2021-08-03 Blender Rendering Meeting


  • Brecht Van Lommel (Blender)
  • Kévin Dietrich (Blender)
  • Jeroen Bakker (Blender)
  • Patrick Mours (NVIDIA)
  • Brian Savery (AMD)
  • Christophe Hery (Facebook)
  • Feng Xie (Facebook)



  • Cycles X: the main remaining work for feature completeness is in tiled rendering, which is under development by Sergey. Volume sampling quality is now similar to master, but unlike before works on both the CPU and GPU. Still room for improvement here. Next major task for Brecht is probably many light sampling.
  • Anisotropic SSS needs a bit more review and help with versioning, Brecht will look at the patch.
  • Christophe plans to contribute support for manifold next event estimation after the anisotropic SSS patch is finished.
  • Feng has patches to make OptiX BVH builds significantly faster (especially with multiple devices). Plan is to port that to cycles-x and then contribute the patch.
  • Kévin will write a blog post regarding the work done in cooperation with Facebook, regarding Alembic procedural and performance.
  • Final Alembic procedural UI still needs to be reviewed, Brecht will ask Sybren to look at it again.
  • AMD has most cycles-x tests passing with their implementation. Drivers and patch for testing still expected to come within a few weeks.
  • Unclear yet when exactly cycles-x will be merged into master. Optimistically this might be in a month to be ready for 3.0.

Eevee & Viewport

  • Jeroen is currently working on other areas of Blender, so Vulkan development is not expected to see big changes in the coming weeks.
  • Kévin’s OpenSubdiv support is now feature-complete, except for dependency graph handling where other objects depend on the subdivided mesh. Features like loose edges and snapping are working.

Render API

  • Brian will create a task with an overview of improvements that can be made in the render API for renderers other than Cycles and Eevee. We hope that with the help from developers of various external renderers some important issues can be tackled here.

Practical Info

This is a weekly video chat meeting for planning and discussion of Blender rendering development. Any contributor (developer, UI/UX designer, writer, …) working on rendering in Blender is welcome to join and add proposed items to the agenda.

For users and other interested parties, we ask to read the meeting notes instead so that the meeting can remain focused.

  • Agenda
  • Google Meet
  • Time: Tuesday 5 to 6 PM Amsterdam Time
  • Next Meeting: August 17, 2021

MNEE seems to be about caustics, so I am curious about its relation with path guiding. Is MNEE going to be a replacement for path guiding or is it complementary?

Also I recently have read about MNEE in a paper “Slope-Space Integrals for Specular Next Event Estimation” (Loubet et al):

This is in the lower right of the second page and upper left of the third page, under the related work section.
It sounds like from the newer paper’s perspective the MNEE technique seems to have some limitations.
Also, how is this Slope-Space SNEE algorithm? What do the devs think about this?

1 Like

MNEE and path guiding are complementary. I don’t know the details of the MNEE implementation being contributed, but there’s always newer papers that can be used to improve it.


Cycles X: the main remaining work for feature completeness is in tiled rendering, which is under development by Sergey.

@sergey Any chance you’d be willing to do some tiled rendering on a Blender Developer’s YouTube livestream? The videos are always amazing and tiled rendering would be really interesting to learn more about. Seems to come up in forums/chat a bit so could be helpful to have something recorded.

Do I need to download the latest driver? Currently rdna does not support rocm

1 Like

There is no AMD GPU support in the cycles-x branch, it’s under development internally by AMD.


So what about Ambient occlusion? Is it going to be a compositor-only effect? I don’t remember reading anything about its fate…

I think raytraced shaders (bevel and AO) were implemented already, am I wrong?

It’s already in. See ⚓ T87837 Cycles X - Restoring Features

1 Like

They are you’re right. But I mean AO in the world panel

Oh, ok, got it :slight_smile:

No idea about that to be honest.

I was asking because the claim is “feature completeness” and classic AO is not mentioned. Maybe it’s been discarded by design? Sounds strange to me since AO, although an old school lighting trick, is a common feature still used to “just a bit” enhance lighting. Imho anything that can be seen live on the viewport is better than having it post-render


Yes, I get your point, but I don’t know the answer, I presume only @brecht can answer this :slight_smile:

(Hey glare and bloom! I’m talking to you!) :stuck_out_tongue_winking_eye:

@brecht sorry to ask, but since OSL is not mentioned explicitly (regarding Cycles-X) - is Cycles-X will support OSL on the same level as Cycles? Especially interesting in raycasts, and object/geometry data querying (triangle verts, UVs/VCs/etc)

It should be supported on the same level, at least our OSL tests are passing in the cycles-x branch.


It was intentionally removed, as it’s a Blender Internal feature that was only added for compatibility. Personally I think the fast GI options are a good replacement if you want old school AO lighting, and I’d rather not have two different AO mechanisms. It could be added back if there is really a need for it.


Ok, nice to know.
I guess the fast GI feature is enough for the purpose of that “just a bit” I mentioned.
As I wrote somewhere thou, having the chance to use an override world for fast GI would be a huge help, giving the opportunity to mimic the full GI even with fast GI as low as 1 bounce
Edit: Maybe using an isDiffuse lightpath on the world shader would work just fine, still have to try this…

When will Apple GPU get supported?

Hello Community. Is there any Update from AMD on CyclesX. Now even regular blender 3.0 builds are broken.