Cycles for Rhino

So let me see if I am understanding this. In the latest WIP when I use rhino render it is not using the Cycles engine. It is only the Raytraced viewport which uses the Cycles engine. Is this correct?

@jtreitz, correct. If you really, really, really want to try Cycles as a plug-in you could browse over to

and get the CyclesForRhino plug-in. Note though that this plug-in is not very much tested, though it should work OK. I just ensured the latest code as of now is now released as v0.0.5.

/Nathan

1 Like

I see, there it is. I missed the “Raytraced mode”. Thanks.

Björn Syse
Syse Industriell Design
http://www.syse.se
+46 706281244

@nathanletwory This plug-in is for the WIP only?

I’m having all sorts of issues with cycles now. Just upgraded to a GTX 1070 and it just won’t run.
If I force it to use the cpu then it works fine (not as fast as Neon though
 :unamused: )
And when I force it to use the GPU then it crashes. It was four simple boxes on a plane in the default scene. Not maximized view.

I haven’t created .cubin files for the latest generation cards yet. Before I do I have to merge latest Cycles upstream master into our branch for necessary fixes for these cards.

@jtreitz, yes.

/Nathan

Nathan, while you are at that, please do the same for the GTX 1060.
I have the same problem as Holo.

-C-H-A-R-L-E-S-

I will post a new message when I have done this. Until then please don’t try rendering on your very new card. It makes me jealous!

/Nathan

Out of curiosity why isn’t the card supported automatically? I thought Cuda was supported thorugh the driver so the software didn’t have to worry about adding specific cards. I did have to manually add the card to the .txt file for PremierePro though.

The Cycles CUDA (and OpenCL) kernels need to be built for your card architecture before they can be used. To be able to do so one would need the NVidia CUDA SDK installed, which isn’t a dependency we’d want for our users to have to install. As such I have disabled any attempts of compiling kernels when they aren’t found.

If you want to give it a shot you could clone the CCSycles git repository, install the CUDA SDK and modify the cubin build script a bit so that you can compile the sm_60 (or sm_61? I have to look up) version of the kernel. Then just drop it in the right place and you probably could have it run well.

/Nathan

:heart_eyes::heart_eyes::heart_eyes:
I just love that you have such high hopes for me!
But I’ll wisely pass and I’ll gladly wait for you to do your magic!

Cheers!

What are we devs without our beloved users, eh
 :wink:

/Nathan

@nathanletwory

(GTX 1060)

Nathan, when do you think this could happen?

-C-H-A-R-L-E-S-

Probably during October.

/Nathan

Late September???
Pretty please?

Cycles is now close to surpassing Neon for my need, but I can only use it at work as I have the 1070 at home.

Here is a small sample of how I use Cycles now. Default environment, a single rectangular light and it gives a nice emotional feel of a real scale model. Love it.

3 Likes

That is a nice looking example :slight_smile:

It’d be nice to have it done ASAP, but I am afraid there are not enough hours left to get the huge pending Cycles merge from upstream into rhino done :slight_smile:

Ah well
 as long as you have empathy for the pain of having to wait
 :wink:

Does this mean what I think it does: you are taking the latest and greatest Cycles release from Blender and putting it in Rhino? If so, will it just have pretty much the pieces that Rhino has now, or will it be the current whole thing from Blender? Does it mean that you’ve pretty much got all the wiring from Rhino to Blender Cycles done and you just need to hook up each new version?

I always have empathy for the people with waiting pains. :slight_smile:

The latest and greatest up until the moment of merging, yes. Not all stuff will be directly available to Rhino users though, since that needs more wiring - like micro displacement stuff, pointiness for meshes (needs an extra processing step that I haven’t implemented yet), and so on.

But the bulk of the stuff is already exposed, and in use by our integration in one way or another :slight_smile:

The merge is needed to ensure proper functioning of CUDA cards all the way from Fermi cards up to the latest Pascals. The drawback of having to do the merge is that a lot has changed on the lower level of code, meaning I have to go over all changes with a fine-toothed comb and a whole lot of patience.

/Nathan

2 Likes

We sympathize

:innocent:

3 Likes