPU renders and add a few GPU equipped nodes for Redshift.

[Daryl Dunlap]

Congrats on getting this up and going.  I signed up for access.  Do you have any ballpark for where you think your price point will come in at?

[Mirko Jankovic]

grats on progress.

I guess that renderer is tightly bind to CUDA and OpenCL versions like for AMD cards are not in plan?

[Nicolas Burtnyk]

Unfortunately, I can't provide any solid pricing info yet, but we'd like to keep the price accessible to everyone so you can expect it to be priced competitively (and perhaps cheaper) compared to the other renderers.  We're also considering a couple of pricing tiers, but that's still all TBD.

[Steven Caron]

http://www.redshift3d.com/support/faqs/

:)

[Nicolas Burtnyk]

Redshift does indeed use CUDA and is therefore currently limited to NVidia hardware.

However OpenCL support is definitely in our plan, though not in the immediate term (in other words it's not something we're working on at the moment).

On Wed, Mar 13, 2013 at 12:32 PM, Mirko Jankovic <mirkoj....@gmail.com> wrote:

[Daryl Dunlap]

Steve, thanks for the link to the FAQ.  I'm also pleasantly surprised to see some Documentation there as well. :-]

[Matt Morris]

Looking forward to seeing some more samples, just wondering why you continue to support 32-bit, given that autodesk have dropped support in the more recent versions? Does it add much overhead? Scuse my uninformed ponderings ;)

--
www.matinai.com

[Nicolas Burtnyk]

Yes we have some docs up on the site too!  Thanks Daryl for mentioning that.

Please be aware though that the docs are not 100% complete and not super clean.  But they do contain lots of good info and should definitely help people get started.

[François Painchaud]

Great stuff! With the proper low/no GI configuration, would it be possible to use Redshift to do near-realtime playback in Softimage?

[Eugen Sares]

Congratulations, too, for your achievements! It's so cool to see how
small teams of idealists can take the technological lead... wouldn't you
expect something like this from nVidia themselves?

Am 13.03.2013 21:03, schrieb Fran�ois Painchaud:
> Great stuff! With the proper low/no GI configuration, would it be
> possible to use Redshift to do near-realtime playback in Softimage?

Like a replacement for the HQ viewport? Ha!
Which rises the question: does it support progressive refinement?

[Nicolas Burtnyk]

We support 32 bit because it's not any more difficult to do so and doesn't impact the performance or anything on 64 bit.

Basically, why not? :)

[Stefan Kubicek]

> Like a replacement for the HQ viewport? Ha!
> Which rises the question: does it support progressive refinement?

From http://www.redshift3d.com/products/redshift :

"1-Click Progressive Rendering

Redshift provides a progressive rendering mode that allows you to conveniently and interactively preview your scene without waiting for GI prepasses. This mode uses progressive refinement to give you draft quality results almost immediately while continuously refining the quality over time."




--
-------------------------------------------
Stefan Kubicek
-------------------------------------------
keyvis digital imagery
Alfred Feierfeilstraï¿œe 3
A-2380 Perchtoldsdorf bei Wien
Phone: +43/699/12614231
www.keyvis.at ste...@keyvis.at
-- This email and its attachments are --
--confidential and for the recipient only--

[Daryl Dunlap]

Nicolas, the docs say you support model instancing.  But, do you also support instancing via PointClouds (Particle instancing)?

[Nicolas Burtnyk]

@François: Unfortunately due to some fixed overhead on our side and things like scene translation, Redshift will not render at 30 frames per second even with all features turned down/off.  That being said, depending on resolution and settings, you could get renders on the order of 1-5 seconds or possibly less for very simple things (like shaderballs).  Also, progressive rendering can give you (noisy) results that approximate the final render very quickly (see below).

@Eugen: Redshift has progressive rendering mode which provides noisy results almost instantly and refines over time.  Final renders don't use this mode, but it's very useful while setting up your scene and tweaking shading/lighting.

On Wed, Mar 13, 2013 at 1:19 PM, Eugen Sares <soft...@keyvis.at> wrote:

Congratulations, too, for your achievements! It's so cool to see how small teams of idealists can take the technological lead... wouldn't you expect something like this from nVidia themselves?

[Nicolas Burtnyk]

At this moment Redshift does not understand particle instancing.  However we can add support for this very easily.

[Daryl Dunlap]

Yeah, it very straightforward, you read the Particle 'Shape' attribute.  Then compare that value against the ICEShapeType enum, if the ICEShapeType is Reference, then you have a Particle Instance of a object in the scene.  You can then take that value and get the actual Object by various means, one of them being thru Dictionary.GetObject().  Once you have the object, you can read it's Type value to determine what kind of object (model, polymesh, light, etc.) is being instanced, and then go from there.

[Nicolas Burtnyk]

Hey guys,

Thanks for the great responses.  I think I've responded to everyone who sent an alpha request, but if you think I missed you, please shoot me an email to remind me :)

Also, I wanted to share some render times as we ran some more tests this afternoon comparing the GTX 470, GTX 670 and GTX Titan (which we actually received after the announcement went out) for the scenes we posted.

Some of this info is also on our announcement thread on CGTalk (http://forums.cgsociety.org/showthread.php?f=59&t=1098062).

Gargoyle 1280x720 (jp_studio_icp_1280.png)

GTX 470: 35 seconds

GTX 670: 27 seconds

GTX Titan: 17 seconds

Car 1024x683 (mazda_1024.png)

GTX 470: 75 seconds

GTX 670: 65 seconds

GTX Titan: 39 seconds

Evermotion Living Room 1200x1000 (AI_V8_S10_1200.png)

GTX 470: 155 seconds

GTX 670: 123 seconds

GTX Titan: 77 seconds

Classroom 1024x512 (classroom.png)

GTX 470: 129 seconds - ok I exaggerated a bit when I said 2 minutes :)

GTX 670: 96 seconds

GTX Titan: 49 seconds

Please also bear in mind that we're still just in alpha and constantly improving performance.  We're kind of obsessed with speed :)

[Ahmidou Lyazidi]

77 secs for the Living room, that's impressive!!
-----------------------------------------------
Ahmidou Lyazidi
Director | TD | CG artist
http://vimeo.com/ahmidou/videos


2013/3/14 Nicolas Burtnyk <nic...@redshift3d.com>:

>>> Le 13/03/2013 19:18, Steven Caron a �crit :

[Manuel Huertas Marchena]

Hi Nicolas,

This looks very interesting, props to you guys, I am looking forward for the video as well, if possible!
I have a question, what about framebuffer support in softimage with redshif, is it similar worflow as mr/vray? ...Haven't found that on the documentation, maybe I missed that.

Thanks

Cheers


-Manuel

> Date: Thu, 14 Mar 2013 13:01:28 +1100
> Subject: Re: Announcing Redshift - Biased GPU Renderer
> From: ahmid...@gmail.com
> To: soft...@listproc.autodesk.com

[Jason S]

I find the avoidance of GPU ram limitations to be quite something,
which is usually a (if not *the*) major drawback of GPU rendering.

That along with aproximation methods which are also unusual,
made me quite curious.

Sent my trial request too :)

congrats!
Jason

[Nicolas Burtnyk]

Hi Manuel,

Redshift doesn't currently support multiple framebuffers (or "render elements"), but it's in the plan.

-Nicolas

[pet...@skynet.be]

> Please also bear in mind that we're still just in alpha and constantly improving performance.  We're kind of obsessed with speed :)

 

speed is great of course – but IMO it’s not the most important factor.

 

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely “it has to be rendered in X amount of time” – clients couldn’t care less. It has to be good enough first and rendered in time for delivery.

 

it’s been a long time I’m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff – but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering – that can be very precisely tweaked by the user.

 

It’s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require – so in turn you have to bite the bullet and go back to good old offline rendering – and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features – not just GI / AO / softshadows / softreflections – but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair – and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

 

Mental ray never became the standard it was because of speed – but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it’s features are getting long in the tooth as well, which opens the door wide open for others – but it remains a reference for what a renderer should at least aspire to.

 

just some thoughts and hints of what matters to me when considering a new renderer.

[Tim Crowson]

Those are some impressive render times!
-Tim

--

[Andy Moorer]

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX  elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

[pet...@skynet.be]

you are right of course, as always.

 

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

 

 

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

 

[Nicolas Burtnyk]

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

[Steven Caron]

thanks a lot for the video, gives me a good idea of the integration. honestly i have no time for testing... but when you guys announce a price i will see if its right for me to jump on :)

[Alok Gandhi]

Wow !

