Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
The first post of this thread is a WikiPost and can be edited by anyone with the appropiate permissions. Your edits will be public.


macrumors 65816
Original poster
Oct 27, 2021
This wikipost is a summary of the state of the art 3D modeling and rendering software for Apple Silicon. You can improve it by adding new software, completing missing information and posting benchmarks.

To keep this wikipost as useful as possible, please use the other thread if you have any comments or questions about this wikipost.

3D DCC software​

last update
Blender3.511 or higheryesyes29/03/2023
Cinema 4D2023.210.15.7 or higheryes?29/03/2023
Houdini*19.5*10.15 or higheryes?20/07/2022
LightWave**2020.310.12 or highernono21/05/2021

3D Sculpting software​

applicationversionmacOSnativeoptimizedlast update
Zbrush2023.113.2.1 or higheryes?29/03/2023

3D rendering software​

applicationversionmacOSnativeoptimizedlast update
Arnold7.1.4.0 - 17/11/202210.13 or higheryes7.2.1.0 - 29/03/2023
Cinema 4D native2023.210.15.7 or higheryes?29/03/2023
Cycles3.511 or higheryesyes29/03/2023
Eevee3.511 or higheryes?29/03/2023
Pixar Renderman
Redshift3.5.1413 or higher, 16GB or higheryesyes29/03/2023

* Houdini went "Gold" status on Apple Silicon on 22/02/2023
** LightWave does not run native yet. Users have identified contexts that will crash LW on Silicon.

Last edited:


macrumors regular
Dec 20, 2022
First of all, I did this research for myself. I hope it will be useful for anyone who looks for information about the topic.

As I need a subject for testing, I prepared a simple set up with Earthquake model from ZBrush. I unwrapped UV’s and export a high poly mesh as FBX file. Whole scene contains more than 3.153.000 polygons and file size is 305 MB.

I did it that way because I want to import same FBX file for each 3D DCC/Render engine.

I use same set up for every Render engine with it’s own standart material and quad lights which are positioned exact same xyz.
For obvious reasons I do not try to get out exact same final render at the end also do not want to spend time for reach better renders.

Anyone could achieve shorter render times or better renders with own render settings or own tricks.
But this is not my target for this research. I tend to simulate when a user fist time use a renderer what will be first impression.

First I rendered out via Redshift. It's default render settings as CPU renderer and as GPU renderer on latest C4D version.
( Redshift's default render settings Unified Sampling is off ).


Than I build up the scene with Arnold Renderer and also rendered out with default settings. So this gave me an idea how far can Redshift go as CPU renderer.


Most interesting thing is Redshift GPU rendering 00:12 seconds and Arnold CPU rendering 00:13 seconds. They had equal render times for this simple scene. Sure there could be different scenarios with different scene structures and results could vary.

Than I decided to feed them with more polygons. I duplicate Earthquake 12 times and repositioned the lights for both scenes.


As you can see from memory tab Redshift demands nearly double amount of memory from system.
This shows me if I decided to use CPU rendering on AS Mx Arnold could provide reliable and easy solution.
Indeed GPU rendering is way faster but sometimes for uniqe/high memory demand scenes you had to be use CPU renderer as a life saver. Also Arnold has got some special abilities/shaders or workflows.

Also I imported single Earthquake scene in to Blender and set up scene with Blenders own shader and quad lights. Rendered out Blender's Default Cycles render settings.


Blender is not consistent for every single render time. Sometimes there were %40 differences for every single render. So I rendered out scene 5 different times. This results are the best render times I could achieved.
A few render ( Cycles GPU/Metal/Full Kernel Settings ) ends up over 34 seconds. On the other hand CPU rendering is very consistent and render time differences were only %5.

What I get from this research for myself is : ***Redshift is fastest and reliable GPU renderer on AS M1 Ultra.
Arnold is the most consistent and quite fast renderer on AS M1 Ultra.
I can not add V-Ray or Corona to this multiple comparison test because when you download V-Ray Demo everything works great but when your Demo period is over it messes up whole C4D and C4D plugins. It has to be removed from evert directory/folder or subsystem otherwise C4D will not start up. I have experienced this on Mac and PC same time.

As anyone can understand Blender is way behind on that subject. I expected better results from Blender but turns out my feeling was completely wrong. Blender's own Demo Scenes are fine internally. I never seen a correct converted Blender Demo Scene via other 3D DCC.
Last Blender Demo Secene is blender_35_splash_nicole_morena. M1 Ultra 48 GPU rendered this scene 00:14 seconds. I think M2 Max 38 GPU could achieve a result around 00:10 seconds.
Scene link :

That's all for now
*** I am going to share Redshift v.s. Octane next ;)


macrumors regular
Dec 20, 2022
This time I compare Redshift and Octane with simple Earthquake scene and multi Earthquake scene.

For this I try to equalize both render engines.
For both of them : Denoise off
Sample count 128
Adaptive Sampling 0.01
I turned Redshift Bucket rendering off and Progressive rendering on. Because Octane uses Progressive rendering or I do not know how to change it to Bucket rendering.

Results are interesting


For this simple scene Octane uses half amount of memory and rendered out nearly %50 faster.
Both of them used same amount of Swap Memory 643.8 MB. but Redshift uses twice as more total memory.
Redshift 19.86 GB / Octane 9.84 GB total Memory.

Things go weird when I pushed both of them for Multi Earthquake Scene. There is 12 Earthquake on the scene.

Same Render Settings above.


Now Octane demands at least %40 more memory than Redshift.
System was under memory pressure while Octane rendered out this scene.
On the other hand Redshift used same amount of memory at both stages for this scene.
Redshift 48.31 GB / Octane 62.64 GB total Memory.

Another interesting point is Octane used 12.78 GB Swap Memory Redshift is only using 1.014 MB Swap Memory.

Because there is 12 Earthquake on the scene. Both of them used same amount of memory when there is only 1 Earthquake on the scene it was 643.8 MB.
Now there is 12 Earthquakes so 643x12 is 7.716. There is a Memory Gap nearly 5 GB on the Octane side, I don't know is this a memory issue or default for Octane.
This is weird because Redshift used only 1.014 MB Swap Memory for 12 Earthquakes.
When I read about Octane I have found this. Octane converts all meshes to its native rendering solution.
Scott Benson explains that very clear :
Octane triangulates any mesh fed into it. If you build a model with a combination of triangles, quads and n-gons in C4D and start the live viewer, Octane will convert all of that to triangles first before doing anything else. This is why with lower density meshes, there might be cut corners or zigzag geometry if you have sharp geometric shapes in your displacement where you might expect a smoother extruded look.
This also explains why when you create a model with a certain amount of quads, Octane will show twice as many polygons as you think you have. Triangulation isn’t necessarily good or bad, you just want to be aware that this is what’s happening.
Here is the link
Also for Triangulation

Anyone could check out Octane documents which are prepared very good and detailed.

What I understand from this experiment is rendering time gap between Redshift and Octane somehow narrows a lot when scene has got more objects or much more poly count.

Besides that Octane is the fastest Render Engine you can find.
If you have bought any Mac from Apple you can use OctaneX free on any 3D DCC even with Unreal Engine.
As they say on Octane forums ''Happy Mac GPU rendering!''
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.