Research: Blender 2.83 CYCLES vs Maya 2020.2 ARNOLD (4K) – Speed 400%

+++ VIDEO INFORMATION +++

The video shows the different render times in 400% speed.

Top left: CYCLES CPU
Top right: CYCLES GPU
Bottom left: ARNOLD CPU
Bottom right: ARNOLD GPU

01:53 Compare results

Render time might be affected by recoding with OBS Studio.

+++ SOFTWARE +++

Blender 2.83 LTS (03. May 2020) – Flimic Medium Contrast
Maya 2020.2 – ARNOLD 6.0.3.0 – Unity neutral tone-map

+++ HARDWARE +++

XMG NEO 15
32 GB RAM
i7-8750H @ 2.20 GHz (6 cores / 12 proc)
GTX 1060 (6 GB)

+++ OPERATION SYSTEM +++

Win 10 1909 / Nvidia Studio driver 442.92

+++ SCENE AND DATA INFORMATION ++++

The still life scene was originally set up in Blender 2.79 with photogrammetry models by Oliver Harries:

Special thanks to Oliver Harries for these fantastic free and very detailed models.

All models will use a color, roughness and normal map. All maps were scaled down from 8K to 2K for this test.

Lighting with one High Dynamic Range image.

The Kiwi's fur has been converted into meshes for this test and used for both scenes.

Both scenes will use Physically Based Rendering by aiStandard (Arnold) and Principled (Cycles) and 64 samples for Camera (AA) and all other passes (Branched Path Tracing in Blender).

+++ CONCLUSION +++

This test is using a simple scene without special shader or render intensive effects. ARNOLD is known for handling very complex scenes and high-quality rendering. The render setting and PBR shader were setup as close as possible using the same textures. The scenes use the same HDR image for lighting. The results at the end are very similar. There are differences in the Global Illumination result at the same Light Paths/Ray Depth settings. The CYCLES rendering shows more light at darker areas. There's less noise in the CYCLES rendering.

CPU: CYCLES is 90% faster
GPU: CYCLES is 25% faster

[mohamedatef]

The real magnificent work here is how you managed to almost exactly reproduce the same render output in the two engines.

[joethorpe]

Everyone talking about the render speed, but I just want to eat some of that delicious fruit.

[NarekAvetisyan]

Wow, it's amazing how similar they look. And Blender is free!

[mihailazar]

I genuinely did not expect cycles to be faster ... Now compare it to Octane

[arloc_official]

oh boi didnt expect cycles to be that fast. now im even more proud to be a blender user :D

[jorge]

64x64 is too small for gpu rendering. Try 256 at least. It will be faster.

[nikolaostsimpetonidis]

Great test. Many thanks for your effort !

[wacom]

hey thanks so much for this video! don't listen to the toxic comments a lot of people appreciate what you are doing!

[millie]

And the winner is... Blender. For one simple reason - It’s free. That means all of us that don’t work in VFX can create whatever we like without huge cost implications. Maya simply isn’t cost effective as a freelancer and / or hobbyist. Look at the two results. With a little post processing you could make the two look identical and no one would ever know. But when you’ve got £200 more in your bank at the end of the month that’s when you’ll notice 😉

[eyewitness]

Were you just rendering on the GPU or CPU, or have you tried the GPU+CPU option in later blenders? :) Also, 1/4th the samples plus the intel Denoise node in compositor is just pure witchcraft.

[stevenhoxville]

Nice video! Compare LuxCore and Cycles

[polimetakrylanmetylu]

I really enjoyed this material, but I'm not sure if it's really conclusive. I'm surprised that cycles could deliver the same results as the commercial product, however I think that more testing is required. I'm glad that there is hair on kiwi, and that both renderers could handle it. But I think that you should test out more scenes, mainly
Scenes that feature a lot of light sources (to check if both renderers handle sampling all light sources efficiently)
Interior scenes, preferably lighted by hdr through a small window
Scenes that feature a lot of glass objects, to show how both handle caustics (Arnold doesn't advertise being bidirectional path tracer specifically, but if it is, it would have a huge edge over blender's cycles)

Also, you didn't show some info about render settings, mainly bounce count.
That's all I have to say, thanks for attending my Ted talk

[LucasTbooker]

Which settings did you use for min max samples at Arnold's bucket?
I mean it's just too fast. I have a i7 6700 and a gtx 960. Renders take at least 10 min for Arnold and Vray.

[NarekAvetisyan]

Did you use Adaptive Sampling in Cycles? It can reduce the render time up to 40%! And not to mention if you had an RTX card with OptiX rendering that could reduce it even further!

[darioquizhpi]

GPU support of both were developed by Brecht Van Loomel (Cycles creator)

[gabrielgonach]

In Blender, if you're gonna render with GPU you have to change the tile size to around 256-512 pixels. 16-64 pixels is meant for cpu's only so this could be done in less time if settings were right.

[GUMMY_MKII]

What makes Arnold’s GPU post processing slow though?
It seemed to be faster at first

[rakajae]

can you change the render option to processor in cycles? just wondering what happen if they both using procie as the engine.

[skrotov]

does arnold gpu support tile rendering? if you can choose it arnold will probably have same time or maybe it will be faster

[AndreaAvellinoVideo]

Have you guys tried E-CYLCES out? It's blazing fast, I work only with that since one year now.