Sent from my iPhone

[Sylvain Lebeau]

killer!!!!
congrats to you and team Nicolas!!

sly

Sylvain Lebeau // SHED
V-P/Visual effects supervisor
1410, RUE STANLEY, 11E �TAGE MONTR�AL (QU�BEC) H3A 1P8
T 514 849-1555 F 514 849-5025 WWW.SHEDMTL.COM <http://WWW.SHEDMTL.COM>

On 3/14/2013 10:35 PM, Nicolas Burtnyk wrote:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

�

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

�

�

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

�

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX� elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.� We're kind of obsessed with speed :)

�

speed is great of course � but IMO it�s not the most important factor.

�

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely �it has to be rendered in X amount of time� � clients couldn�t care less. It has to be good enough first and rendered in time for delivery.

�

it�s been a long time I�m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff � but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering � that can be very precisely tweaked by the user.

�

It�s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require � so in turn you have to bite the bullet and go back to good old offline rendering � and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features � not just GI / AO / softshadows / softreflections � but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair � and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

�

Mental ray never became the standard it was because of speed � but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it�s features are getting long in the tooth as well, which opens the door wide open for others � but it remains a reference for what a renderer should at least aspire to.

�

[Emilio Hernandez]

Let me tell you that I just put my hands on this baby and wow!!!   This is going to rock the rendering world.  And for Softimage!!!! 

Awsome guys congratulations on this one.  My quadro 3000 finally is awake!!!

2013/3/14 Sylvain Lebeau <s...@shedmtl.com>

killer!!!!
congrats to you and team Nicolas!!

sly

Sylvain Lebeau // SHED
V-P/Visual effects supervisor

1410, RUE STANLEY, 11E ÉTAGE MONTRÉAL (QUÉBEC) H3A 1P8

T 514 849-1555 F 514 849-5025

WWW.SHEDMTL.COM <http://WWW.SHEDMTL.COM>

On 3/14/2013 10:35 PM, Nicolas Burtnyk wrote:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

 

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

 

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

 

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX  elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.  We're kind of obsessed with speed :)

 

speed is great of course – but IMO it’s not the most important factor.

 

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely “it has to be rendered in X amount of time” – clients couldn’t care less. It has to be good enough first and rendered in time for delivery.

 

it’s been a long time I’m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff – but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering – that can be very precisely tweaked by the user.

 

It’s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require – so in turn you have to bite the bullet and go back to good old offline rendering – and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features – not just GI / AO / softshadows / softreflections – but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair – and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

 

Mental ray never became the standard it was because of speed – but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it’s features are getting long in the tooth as well, which opens the door wide open for others – but it remains a reference for what a renderer should at least aspire to.

 

just some thoughts and hints of what matters to me when considering a new renderer.

--

[Eugen Sares]

Guys, this is fantastic! Exactly the simple workflow and high speed anyone should come to expect these days!
It had to be a bunch of true independent "nerds" to pave the path, again... (meant as a compliment!)
I haven't got resources left for testing, but I'm very much looking forward to 1.0.

Am 15.03.2013 03:35, schrieb Nicolas Burtnyk:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

�

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

�

�

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

�

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX� elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.� We're kind of obsessed with speed :)

�

speed is great of course � but IMO it�s not the most important factor.

�

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely �it has to be rendered in X amount of time� � clients couldn�t care less. It has to be good enough first and rendered in time for delivery.

�

it�s been a long time I�m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff � but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering � that can be very precisely tweaked by the user.

�

It�s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require � so in turn you have to bite the bullet and go back to good old offline rendering � and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features � not just GI / AO / softshadows / softreflections � but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair � and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

�

Mental ray never became the standard it was because of speed � but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it�s features are getting long in the tooth as well, which opens the door wide open for others � but it remains a reference for what a renderer should at least aspire to.

�

[Arvid Björn]

Very interesting indeed! Definitely shooting you guys an email! :)

On Fri, Mar 15, 2013 at 10:50 AM, Eugen Sares <soft...@keyvis.at> wrote:

Guys, this is fantastic! Exactly the simple workflow and high speed anyone should come to expect these days!
It had to be a bunch of true independent "nerds" to pave the path, again... (meant as a compliment!)
I haven't got resources left for testing, but I'm very much looking forward to 1.0.


Am 15.03.2013 03:35, schrieb Nicolas Burtnyk:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

 

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

 

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

 

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX  elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.  We're kind of obsessed with speed :)

 

speed is great of course – but IMO it’s not the most important factor.

 

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely “it has to be rendered in X amount of time” – clients couldn’t care less. It has to be good enough first and rendered in time for delivery.

 

it’s been a long time I’m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff – but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering – that can be very precisely tweaked by the user.

 

It’s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require – so in turn you have to bite the bullet and go back to good old offline rendering – and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features – not just GI / AO / softshadows / softreflections – but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair – and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

 

Mental ray never became the standard it was because of speed – but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it’s features are getting long in the tooth as well, which opens the door wide open for others – but it remains a reference for what a renderer should at least aspire to.

[Len Krenzler]

+1!� Absolutely out of this world!� How you guys got all this done so fast is mind blowing.� Integrated into SI too, not just an export plugin.� This is truly ground breaking!

On 3/14/2013 10:06 PM, Emilio Hernandez wrote:

Let me tell you that I just put my hands on this baby and wow!!!�� This is going to rock the rendering world.� And for Softimage!!!!�

Awsome guys congratulations on this one.� My quadro 3000 finally is awake!!!

2013/3/14 Sylvain Lebeau <s...@shedmtl.com>

killer!!!!
congrats to you and team Nicolas!!

sly

Sylvain Lebeau // SHED
V-P/Visual effects supervisor

1410, RUE STANLEY, 11E �TAGE MONTR�AL (QU�BEC) H3A 1P8

T 514 849-1555 F 514 849-5025

WWW.SHEDMTL.COM <http://WWW.SHEDMTL.COM>

On 3/14/2013 10:35 PM, Nicolas Burtnyk wrote:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

�

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

�

�

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

�

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX� elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.� We're kind of obsessed with speed :)

�

speed is great of course � but IMO it�s not the most important factor.

�

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely �it has to be rendered in X amount of time� � clients couldn�t care less. It has to be good enough first and rendered in time for delivery.

�

it�s been a long time I�m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff � but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering � that can be very precisely tweaked by the user.

�

It�s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require � so in turn you have to bite the bullet and go back to good old offline rendering � and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features � not just GI / AO / softshadows / softreflections � but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair � and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

�

Mental ray never became the standard it was because of speed � but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it�s features are getting long in the tooth as well, which opens the door wide open for others � but it remains a reference for what a renderer should at least aspire to.

�

just some thoughts and hints of what matters to me when considering a new renderer.

--

[Mirko Jankovic]

hey I haven't really seen if region rendering is supported as well or only preview window? just wondering

On Fri, Mar 15, 2013 at 2:17 PM, Len Krenzler <l...@creativecontrol.ca> wrote:

+1!  Absolutely out of this world!  How you guys got all this done so fast is mind blowing.  Integrated into SI too, not just an export plugin.  This is truly ground breaking!

On 3/14/2013 10:06 PM, Emilio Hernandez wrote:

Let me tell you that I just put my hands on this baby and wow!!!   This is going to rock the rendering world.  And for Softimage!!!! 

Awsome guys congratulations on this one.  My quadro 3000 finally is awake!!!

2013/3/14 Sylvain Lebeau <s...@shedmtl.com>

killer!!!!
congrats to you and team Nicolas!!

sly

Sylvain Lebeau // SHED
V-P/Visual effects supervisor

1410, RUE STANLEY, 11E ÉTAGE MONTRÉAL (QUÉBEC) H3A 1P8

T 514 849-1555 F 514 849-5025

WWW.SHEDMTL.COM <http://WWW.SHEDMTL.COM>

On 3/14/2013 10:35 PM, Nicolas Burtnyk wrote:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

 

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

 

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

 

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX  elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.  We're kind of obsessed with speed :)

 

speed is great of course – but IMO it’s not the most important factor.

 

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely “it has to be rendered in X amount of time” – clients couldn’t care less. It has to be good enough first and rendered in time for delivery.

 

it’s been a long time I’m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff – but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering – that can be very precisely tweaked by the user.

 

It’s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require – so in turn you have to bite the bullet and go back to good old offline rendering – and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features – not just GI / AO / softshadows / softreflections – but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair – and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

 

Mental ray never became the standard it was because of speed – but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it’s features are getting long in the tooth as well, which opens the door wide open for others – but it remains a reference for what a renderer should at least aspire to.

