Cycles Development Updates


(Simon Storl-Schulke) #867

Render resolution, image format, frame range, Metadata etc… are all things that are independent from the selected Engine, but define the actual output format. The options related to the current Renderengine are in the Rendertab. Seems logical to me.


(Piotr Adamowicz) #868

My view is that if the setting is passed to the render engine on render, then it’s a render setting. If Blender only uses the setting once the render is completed, then it’s an output setting.


(brecht) #869

Here’s the reasoning behind the change.

The idea is that EEVEE and Cycles viewport rendering is now what you start with, and so the first settings you need to tweak are everything that affects the look and quality of the render. Only later do you need to press F12 or save the render to a file, so having Dimensions as the first open panel does not make as much sense anymore.

The number of panels in the render properties also got quite long and tedious to navigate, with not much space left for new features. Color Management and Simplify are important settings for look and quality, and adding them to the already crowded render settings would have made them even longer without this reorganization.

If this is better or not of course depends on your workflow, particularly how much viewport rendering you do compared to pressing F12.


#870

Well, modern ideas and modern human logic are showing signs of being totally disconnected from reality.

But nevermind, what would be really nice is having some visual info on the header or top bar or somewhere in plain sight… about which render engine is used. Since it’s gone…


(Ace Dragon) #871

Just noticed, Lukas Stockner has had a patch submitted which was more or less unnoticed by well…everyone.
https://developer.blender.org/D3900

It’s a new shading mode that is very similar to the multiscatterGGX, but simpler to compute and with no additional noise. Sometimes the dev. site feed goes so fast you miss these things.


(drgci) #872

Lukas is paid developer? And how many full time paid developers blender have?


(oaschwab) #873

I think they have 6 paid developers.


(lukasstockner97) #874

Technically yes right now, but not by the Blender Foundation/Institute, so all those patches are made in my free time.


(cet77) #875

Fantastic work. :slight_smile: Thank you


(oaschwab) #876

Yes your amazing work does not go unnoticed. Thank you Lucas.


(mib2berlin) #877

Hi, Brecht add support for Nvidia RTX cards to buildbot.

https://developer.blender.org/rBf63da3dcf59f87b34aa916b2c65ce5a40a48fd92

Cheers, mib


(Grimm) #878

Yay! I will test it today or tomorrow. :smiley:


(CarlG) #880

How would I go about using this? I have no idea who this buildbot is :stuck_out_tongue:

I still haven’t ordered a new computer since my main computer broke completely down.


(NewVisitor) #881

(CarlG) #882

Oh, ok. That’s buildbot? I’ve been using it but I never knew it was named as such. So I guess if I download those they (2.79 and 2.80beta) will come with RTX support?


(Grimm) #883

This version is working great on my RTX 2070, no more CUDA sm_75 errors. :smile:


(Ace Dragon) #884

OpenCL rendering has been disabled on MacOS, maybe permanently.
https://lists.blender.org/pipermail/bf-committers/2018-December/049695.html

The reason is because the implementation is more or less unfixable on the Mac. since Apple has deprecated the API. Please note not to place any blame on AMD since they have no control over the platform’s drivers.

At least Apple users have OpenSubDiv now, but this shows they are still on the platform with the highest risk of degradation in experience.


(anaho) #885

I recently asked the question if MIS on lamps is buggy when using the branched path intergrator with sample all lights enabled. I would like to demonstrate the problems on this simple scene.
The scene is modelled in real world scale, consists of two area lamps, is enclosed, all caustics are disabled, the sampling method is Sobol.

Regular Path Tracer ( MIS OFF/MIS ON)

So far so good. Enabling MIS effective removes glossy noise on the objects corners. It looks overall better with MIS enabled.

Branched Path Tracer with sample all lights enabled( MIS OFF/MIS ON)

While enabling MIS also removes glossy noise around the objects corner, it also amplifies/ introduces noise in the transparent part of the object. It is no longer clear, which result is preferable from a noise ratio perspective.

A more extreme example of this can be seen when rendering the Agent Interior CPU file in BPT mode. There is a lot of noise introduced just from using MIS.

Edit: Lukas already replied before the renders on the Agent scene were finished :slight_smile:

MIS OFF:


MIS ON:

In this scene MIS seems completely unusuable.

Is this a known bug?


(lukasstockner97) #886

There is a known problem with MIS in BPT mode, yes - it doesn’t cause wrong results, but noise is higher than it could be. However, I can’t tell whether it causes this particular problem without testing.

I originally wanted to handle it together with the lamp BVH GSoC project, but I should probably just work on a standalone fix.


(anaho) #887

Thanks for the reply . I was unsure whether I should post a bug report so I wanted clearification on this.