Cycles-X

Is it normal for exr files to look different from the final render?

Left: render - Right: exr file

The exr files look the same with half and full float.

Tested both ordinary and manually built passes and both emissions work for me, latest cyclesx ofc

You have different tonemapping turned on probably, right is more satured and gamma is raised

1 Like

Great news! Thanks for sharing your work with everyone. I hope it get through for inclusion :clap:

1 Like

I think maybe there’s just something wrong with the Affinity Photo program I’m opening it with.

Just for anyone using Rokens node setup, I would just be aware that the ‘denoise’ part is actually denoising the raw footage aswell.

What I would suggest is havign two denoise nodes in, one specifically on the shadow catcher, and one on the 3D part… that way its all denoised before any colour manipulations that are going on.

In blender the default rendered view would be filmic. In affinity it opens with linear srgb which might explain the difference. You can get the filmic macros for affinity here https://jamesritson.co.uk/resources.html or set blender to srgb instead of filmic and see if you get comparable results.

1 Like

Persistant Data in the render settings is causing this.

1 Like

Thank you for the tip.
I’ll try it.

I’ll never think about that.

oh thanks for sharing the finding.
Hopefully it’s a bug!

Just a heads up, many bugs related to SSS shading has been fixed in the last couple of days. That means with just a few exceptions that can easily be worked around (like plugging something into an empty normal input), Cycles X should now render about everything created in master without regressions (so many of us can now move completely to Cycles X and enjoy things like the improved adaptive sampling, vastly improved denoising, and the more complete random walk SSS).

7 Likes

world ambient oclusion still missing in cycles X :frowning:
*fast gi dont replace it.

I never really used it. What’s missing from Fast GI?

AO in World Properties works like a global ambient light, reducing noise and getting some ‘extras light bounces’, adjusting correct AO distance and factor you can get very clean and fast lighting with 2 or 3 gi bounces, keeping nice color bleeding… some times better than increase gi bounces.

3 Likes

Yes AO is nice combined with fast GI to gain back the light you lose when using very low bounces.
If we could decouple the fast GI lighting from world…

2 Likes

Persistent Data has been broken since its implementation. 90% of Persistent Data is bugs at the moment.

I also had crashes with it.

Wow - been driving me nuts for days. When did tile rendering happen in Cycles-X (and I think it still has problems). I’ve been trying to render a scene at QHD, and it was being chopped at the left. I spent ages going through everything, and eventually put it down to CPU rendering. I optimised the scene to get it to fit into VRAM, and same issue. Then spotted the tile size setting is back, so I turned auto tile size off, and voila, rendered as expected. (The clue was when I let it render the whole scene regardless instead of stopping it, and the cropped part rendered afterwards.

Restored the scene to include all the bells and whistles (mainly adaptive SubD) so it no longer fit in VRAM, switched back to CPU and it worked.

BUT, using conventional wisdom, I dropped the tile size to 32 (for CPU rendering) and it hangs after a couple of rows. Nevertheless, I now managed render the scene (rendering full frame rather than tiles), but when did this happen?

For info, the default tile is 2048, which is 1920 @ around 106%

Oh, and just realised that the displacement node (not modifier) is working without Adaptive SubD. You have to have a highly subdivided mesh, but still a faster render and less memory. When did that happen?

What’s the difference between cycles and octane? How to make cycles create volumetric light scattering and propagation like in octane?