 

just some thoughts and hints of what matters to me when considering a new renderer.

--

[Emilio Hernandez]

Everything is supported Mirko!  It is like having the old and crumpy MR reborn with power, speed and awsome result.  Integration with Softimage is seamless.

2013/3/15 Mirko Jankovic <mirkoj....@gmail.com>

--

[Mirko Jankovic]

uuuuu soo nice! now just to wait for OpenCL version whenever it comes.. I moved away from nvidia completely :)

[Len Krenzler]

You might want to move back just for this...just sayin'...

On 3/15/2013 7:35 AM, Mirko Jankovic wrote:

uuuuu soo nice! now just to wait for OpenCL version whenever it comes.. I moved away from nvidia completely :)

On Fri, Mar 15, 2013 at 2:33 PM, Emilio Hernandez <emi...@e-roja.com> wrote:

Everything is supported Mirko!� It is like having the old and crumpy MR reborn with power, speed and awsome result.� Integration with Softimage is seamless.

2013/3/15 Mirko Jankovic <mirkoj....@gmail.com>

hey I haven't really seen if region rendering is supported as well or only preview window? just wondering

On Fri, Mar 15, 2013 at 2:17 PM, Len Krenzler <l...@creativecontrol.ca> wrote:

+1!� Absolutely out of this world!� How you guys got all this done so fast is mind blowing.� Integrated into SI too, not just an export plugin.� This is truly ground breaking!

On 3/14/2013 10:06 PM, Emilio Hernandez wrote:

Let me tell you that I just put my hands on this baby and wow!!!�� This is going to rock the rendering world.� And for Softimage!!!!�

Awsome guys congratulations on this one.� My quadro 3000 finally is awake!!!

2013/3/14 Sylvain Lebeau <s...@shedmtl.com>

killer!!!!
congrats to you and team Nicolas!!

sly

Sylvain Lebeau // SHED
V-P/Visual effects supervisor

1410, RUE STANLEY, 11E �TAGE MONTR�AL (QU�BEC) H3A 1P8

T 514 849-1555 F 514 849-5025

WWW.SHEDMTL.COM <http://WWW.SHEDMTL.COM>

On 3/14/2013 10:35 PM, Nicolas Burtnyk wrote:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

�

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

�

�

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

�

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX� elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.� We're kind of obsessed with speed :)

�

speed is great of course � but IMO it�s not the most important factor.

�

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely �it has to be rendered in X amount of time� � clients couldn�t care less. It has to be good enough first and rendered in time for delivery.

�

it�s been a long time I�m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff � but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering � that can be very precisely tweaked by the user.

�

It�s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require � so in turn you have to bite the bullet and go back to good old offline rendering � and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features � not just GI / AO / softshadows / softreflections � but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair � and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

�

Mental ray never became the standard it was because of speed � but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it�s features are getting long in the tooth as well, which opens the door wide open for others � but it remains a reference for what a renderer should at least aspire to.

�

just some thoughts and hints of what matters to me when considering a new renderer.

--

-- 
_________________________________________________

Len Krenzler - Creative Control Media Productions

Phone: 780.463.3126

www.creativecontrol.ca - l...@creativecontrol.ca

--

[Mirko Jankovic]

well honestly... I'm on gaming cards because pro cards really are not justified with price in my case, and with gaming line ati right now is twice the speed of nvidia really... so just for rendering t o sacrifice all viewport performance.. I'm not sure that is something I would be willing to do :) not sure how mixing cards on same board would work with different drivers and everything to have one nvidis just for rendering.. anyway that is all different story and not really relevant in this case. in any case this is so refreshing

On Fri, Mar 15, 2013 at 2:39 PM, Len Krenzler <l...@creativecontrol.ca> wrote:

You might want to move back just for this...just sayin'...

On 3/15/2013 7:35 AM, Mirko Jankovic wrote:

uuuuu soo nice! now just to wait for OpenCL version whenever it comes.. I moved away from nvidia completely :)

On Fri, Mar 15, 2013 at 2:33 PM, Emilio Hernandez <emi...@e-roja.com> wrote:

Everything is supported Mirko!  It is like having the old and crumpy MR reborn with power, speed and awsome result.  Integration with Softimage is seamless.

2013/3/15 Mirko Jankovic <mirkoj....@gmail.com>

hey I haven't really seen if region rendering is supported as well or only preview window? just wondering

On Fri, Mar 15, 2013 at 2:17 PM, Len Krenzler <l...@creativecontrol.ca> wrote:

+1!  Absolutely out of this world!  How you guys got all this done so fast is mind blowing.  Integrated into SI too, not just an export plugin.  This is truly ground breaking!

On 3/14/2013 10:06 PM, Emilio Hernandez wrote:

Let me tell you that I just put my hands on this baby and wow!!!   This is going to rock the rendering world.  And for Softimage!!!! 

Awsome guys congratulations on this one.  My quadro 3000 finally is awake!!!

2013/3/14 Sylvain Lebeau <s...@shedmtl.com>

killer!!!!
congrats to you and team Nicolas!!

sly

Sylvain Lebeau // SHED
V-P/Visual effects supervisor

1410, RUE STANLEY, 11E ÉTAGE MONTRÉAL (QUÉBEC) H3A 1P8

T 514 849-1555 F 514 849-5025

WWW.SHEDMTL.COM <http://WWW.SHEDMTL.COM>

On 3/14/2013 10:35 PM, Nicolas Burtnyk wrote:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

 

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

 

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

 

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX  elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.  We're kind of obsessed with speed :)

 

speed is great of course – but IMO it’s not the most important factor.

 

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely “it has to be rendered in X amount of time” – clients couldn’t care less. It has to be good enough first and rendered in time for delivery.

 

it’s been a long time I’m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff – but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering – that can be very precisely tweaked by the user.

 

It’s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require – so in turn you have to bite the bullet and go back to good old offline rendering – and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features – not just GI / AO / softshadows / softreflections – but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair – and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

 

Mental ray never became the standard it was because of speed – but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it’s features are getting long in the tooth as well, which opens the door wide open for others – but it remains a reference for what a renderer should at least aspire to.

 

just some thoughts and hints of what matters to me when considering a new renderer.

--

-- 
_________________________________________________

Len Krenzler - Creative Control Media Productions

Phone: 780.463.3126

www.creativecontrol.ca - l...@creativecontrol.ca

--

[Tim Leydecker]

You may want to google the GTX Titan (GK110) for speed comparisons
and the older GTX 680/670 (GK104) for price/performance comparisons, too.

I will likely switch from a Quadro to one of the above cards even
thought the newest Quadro cards have a more competitive price
than one has come to expect from nVidia/PNY in the past.

The reasons for the switch in my upgrade plan is that I want to play more
(including actual gaming) with tools like Marmoset, dDo, nDO and zBrush.

Cheers,

tim

On 15.03.2013 14:51, Mirko Jankovic wrote:
> well honestly... I'm on gaming cards because pro cards really are not justified with price in my case, and with gaming line ati right now is twice the speed of nvidia really... so
> just for rendering t o sacrifice all viewport performance.. I'm not sure that is something I would be willing to do :) not sure how mixing cards on same board would work
> with different drivers and everything to have one nvidis just for rendering.. anyway that is all different story and not really relevant in this case. in any case this is so refreshing
>
>
> On Fri, Mar 15, 2013 at 2:39 PM, Len Krenzler <l...@creativecontrol.ca <mailto:l...@creativecontrol.ca>> wrote:
>
> You might want to move back just for this...just sayin'...
>
>
> On 3/15/2013 7:35 AM, Mirko Jankovic wrote:
>> uuuuu soo nice! now just to wait for OpenCL version whenever it comes.. I moved away from nvidia completely :)
>>
>>
>> On Fri, Mar 15, 2013 at 2:33 PM, Emilio Hernandez <emi...@e-roja.com <mailto:emi...@e-roja.com>> wrote:
>>
>> Everything is supported Mirko! It is like having the old and crumpy MR reborn with power, speed and awsome result. Integration with Softimage is seamless.
>>
>>

