It is not about render speed, but setup speed.

In my case (may be different from everybody) I want to spend as little time as possible doing the most amount of iterations and I am willing to sacrifice displacement for the tests, motion blur and quite a bit more PROVIDED when I hit the render button at night with Mantra the next day I can see it with all the bells and wisttles.

It is development time vs render time, right now I feel development time is the bottleneck and engines like VRay and VRayRT address exactly this area.

Specially considering the comiditization of rendering on the cloud which is around the corner and the possibility of sending your render to a limitless renderfarm I think this is going to be even more acute.

Regarding CPU vs GPU, sure enough there has been lots of limitations but they are falling like dominoes by the day so I would not assume the limits of a year ago still remain. For example, the texture limits have been improved massively.

Again, this may not be everybody's view but the potential is too big to ignore. Just have a look at the GDC 2013 demos for nvidia VCA with OctaneCloud and you will instantly see what i am after when I ask for GPU rendering.

:-)

jb

pberto
Robert,

I would rather prefer if resources are spent on making Mantra faster. For that example, which of course is not globally representative, of 3 min Vs 17 seconds IMO does not justify the speed gain. No real displacement, no real transformation and deformation motion blur, no volumes, no shading language, no subdivision surfaces …)

Sorry but IMO GPU renderers have still a long way to reach the quality and reliability (cross-platform & hw-agnostic) of a good offline renderer. If the GPU can assist the CPU render in a reliable way this would be already a big success (and you will still hit driver and cross-platform issues).

P

I'd be interested to hear how much more productive GPU renders are in real-world productions, say for look dev, or is it only really good for demos?

Some guys from Si-Community are testing RedShift already http://www.si-community.com/community/viewtopic.php?f=4&t=3489&start=40 [si-community.com]

Check out Thea or octane for materials dev and sure you will see the potential.

MartybNz
I'd be interested to hear how much more productive GPU renders are in real-world productions, say for look dev, or is it only really good for demos?

Looks like the Render View pane with it's Relighting Buffer goes in some way to addressing speeding up LookDev.

I would love to be able to have a look dev environment inside houdini that allowed me to define the materials at least as fast as this.

http://www.youtube.com/watch?v=F7r69634fVM [youtube.com]

Then build a gallery of materials for the shot and from then on is paint by numbers kind of thing for the assets.

I am willing to sacrifice lots of thing in this environment (displacement mapping, etc) but I would like the parameters of my shader (like the PBR layered material) to be carried across to the main renderer.

I am not inventing nothing new, this is the VrayRT vs Vray approach.

I have been using this system in the past inside XSI with mental ray although the shader development phase was super-slow as the render engine was quite slow too.

Hope it makes sense.

Built a quick test and I'm more impressed with Mantra than ever, in fact it would be faster than Octane as this machine doesn't have Nvidia!)

The best would be to get a like for like test, Octane vs Mantra. Can you build a test that we can all run?

these links to canned demos are interesting but get too boring as they always show strengths.

to compair renders is always hard to judge and i'm not a fan of it.

gpu renders has got speed but not all functions build in or has some other limits.

for mantra it would be nice to have a hybrid solution like that what the luxrender has been done.

http://www.luxrender.net/wiki/GPU [luxrender.net]

it doesn't give the full speed like a gpu render but it will give some boost in rendering and the investment in to the development should be lower.
when i'm correct the luxrender gpu solution was in the 2000 loc area.
or
sesi makes a diffrent way like a vex => LLVM => GPU/PHI concept…..

sesi will know what they are doing… :-)

Reading the VrayRt help files shows it's limits, i.e. only triangle meshes. etc

http://help.chaosgroup.com/vray/help/rt100/render_gpu.htm#load_balance [help.chaosgroup.com]

It appears if you use Mantra with similar limitations it too runs quite smoothly.

I may explore how to fine tune the settings to get similar performance but so far it has been very very far from the video I posted.

Sure I can mimic the workflow, I was already doing it a few years ago with Mental Ray and now with Mantra but the speed is critical, there is a line that once is crossed you just can't go back without asking… politely of course

Hi all,

Has anyone tested Redshift GPU renderer? It has a very good integration with Softimage and Maya and is being actively developed with good customer service and support. The renderer has supported hair and fur, and after version 1.0 they will focus on volumetric rendering. Their beta version is being used in a production environment already. The renderer is designed using an out of core architecture where the scene geometry won't be limited by the gpu RAM amount.

Check it out at redshift3d.com. I would suggest to the community to contact them and ask them to port their renderer to Houdini. They have a plan to support Houdini but their main focus after supporting Softimage and Maya would be to support 3dMax and other dcc packages after that. Probably if the community can suggest to them to also develop a version for Houdini at the same time, that would be awesome.

Thanks.

Cheers,

faizol
Has anyone tested Redshift GPU renderer?

Have you tested it?

MartybNz
Have you tested it?

Yes, I did for several months until I changed my system to using Linux completely due to some programming requirements earlier this year.

Since then some features have been added to the renderer, most notably hair/fur and Softimage's ICE attributes. It's very stable, can handle large scenes and renders very quickly. It's not as scalable as Octane though if used in a multi-gpus setup, but the fundamental architecture (out of core design) enables the renderer to handle large and complex geometry, even if the geometry size is bigger than the gpu ram, which is something that can't be done by Octane.

I was able to produce flicker free final animation render using the renderer in a very short time. Check out their websites, I know a few projects which are using the renderer in full production environment.

If the community can convince the developers, then perhaps they can have their renderer compiled for Houdini integration on all platforms during their participating media rendering development. The code for their renderer is designed to be multi-platform from the very beginning.

Cheers,

Nice! I've signed up to their site though it's only windows and Nvidia right now, so will have to wait…

All is not lost though, I thought I saw a recent job posting at SideFx for rendering skills, and if you had gpu programming it was an additional plus

Looks very cool - would be awesome to get CUDA renderer for Houdini -

thats a suprise for me, the scene renderd in octane:

https://www.youtube.com/watch?v=gLyhma-kuAw&list=UUqPmc3i0TE0oyetAKlv1Z2A [youtube.com]

it's not bad.

I´ve been seen a lot of GPU renders solution lately like redshift which is really impressive no limits for GPU memory like before, anyone knows if Sidefx is planing to make Mantra render with GPU for a further release of Houdini?

To efficiently support ANY renderer requires a significant amount of time and energy, given the quirks and specialties of each renderer. Rudimentary support can be achieved in weeks, but mature, production-worthy support can take months or years of developer time.

I think, at this point, if you want support for a renderer, pressure the vendor to write the support, instead of Side Effects.

This was an age old issue that the Renderman Spec was meant to solve, right?

I agree with you it really takes time and effort and I am not pressuring anyone I am just asking about the possibility which would be awesome IMHO if we could have it with Mantra don´t you think? Don´t get me wrong Mantra is awesome the way it is and it´s my primary choice to render anything.
And If I am not mistaken, since you mentioned renderman, even Pixar is starting to use this technology like others so why not look at this tendency right now?

Cheers!

i would like to see the integration of the cycles render into houdini because there won't be licences fees or conflicts with other company's when you want to deploy it directly in to houdini and the development is going on fast.

but the task for a solid integration is surly not small and takes some time.

for my point of view is that it makes for sesi only sense when they invest time in there product and not in others implementation.