>> 2013/3/15 Mirko Jankovic <mirkoj....@gmail.com <mailto:mirkoj....@gmail.com>>

>>
>> hey I haven't really seen if region rendering is supported as well or only preview window? just wondering
>>
>>
>> On Fri, Mar 15, 2013 at 2:17 PM, Len Krenzler <l...@creativecontrol.ca <mailto:l...@creativecontrol.ca>> wrote:
>>
>> +1! Absolutely out of this world! How you guys got all this done so fast is mind blowing. Integrated into SI too, not just an export plugin. This is truly
>> ground breaking!
>>
>>
>> On 3/14/2013 10:06 PM, Emilio Hernandez wrote:
>>> Let me tell you that I just put my hands on this baby and wow!!! This is going to rock the rendering world. And for Softimage!!!!
>>>
>>> Awsome guys congratulations on this one. My quadro 3000 finally is awake!!!
>>>
>>>

>>> 2013/3/14 Sylvain Lebeau <s...@shedmtl.com <mailto:s...@shedmtl.com>>

>>>
>>> killer!!!!
>>> congrats to you and team Nicolas!!
>>>
>>> sly
>>>

>>> *Sylvain Lebeau // SHED**
>>> *V-P/Visual effects supervisor
>>> 1410, RUE STANLEY, 11E �TAGE MONTR�AL (QU�BEC) H3A 1P8
>>> T 514 849-1555 F 514 849-5025WWW.SHEDMTL.COM <http://www.shedmtl.com/><<http://www.shedmtl.com/>http://WWW.SHEDMTL.COM>

>>> On 3/14/2013 10:35 PM, Nicolas Burtnyk wrote:
>>>> Hey guys,
>>>>
>>>> I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.
>>>>
>>>> http://youtu.be/fjCguRdSlV0
>>>>
>>>> -Nicolas
>>>>
>>>>
>>>>
>>>> On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be <mailto:pet...@skynet.be>> wrote:
>>>>
>>>> you are right of course, as always.
>>>> what is really needed is a fine balance between quality and speed,
>>>> at a pricepoint that is affordable yet high enough to sustain development,
>>>> and available before my retirement.

>>>> *From:* Andy Moorer <mailto:andym...@gmail.com>
>>>> *Sent:* Thursday, March 14, 2013 9:02 PM
>>>> *To:* soft...@listproc.autodesk.com <mailto:soft...@listproc.autodesk.com>
>>>> *Subject:* Re: Announcing Redshift - Biased GPU Renderer

>>>> Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time
>>>> than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow
>>>> time to make changes and get them rendered... A major limitation to working with rendered VFX elements versus composite effects which can often be
>>>> altered in near realtime.
>>>>
>>>> Sent from my iPad
>>>>
>>>> On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be <mailto:pet...@skynet.be>> wrote:
>>>>
>>>>> > Please also bear in mind that we're still just in alpha and constantly improving performance. We're kind of obsessed with speed :)

>>>>> speed is great of course � but IMO it�s not the most important factor.
>>>>> over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely �it
>>>>> has to be rendered in X amount of time� � clients couldn�t care less. It has to be good enough first and rendered in time for delivery.
>>>>> it�s been a long time I�m looking forward for a viewport/GPU mental ray replacement in softimage.
>>>>> Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff � but more importantly, it should have the bells and
>>>>> whistles of a modern raytracer, and deliver production quality rendering � that can be very precisely tweaked by the user.
>>>>> It�s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling
>>>>> problem or no ability to finetune this or that effect or simply lack of a feature you really require � so in turn you have to bite the bullet and go
>>>>> back to good old offline rendering � and the corresponding rendertimes will be twice as frustrating.
>>>>> Very extensive support for lighting features � not just GI / AO / softshadows / softreflections � but also SSS, raytraced refractions, motion blur,
>>>>> volumetrics, ICE support, instancing, hair � and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.
>>>>> Mental ray never became the standard it was because of speed � but because of what one can achieve with it. (and then you have to turn off a few things

>>>>> left and right for final renders in order to make rendertimes acceptable)

>>>>> Obviously in this day and age it�s features are getting long in the tooth as well, which opens the door wide open for others � but it remains a

>>>>> reference for what a renderer should at least aspire to.
>>>>> just some thoughts and hints of what matters to me when considering a new renderer.
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>>
>>
>>

>> --
>> _________________________________________________
>>
>> Len Krenzler - Creative Control Media Productions
>>

>> Phone:780.463.3126 <tel:780.463.3126>
>>
>> www.creativecontrol.ca <http://www.creativecontrol.ca> -l...@creativecontrol.ca <mailto:l...@creativecontrol.ca>
>>
>>
>>
>>
>>
>> --

>>
>>
>
>
> --
> _________________________________________________
>
> Len Krenzler - Creative Control Media Productions
>

> Phone:780.463.3126 <tel:780.463.3126>
>
> www.creativecontrol.ca <http://www.creativecontrol.ca> -l...@creativecontrol.ca <mailto:l...@creativecontrol.ca>
>
>

[Emilio Hernandez]

Well Mirko as Len said.  You might just reconsider going back to Nvidia.  CUDA is coming strong on a lot of apps.  And getting first than ATI.  You can buy a GTX 470 for 200 bucks.

2013/3/15 Mirko Jankovic <mirkoj....@gmail.com>

--

[Mirko Jankovic]

actualy I already have an 580 in another comp so that itself is not problem :)

[Tim Crowson]

I've been really impressed with the performance and integration so far. I still need to throw some heavy scenes at it thow. But considering what it can do on a single card, I can't wait to see how it will run once multiple cards are supported.

Either way, this is already a win for the Softimage community. Big thanks to Nicolas and his team!

-Tim

On 3/15/2013 9:32 AM, Mirko Jankovic wrote:

actualy I already have an 580 in another comp so that itself is not problem :)

On Fri, Mar 15, 2013 at 3:29 PM, Emilio Hernandez <emi...@e-roja.com> wrote:

Well Mirko as Len said.� You might just reconsider going back to Nvidia.� CUDA is coming strong on a lot of apps.� And getting first than ATI.� You can buy a GTX 470 for 200 bucks.

2013/3/15 Mirko Jankovic <mirkoj....@gmail.com>

well honestly... I'm on gaming cards�because�pro cards really are not justified with price in my case, and with gaming line ati right now is twice the speed of nvidia really... so just for rendering t o sacrifice all viewport performance.. I'm not sure that is something I would be willing to do :) not sure how mixing cards on same board would work with�different�drivers and everything to have one nvidis just for rendering.. anyway that is all different story and not really relevant in this case. in any case this is so refreshing

On Fri, Mar 15, 2013 at 2:39 PM, Len Krenzler <l...@creativecontrol.ca> wrote:

You might want to move back just for this...just sayin'...

On 3/15/2013 7:35 AM, Mirko Jankovic wrote:

uuuuu soo nice! now just to wait for OpenCL version whenever it comes.. I moved away from nvidia completely :)

On Fri, Mar 15, 2013 at 2:33 PM, Emilio Hernandez <emi...@e-roja.com> wrote:

Everything is supported Mirko!� It is like having the old and crumpy MR reborn with power, speed and awsome result.� Integration with Softimage is seamless.

2013/3/15 Mirko Jankovic <mirkoj....@gmail.com>

hey I haven't really seen if region rendering is supported as well or only preview window? just wondering

On Fri, Mar 15, 2013 at 2:17 PM, Len Krenzler <l...@creativecontrol.ca> wrote:

+1!� Absolutely out of this world!� How you guys got all this done so fast is mind blowing.� Integrated into SI too, not just an export plugin.� This is truly ground breaking!

On 3/14/2013 10:06 PM, Emilio Hernandez wrote:

Let me tell you that I just put my hands on this baby and wow!!!�� This is going to rock the rendering world.� And for Softimage!!!!�

Awsome guys congratulations on this one.� My quadro 3000 finally is awake!!!

2013/3/14 Sylvain Lebeau <s...@shedmtl.com>

killer!!!!
congrats to you and team Nicolas!!

sly

Sylvain Lebeau // SHED
V-P/Visual effects supervisor

1410, RUE STANLEY, 11E �TAGE MONTR�AL (QU�BEC) H3A 1P8

T 514 849-1555 F 514 849-5025

WWW.SHEDMTL.COM <http://WWW.SHEDMTL.COM>

On 3/14/2013 10:35 PM, Nicolas Burtnyk wrote:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

�

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

�

�

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

�

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX� elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.� We're kind of obsessed with speed :)

�

speed is great of course � but IMO it�s not the most important factor.

�

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely �it has to be rendered in X amount of time� � clients couldn�t care less. It has to be good enough first and rendered in time for delivery.

�

it�s been a long time I�m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff � but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering � that can be very precisely tweaked by the user.

�

It�s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require � so in turn you have to bite the bullet and go back to good old offline rendering � and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features � not just GI / AO / softshadows / softreflections � but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair � and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

�

Mental ray never became the standard it was because of speed � but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it�s features are getting long in the tooth as well, which opens the door wide open for others � but it remains a reference for what a renderer should at least aspire to.

�

just some thoughts and hints of what matters to me when considering a new renderer.

--

-- 
_________________________________________________

Len Krenzler - Creative Control Media Productions

Phone: 780.463.3126

www.creativecontrol.ca - l...@creativecontrol.ca

--

-- 
_________________________________________________

Len Krenzler - Creative Control Media Productions

Phone: 780.463.3126

www.creativecontrol.ca - l...@creativecontrol.ca

--

--

�

[Len Krenzler]

I only have a GTX470 and it flies even with that!� I'm testing a scene right now with 4.5 mil polys and a 12k HDR lighting texture as well as other large textures and no problem.�

[Guillaume Laferriere]

I wonder how this looks with a render region with alpha blending turned on.
The renderer would need to output RGBA and support the render region. Does it?

GL
From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Len Krenzler
Sent: Friday, March 15, 2013 11:13 AM

To: soft...@listproc.autodesk.com
Subject: Re: Announcing Redshift - Biased GPU Renderer

I only have a GTX470 and it flies even with that! I'm testing a scene right now with 4.5 mil polys and a 12k HDR lighting texture as well as other large textures and no problem.

On 3/15/2013 8:58 AM, Tim Crowson wrote:
I've been really impressed with the performance and integration so far. I still need to throw some heavy scenes at it thow. But considering what it can do on a single card, I can't wait to see how it will run once multiple cards are supported.

Either way, this is already a win for the Softimage community. Big thanks to Nicolas and his team!

-Tim
On 3/15/2013 9:32 AM, Mirko Jankovic wrote:
actualy I already have an 580 in another comp so that itself is not problem :)

On Fri, Mar 15, 2013 at 3:29 PM, Emilio Hernandez <emi...@e-roja.com<mailto:emi...@e-roja.com>> wrote:
Well Mirko as Len said. You might just reconsider going back to Nvidia. CUDA is coming strong on a lot of apps. And getting first than ATI. You can buy a GTX 470 for 200 bucks.

2013/3/15 Mirko Jankovic <mirkoj....@gmail.com<mailto:mirkoj....@gmail.com>>
well honestly... I'm on gaming cards because pro cards really are not justified with price in my case, and with gaming line ati right now is twice the speed of nvidia really... so just for rendering t o sacrifice all viewport performance.. I'm not sure that is something I would be willing to do :) not sure how mixing cards on same board would work with different drivers and everything to have one nvidis just for rendering.. anyway that is all different story and not really relevant in this case. in any case this is so refreshing

On Fri, Mar 15, 2013 at 2:39 PM, Len Krenzler <l...@creativecontrol.ca<mailto:l...@creativecontrol.ca>> wrote:
You might want to move back just for this...just sayin'...


On 3/15/2013 7:35 AM, Mirko Jankovic wrote:
uuuuu soo nice! now just to wait for OpenCL version whenever it comes.. I moved away from nvidia completely :)

On Fri, Mar 15, 2013 at 2:33 PM, Emilio Hernandez <emi...@e-roja.com<mailto:emi...@e-roja.com>> wrote:
Everything is supported Mirko! It is like having the old and crumpy MR reborn with power, speed and awsome result. Integration with Softimage is seamless.

2013/3/15 Mirko Jankovic <mirkoj....@gmail.com<mailto:mirkoj....@gmail.com>>

hey I haven't really seen if region rendering is supported as well or only preview window? just wondering

On Fri, Mar 15, 2013 at 2:17 PM, Len Krenzler <l...@creativecontrol.ca<mailto:l...@creativecontrol.ca>> wrote:
+1! Absolutely out of this world! How you guys got all this done so fast is mind blowing. Integrated into SI too, not just an export plugin. This is truly ground breaking!

On 3/14/2013 10:06 PM, Emilio Hernandez wrote:

Let me tell you that I just put my hands on this baby and wow!!! This is going to rock the rendering world. And for Softimage!!!!
Awsome guys congratulations on this one. My quadro 3000 finally is awake!!!

2013/3/14 Sylvain Lebeau <s...@shedmtl.com<mailto:s...@shedmtl.com>>

killer!!!!
congrats to you and team Nicolas!!

sly
Sylvain Lebeau // SHED
V-P/Visual effects supervisor

1410, RUE STANLEY, 11E ÉTAGE MONTRÉAL (QUÉBEC) H3A 1P8
T 514 849-1555 F 514 849-5025 WWW.SHEDMTL.COM<http://www.shedmtl.com/> <http://WWW.SHEDMTL.COM>

On 3/14/2013 10:35 PM, Nicolas Burtnyk wrote:
Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be<mailto:pet...@skynet.be>> wrote:
you are right of course, as always.

what is really needed is a fine balance between quality and speed,
at a pricepoint that is affordable yet high enough to sustain development,
and available before my retirement.

From: Andy Moorer<mailto:andym...@gmail.com>
Sent: Thursday, March 14, 2013 9:02 PM
To: soft...@listproc.autodesk.com<mailto:soft...@listproc.autodesk.com>
Subject: Re: Announcing Redshift - Biased GPU Renderer

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be<mailto:pet...@skynet.be>> wrote:
> Please also bear in mind that we're still just in alpha and constantly improving performance. We're kind of obsessed with speed :)

speed is great of course - but IMO it's not the most important factor.

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely "it has to be rendered in X amount of time" - clients couldn't care less. It has to be good enough first and rendered in time for delivery.

it's been a long time I'm looking forward for a viewport/GPU mental ray replacement in softimage.
Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff - but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering - that can be very precisely tweaked by the user.

It's very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require - so in turn you have to bite the bullet and go back to good old offline rendering - and the corresponding rendertimes will be twice as frustrating.
Very extensive support for lighting features - not just GI / AO / softshadows / softreflections - but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair - and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

Mental ray never became the standard it was because of speed - but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)
Obviously in this day and age it's features are getting long in the tooth as well, which opens the door wide open for others - but it remains a reference for what a renderer should at least aspire to.

just some thoughts and hints of what matters to me when considering a new renderer.





--

--

_________________________________________________



Len Krenzler - Creative Control Media Productions

Phone: 780.463.3126<tel:780.463.3126>

www.creativecontrol.ca<http://www.creativecontrol.ca> - l...@creativecontrol.ca<mailto:l...@creativecontrol.ca>



--

--

_________________________________________________



Len Krenzler - Creative Control Media Productions

Phone: 780.463.3126<tel:780.463.3126>

www.creativecontrol.ca<http://www.creativecontrol.ca> - l...@creativecontrol.ca<mailto:l...@creativecontrol.ca>



--


--

--

_________________________________________________



Len Krenzler - Creative Control Media Productions



Phone: 780.463.3126

www.creativecontrol.ca<http://www.creativecontrol.ca> - l...@creativecontrol.ca<mailto:l...@creativecontrol.ca>

[Nour Almasri]

This is a great achievement Nicolas .

It's really surprise how many amazing tech are available in those days ( Fabric Engine - Arnold  - Alembic )  and who do it is not the biggest (resources, money) company , this renderer look promising and will be on of them , keep up the great work guys .

[Len Krenzler]

Yes it does.

> 1410, RUE STANLEY, 11E �TAGE MONTR�AL (QU�BEC) H3A 1P8

www.creativecontrol.ca - l...@creativecontrol.ca

[olivier jeannel]

I Just watched the video this evening (not at my office these days
unfortunatly) I'm completly blown away by what I saw ! I'm smilling
stupidly atm....
I really hope the particle instance integration won't take long so that
I can throw something to render...
Nicholas that's super great stuff.

Would you share some demo/tutorial scenes for us to have a quickstart ?
(never rendered with that kind of engine before)

Awesome anyway !


Le 15/03/2013 18:12, Len Krenzler a �crit :

[Christian Gotzinger]

Excuse the language, but: Holy shit! Mighty impressive stuff!

[Mirko Jankovic]

Just thinking.. you really have to wonder why AD as huge company with resources that are probably hard to imagine by any of us never got to make ANYTHING nearly great as anything like what we see from 3rd party guys around. 

If you think about it like 99% of progress is NOT made by big companies. Really need to think in which direction money flow could start to change.. away from AD-like and more towards guys like Redshift team :)

All the best!

[Len Krenzler]

No kidding!  I can't imagine why Nvidia/Arc/MR whatever they're called now couldn't have done this years ago.  Between them and AD they can't even get they're basic features working.

Money should be directed to these 3rd party guys as much as possible.  AD must have called me about 10 times to renew my sub.  This time I said no, I think I'll spend that on the guys getting results.

If you haven't tried this yet, do!

-- 
_________________________________________________

Len Krenzler - Creative Control Media Productions

Phone: 780.463.3126

www.creativecontrol.ca - l...@creativecontrol.ca

[Mirko Jankovic]

just wondering.. if everything is on GPU in theory it should work fine even with some slower older CPU?

Got some older comp laying around and both PCI slots in comp are filled, maybe could use that one for GPU rendering station for testing :)

any thoughts?

[Sven Constable]

It is not that simple. I think to make a renderer that looks very promising is one thing. Establish it in the market is the hard part. For example lets take Arnold… it took them over ten years to make it something we consider a product and today it's (officially) still in beta. There were other renderers (I don't remember right now. Brazil was one of them). Great renderer, faster than some others. Now its abandoned. This was actually a renderer used some years, but there were many others that didn't survive their first couple of years while in developing.

 

Redshift looks indeed very nice and promising. I hope it will make its way into the market.

 

sven

 

 

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Mirko Jankovic
Sent: Friday, March 15, 2013 20:54
To: soft...@listproc.autodesk.com
Subject: Re: Announcing Redshift - Biased GPU Renderer

 

Just thinking.. you really have to wonder why AD as huge company with resources that are probably hard to imagine by any of us never got to make ANYTHING nearly great as anything like what we see from 3rd party guys around. 

[Steven Caron]

arnold is NOT in beta... but your point about market success is made. 

[Vladimir Jankijevic]

I have to back up Steven. Arnold is NOT in beta. I had more to say about this subject but it's not the place for that. 

I'm really curious what the Redshift guys are able to deliver for a production environment. I'll keep an eye on this for sure!

--
---------------------------------------
Vladimir Jankijevic
Technical Direction

Elefant Studios AG
Lessingstrasse 15
CH-8002 Zürich

+41 44 500 48 20

www.elefantstudios.ch
---------------------------------------

[pet...@skynet.be]

well, I had to join a betalist to get info and test it, downloadables were called beta, and so was the quote (afaik – it wasn’t adressed to me) – I was under the impression that it is still beta, because that’s what it said on the surface.

Smoke and mirrors? Semantics?

Perhaps – but not to the producer who had to make the decision on purchasing.

 

 

 

From: Vladimir Jankijevic

Sent: Saturday, March 16, 2013 12:03 AM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

 

[Stephen Blair]

SItoA beta and Mtoa beta, not Arnold beta

[Sven Constable]

ok, I stand corrected. Thanks for clearing that up.

 

From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Vladimir Jankijevic
Sent: Saturday, March 16, 2013 0:04
To: soft...@listproc.autodesk.com
Subject: Re: Announcing Redshift - Biased GPU Renderer

 

I have to back up Steven. Arnold is NOT in beta. I had more to say about this subject but it's not the place for that. 

[pet...@skynet.be]

fair enough – I was indeed referring to SItoA and MtoA , not Arnold standalone.

[Steven Caron]

still, sitoa isn't beta anymore. peter, that was years ago when you started evaluation, right?

it needs to be known that the reason it's not distributed widely isn't because arnold or sitoa is beta software. it's because they can't support everyone that wants access at this time. so no reason to make an announcement about lifting the beta tag. the sitoa list isn't called the beta list by its users or moderators anymore.

sorry for derailing nic's thread.

*written with my thumbs

[Nicolas Burtnyk]

Hi Peter,

First of all, let me apologize for taking forever to respond.  We've had a pretty crazy last couple of days with the alpha launch.

You're absolutely right that speed is worth very little or even nothing if you can't actually get the image you or the client wants out of the damned thing, whether it's due to missing features, stability issues, limitations on content complexity, lack of flexibility or ease of use.  We're very sensitive to that and while I can't claim that we're there yet, we do plan to have all the bells and whistles, stability, flexibility and ergonomics to make Redshift a legitimate choice for production rendering.

That being said, speed can be important for a number of reasons.  A big one is iteration times.  Everything else being equal faster rendering results in better images because you have more opportunity to iterate, experiment, tweak and generally be creative.

Another one is cost.  This will vary a lot for different types of users, but if you suddenly don't need a render farm because your workstation renders just as fast, you've saved money.  Or if you need a farm with only 100 nodes instead of 1000, you've saved some more money.

I should point out that Redshift doesn't just do basic raytracing and GI but actually already supports many of the features you mentioned.  We do point-based SSS, motion blur (not deformation blur yet, but we're working on that right now), instancing and refractions.  For a 3rd party renderer, I would say that our support for the native Softimage shaders is probably about on par or possibly better than the others.

And we're not done yet!  Proper ICE support is a big one, as is proper support for AOV/render channels.  Hair is another.  These are all in the plan and have already had some significant thought (and in some cases initial work) put into them.

-Nicolas

On Thu, Mar 14, 2013 at 1:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.  We're kind of obsessed with speed :)

 

speed is great of course – but IMO it’s not the most important factor.

 

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely “it has to be rendered in X amount of time” – clients couldn’t care less. It has to be good enough first and rendered in time for delivery.

 

it’s been a long time I’m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff – but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering – that can be very precisely tweaked by the user.

 

It’s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require – so in turn you have to bite the bullet and go back to good old offline rendering – and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features – not just GI / AO / softshadows / softreflections – but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair – and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

 

Mental ray never became the standard it was because of speed – but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it’s features are getting long in the tooth as well, which opens the door wide open for others – but it remains a reference for what a renderer should at least aspire to.

[Nicolas Burtnyk]

Hi Mirko,

Redshift does use the CPU for a couple things here and there so the CPU is not irrelevant to the performance, but it's not a big contributor.

For example, the RT hierarchy construction (construction of the acceleration structure for raytracing) is done on the CPU as is texture conversion to our optimized tile format.  Also the screen-space adaptive tessellation and some rebalancing of tree data structures for the point-based techniques run on the CPU.

All in all though the GPU spec is what is really going to make the difference for Redshift, assuming we're not talking about a 486 or something :)

So yeah I encourage you to dust off that old PC, pop in the GTX580 and take it for a spin.  I think you'll be pleasantly surprised.

-Nicolas

[pet...@skynet.be]

no, that was last summer.

 

if anything needs to be known about a software, an official website with actual information would be a good starting point.

 

anyways – lets not spoil Redshift’s thread with talk about other software.

[pet...@skynet.be]

thanks Nicolas – sounds very good.

the images as well as the video look very promising – my interest is certainly aroused .

 

From: Nicolas Burtnyk

Sent: Sunday, March 17, 2013 7:33 AM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

 

[Stefan Kubicek]

Hi Nicolas,

I'm curious in how far GPU memory impacts render time. To put it differently: Assuming the amount of cores is what makes the biggest difference in render time, what's the expected speed differences comparing a graphics card with 1gb to one equipped with 2gb or more?

Cheers,

Stefan

--
-------------------------------------------
Stefan Kubicek
-------------------------------------------
keyvis digital imagery
Alfred Feierfeilstraï¿œe 3
A-2380 Perchtoldsdorf bei Wien
Phone: +43/699/12614231
www.keyvis.at ste...@keyvis.at
-- This email and its attachments are --
--confidential and for the recipient only--

[olivier jeannel]

To add to the subject, is there a Redshift benchmark with different
graphic cards ?
Will be fun to have renderfarm only filled with graphic cards :)
-"Where's your renderfarm ?"
-"It's the little box on floor..."



Le 18/03/2013 12:49, Stefan Kubicek a ï¿œcrit :

[Stefan Andersson]

That was pretty "neat"! :) I can't wait to see some more test!

regards
stefan

On Fri, Mar 15, 2013 at 3:35 AM, Nicolas Burtnyk <nic...@redshift3d.com> wrote:

Hey guys,

I'm going to respond to the last few messages regarding the importance of speed later, but in the meantime here is a video of some live rendering in Softimage.

http://youtu.be/fjCguRdSlV0

-Nicolas

On Thu, Mar 14, 2013 at 1:17 PM, <pet...@skynet.be> wrote:

you are right of course, as always.

 

what is really needed is a fine balance between quality and speed,

at a pricepoint that is affordable yet high enough to sustain development,

and available before my retirement.

 

From: Andy Moorer

Sent: Thursday, March 14, 2013 9:02 PM

To: soft...@listproc.autodesk.com

Subject: Re: Announcing Redshift - Biased GPU Renderer

Well said, but speed is still important, deadlines are tight and particularly in the iterative direction phase often re-rendering takes much more time than making a directed change. "Dailies" reflect this... A series of several directed tweaks to a shot can stretch over several days in part to allow time to make changes and get them rendered... A major limitation to working with rendered VFX  elements versus composite effects which can often be altered in near realtime.

Sent from my iPad

On Mar 14, 2013, at 4:21 AM, <pet...@skynet.be> wrote:

> Please also bear in mind that we're still just in alpha and constantly improving performance.  We're kind of obsessed with speed :)

 

speed is great of course – but IMO it’s not the most important factor.

 

over the years we have all been doing productions with rather long rendertimes, running into hours per frame and more. The bottom line was rarely “it has to be rendered in X amount of time” – clients couldn’t care less. It has to be good enough first and rendered in time for delivery.

 

it’s been a long time I’m looking forward for a viewport/GPU mental ray replacement in softimage.

Hopefully staying below 5 minutes for complex HD images and within 1 minute for more simple stuff – but more importantly, it should have the bells and whistles of a modern raytracer, and deliver production quality rendering – that can be very precisely tweaked by the user.

 

It’s very frustrating to get a promising image very fast, but not being able to make the image really final - some remaining artifacts, sampling problem or no ability to finetune this or that effect or simply lack of a feature you really require – so in turn you have to bite the bullet and go back to good old offline rendering – and the corresponding rendertimes will be twice as frustrating.

Very extensive support for lighting features – not just GI / AO / softshadows / softreflections – but also SSS, raytraced refractions, motion blur, volumetrics, ICE support, instancing, hair – and a good set of shaders and support for the rendertree and as many of the factory shaders as possible.

 

Mental ray never became the standard it was because of speed – but because of what one can achieve with it. (and then you have to turn off a few things left and right for final renders in order to make rendertimes acceptable)

Obviously in this day and age it’s features are getting long in the tooth as well, which opens the door wide open for others – but it remains a reference for what a renderer should at least aspire to.

 

just some thoughts and hints of what matters to me when considering a new renderer.

--
Stefan Andersson | Digital Janitor

blog | showreel | twitter | LinkedIn | cell: +46-73-6268850 | skype:sanders3d

[Mirko Jankovic]

testing it a bit and looks great!

amazing work guys, grats.

any ETA for production ready version?

also reall shame again that it is nvidia only for now. Ati was tested over and over and showing a lot better viewport results in Softimage than nvidia... having this support openCL would be great!

But everything  in it's time. Grats!

[Dan Yargici]

"Ati was tested over and over and showing a lot better viewport results in Softimage than nvidia... "

Really?  I don't remember anyone ever suggesting ATI was anything other than shit!

DAN

[Mirko Jankovic]

http://www.si-community.com/community/viewtopic.php?f=36&t=3526

latest one.

ati rigth now leaves nvidia in the dust

[Raffaele Fragapane]

These days if you hit the right combination of drivers and planet alignment they are OK.

Performance wise they have been ahead of nVIDIA for a while in number crunching, the main problem is the drivers are still a coin toss chance, and that OCL isn't anywhere as popular as CUDA.

With win7 or 8 and recent versions of Soft/Maya they can do well.

nVIDIA didn't help with the crippling of the 6xx for professional use, and pissing off Linus. They are still ahead by a slight margin, for now, but I wouldn't discount AMD wholesale anymore.

If the next generation is as disappointing as Kepler is, and AMD gets both Linux support AND decent (and properly OSS) drivers out, I'm moving time come for the next upgrade. For now I recently bought a 680 because it was kind of mandatory to not go insane with Mari and Mudbox, and because I like CUDA and I toy with it at home.

[Arvid Björn]

My beef with ATI last time I tried FirePro was that it had a hard time locking into 25fps playback in some apps, as if the refresh rate was locked to 30/60. Realtime playback in Softimage would stutter annoyingly IIRC. Plus it seemed to draw text slightly differently in some apps.

Nvidia just feels.. comfy.

[Mirko Jankovic]

well no idea about pro cards.. really never got financial justification to get one, quadro 4000 in old company didn;t really felt anything much better than gaming cards so...

but in gaming segment..

opengl scores in sinebench for example:

gtx 580: ~55

7970: ~90

to start with....

not to mention annoying issue with high segment rotating cube in viewport in SI.

7970 smooth at ~170 fps

with gtx580 bfore that.. to point out that the rest of comp is identical only switched card... for the first 30-50sec frame rate was stuck at something like 17 fps... and after that it kinda jump to ~70-80fps...

in any case with gaming cards ati vs nvidia there is no doubt. and if you are not using CUDA much then no need to even thing which way to go.

Now redshift is game changer heheh but I'm still hoping that OpenCL will be supported and I'm looking forward to test it out with two of 7970 in crossfire :)

btw I'm not much into programming waters but is it really OpenCL programming  that as I understood should work on ALL cards, is that much more complex than for CUDA which is limited to nvidia only? Wouldn't it be more logical to go with solution that is covering a lot more market than something limited to one manufacturer?

[Nicolas Burtnyk]

The answer is... it depends :)

If your scene is very large and doesn't fit in X GB of VRAM, then more VRAM will be a big performance win because you'll be going out of core less.  That being said, even for simpler scenes that easily fit in VRAM, more VRAM can improve performance.  Redshift can use excess VRAM to increase the size of its workloads which results in better utilzation of the GPU.

So more memory is good and more/faster cores is good too, but it's impossible to give you a X factor for the performance difference between a 1GB and 2GB VRAM card with equivalent cores.  It's just too scene dependent.

-Nicolas

          Alfred Feierfeilstraße 3

[Nicolas Burtnyk]

Hey guys,

Just wanted to share a couple of very short videos we made that show the stability of the GI in Redshift.

Unfortunately Youtube's compression kind of murdered the smoothness, but I assure you that any artifacts you see in these videos are from compression and not GI.

http://youtu.be/3c0tHYdd-fg

This video shows the dark side of a deforming gargoyle lit by physical sun & sky.

25 seconds per frame for 1280x720 on a Core i7 3.07Ghz, 12GB RAM w/ NVIDIA Geforce GTX 470.

http://youtu.be/ySCockShxTQ

This video shows the same gargoyle being lit strictly by light bouncing off the floor.  The setup is a white spot light shining onto the floor (off camera).  The red glow you see on the floor around the gargoyle is light that has bounced off the floor, then off the gargoyle.

1 minute per frame for 1280x720 on a Core i7 3.07Ghz, 6GB RAM w/ NVIDIA Geforce GTX 670.

We're still head-down fixing bugs and bringing new features online, but I plan to spend some time making more (and better) videos soon.

-Nicolas

On Tue, Mar 26, 2013 at 7:27 PM, Nicolas Burtnyk <nic...@redshift3d.com> wrote:

[Ahmidou.xsi]

Hi Nicolas, you could maube consider vimeo as a better option.

Cheers

[Raffaele Fragapane]

Benchmarking is more driver tuning than it's videocard performance, and if you want to look at number crunching you should look at the most recent gens.

The 680 has brought nVIDIA back up top for number crunching (forgetting the silver editions or gimmicks like the titan), and close enough to bang for buck best, but AMD's response to that still has to come.

Ironically, though, the 6xx gen is reported as a crippled, bad performer in DCC apps, although I can't say I noticed it myself. It sure as hell works admirably well in mudbox, mari, cuda work, and I've had no issues in maya or soft. I don't really benchmrak or obsess over numbers much though.

When this will obsolesce, I will considering AMD again, probably in a couple years.

For GPU rendering though, well, that's something you CAN bench reliably with the engine, and AMD might still win the FLOP per dollar run there, so it's not to be discounted.

Would be good to know what the redshift guys have to say about it themselves though if they can spare the thought and can actually disclose.

--
Our users will know fear and cower before our software! Ship it! Ship it and let them flee like the dogs they are!

[Jason S]

Yep, doesn't flicker one bit!

[Tim Leydecker]

The GTX Titan is not a gimmick but uses the successor to the chip series
used in the GTX 680, e.g. the GT(X) 6xx series uses the GK104, while
the GTX Titan uses the GK110. You can find the GK110 in the Tesla K20, too.

You could describe the GTX690 as a gimmick, as it uses two GK104 on one card
to maximize performance at the cost of higher powerconsumption, noise and heat.

The performance gain between a GTX680 and a GTX Titan is roughly 35%
and can be felt nicely when using it with higher screenresolutions like
1920x1200 or 2560x1440 and higher antialiasing in games.

That�s where the 6GB VRAM of the GTX Titan come in handy, too.

Cheers,

tim

[Mirko Jankovic]

On the other hand Titan is more expensive than 2 gtx680 if I'm not mistaken... and i bet that with two 680 in SLI, when multi GPU is supported you will have better performance than with 1 titan right?

[Tim Leydecker]

Personally, I�m hesistant to using two or more cards with SLI
because of micro stuttering: http://en.wikipedia.org/wiki/Micro_stuttering

If there would be a solution to that, I�d go with two GTX670 w/4GB VRAM,
as they are the same GK104�s with a 915MHz chipspeed instead of a 1006Mhz
chipspeed as in the reference design GTX680. That could save another 15-35%
percent of investment compared to two single chip GTX 680 cards or one GTX Titan.

Overclocked versions may use slightly different chip/shader speeds.

In any case, as much VRAM as available, as that always helps in many progamms
like Mudbox, Redshift and isn�t much of an added cost (comparing 2GB vs 4GB).

At a company I worked Mari 1.5.x behaved bitchy unless it was given a Quadro
or forced to ignore the actual card�s game heritage. But that may have been
solved with 2.0...

Cheers,

tim

>>> <mirkoj....@gmail.com>**wrote:

>>>> On Thu, Mar 21, 2013 at 10:55 AM, Arvid Bj�rn <arvid...@gmail.com

[Ben Davis]

That's exactly what I'm eager for, having multiple cards in (linked by sli or not) participating in the render. Huge bang for buck potential.

Ben

--

Benjamin Clifford Davis

www.moondog-animation.com 

office:   +33 9 50 04 76 15
mobile: +33 6 88 48 54 50

6 bis avenue des Iles
74000 Annecy
FRANCE

[Mirko Jankovic]

Actual SLI wouldn't need to be active at all. You would still be using single card for viewport and monitors. Other card would be used only iwth renderer.

So there shouldn't be any micro stuttering at all. I think that problem is evident only in gaming, as I'm experiencing it myself with two 7970s, but as said only in gaming when actually using crossfire. 

On Thu, Mar 21, 2013 at 10:55 AM, Arvid Björn <arvid...@gmail.com

[Ben Davis]

I don't think micro stuttering would be a terrible issue as far as GPU rendering goes, it's mostly a frustrating drawback as far as framerates being slightly crippled in gameplay, no?

--

Benjamin Clifford Davis

www.moondog-animation.com 

office:   +33 9 50 04 76 15
mobile: +33 6 88 48 54 50

6 bis avenue des Iles
74000 Annecy
FRANCE

On Thu, Mar 21, 2013 at 10:55 AM, Arvid Björn <arvid...@gmail.com

[Tim Leydecker]

I don锟絫 know how the setup of two or more cards would be best done
for GPU rendering purposes but I would at least try to enable SLI
to get the best framerate/redraw performance in general applications
and games to better justify the investment.

But maybe, if I don锟絫 have to bother about it and just see the GPU renderer
pick up all the available cards I锟絛 be just as happy with the increased
renderspeed and possibilities this gives.

In general I锟絤 most likely hesistant as I锟絭e been burnt by things like
mR unified sampling messing up framebuffers (in xsi2012/mR 3.9.x) or not
correctly supporting satellite rendering (in xsi2012/mR 3.9.x) and seing
those flaws eat up the initial benefit I had hoped for to some extend.

Anyone using VRay RT on Maya or mentalray锟絪 iRay here and able to supply info?

Cheers,

tim

On 27.03.2013 09:34, Ben Davis wrote:
> I don't think micro stuttering would be a terrible issue as far as GPU
> rendering goes, it's mostly a frustrating drawback as far as framerates
> being slightly crippled in gameplay, no?
>
> --
> Benjamin Clifford Davis
>
> www.moondog-animation.com
>
> office: +33 9 50 04 76 15
> mobile: +33 6 88 48 54 50
>
> 6 bis avenue des Iles
> 74000 Annecy
> FRANCE
>
>
> On Wed, Mar 27, 2013 at 9:26 AM, Tim Leydecker <baue...@gmx.de> wrote:
>

>> Personally, I锟絤 hesistant to using two or more cards with SLI
>> because of micro stuttering: http://en.wikipedia.org/wiki/**
>> Micro_stuttering <http://en.wikipedia.org/wiki/Micro_stuttering>
>>
>> If there would be a solution to that, I锟絛 go with two GTX670 w/4GB VRAM,
>> as they are the same GK104锟絪 with a 915MHz chipspeed instead of a 1006Mhz

>> chipspeed as in the reference design GTX680. That could save another 15-35%
>> percent of investment compared to two single chip GTX 680 cards or one GTX
>> Titan.
>>
>> Overclocked versions may use slightly different chip/shader speeds.
>>
>> In any case, as much VRAM as available, as that always helps in many
>> progamms

>> like Mudbox, Redshift and isn锟絫 much of an added cost (comparing 2GB vs

>> 4GB).
>>
>> At a company I worked Mari 1.5.x behaved bitchy unless it was given a
>> Quadro

>> or forced to ignore the actual card锟絪 game heritage. But that may have been

>> solved with 2.0...
>>
>> Cheers,
>>
>> tim
>>
>>
>>
>>
>> On 27.03.2013 08:59, Mirko Jankovic wrote:
>>
>>> On the other hand Titan is more expensive than 2 gtx680 if I'm not
>>> mistaken... and i bet that with two 680 in SLI, when multi GPU is
>>> supported
>>> you will have better performance than with 1 titan right?
>>>
>>>
>>> On Wed, Mar 27, 2013 at 8:55 AM, Tim Leydecker <baue...@gmx.de> wrote:
>>>
>>> The GTX Titan is not a gimmick but uses the successor to the chip series
>>>> used in the GTX 680, e.g. the GT(X) 6xx series uses the GK104, while
>>>> the GTX Titan uses the GK110. You can find the GK110 in the Tesla K20,
>>>> too.
>>>>
>>>> You could describe the GTX690 as a gimmick, as it uses two GK104 on one
>>>> card
>>>> to maximize performance at the cost of higher powerconsumption, noise and
>>>> heat.
>>>>
>>>> The performance gain between a GTX680 and a GTX Titan is roughly 35%
>>>> and can be felt nicely when using it with higher screenresolutions like
>>>> 1920x1200 or 2560x1440 and higher antialiasing in games.
>>>>

>>>> That锟絪 where the 6GB VRAM of the GTX Titan come in handy, too.

>>>>> <mirkoj....@gmail.com>****wrote:

>>>>>> On Thu, Mar 21, 2013 at 10:55 AM, Arvid Bj锟絩n <arvid...@gmail.com

[olivier jeannel]

There was a subject on Redshift Forum about having two grapphic cards.
It seems to be possible to keep a quadro for dispaly (as it is
significantly better at displaying), and have a Titan dedicated to
rendering only (in Redshift you select which card is rendering) as they
have a huge amount of cores and faster memory.
I think I've red somewhere that Titan has 2600 cores against 256 for the
Quadro 4000.
After chating with Nicolas the Titan could be around 4 time faster than
the Quadro4000 ...Which is huge :)

Le 27/03/2013 09:26, Tim Leydecker a �crit :

[Mirko Jankovic]

SLI and crossfire dio not affect viewport performance in any of 3d application. 

On Wed, Mar 27, 2013 at 11:12 AM, olivier jeannel <olivier...@noos.fr> wrote:

There was a subject on Redshift Forum about having two grapphic cards.
It seems to be possible to keep a quadro for dispaly (as it is significantly better at displaying), and have a Titan dedicated to rendering only (in Redshift you select which card is rendering) as they have a huge amount of cores and faster memory.
I think I've red somewhere that Titan has 2600 cores against 256 for the Quadro 4000.
After chating with Nicolas the Titan could be around 4 time faster than the Quadro4000 ...Which is huge :)

On Thu, Mar 21, 2013 at 10:55 AM, Arvid Björn <arvid...@gmail.com

[Ed Manning]

I installed my Titan yesterday, and it bloody screams.

Images soon as I get through this project deadline.

[Daryl Dunlap]

Congrats!  Wish I could get one of those puppies.

[Steven Caron]

i bet, but damn... its an expensive card.

[Ed Manning]

yes, but reasonably less than Tesla and Maximus setups that it outperforms and uses less power than.

And much less than a new computer.  Makes my quad-xeon 2008 Mac Pro a viable workstation/renderbox  for non-CPU tasks.

[Ed Manning]

In what spare time I have I'm setting up a shootout between Octane standalone and redshift in SI.