M1 v M4 Real world comparison - Rendering in Maya

[Allen_Wentz]

I have a MacBook Pro M1 Max and a MacBookPro M4 Max both with 64GB.
I am running Maya 2025.3 and rendering the same 125 frame animation on both.

The M1 rendered in 55 mins.
The M4 rendered in 24 mins.

1) Certainly anything that involves shading, 3D or the like should improve with later Apple SoC like M4 vs. M1. Apple has made specific hardware improvements aimed at ray tracing, etc.

2) Given the intensity of Maya's hardware demands IMO anyone intending an app like Maya should pay for maximum available RAM, not just 64 GB. Apple's Unified Memory Architecture uses RAM quite intensively.

3) [For me] rendering is not a primary important performance characteristic. Because once one commits to a render, one's creative process is already by definition fully interrupted, for minutes, so one will for sure be going to do some other activity. The things I care about are essentially real-time changes that can be fast enough to maintain uninterrupted design thinking versus slow enough to break the thought process. The relevant timings are very fast: seconds or preferably milliseconds. Testing an image blur or color change might be examples.

 

[innerproduct]

So we are talking rendering with Arnold I take it? That is only using the CPU so in that case the uplift is inline with the multi core scores in other CPU renderers like Cycles (going from 250 "points" to "450 "points") If you switch to redshift the uplift will be much higher for simple project that benefit from the GPU advancements. But for more complex scenes, CPU is still king. (try 3delight!)

 

[crazy dave]

But for more complex scenes, CPU is still king. (try 3delight!)

As someone who doesn't do rendering at all, what advantages does CPU rendering still have over GPU rendering?

 

[innerproduct]

As someone who doesn't do rendering at all, what advantages does CPU rendering still have over GPU rendering?

I guess it depends on who you ask, but in general, for one, they are more stable and usually supports more features. I don’t do much right now (so this might have changed lately) in this field but I have had great experiences with both but for different tasks. For small things (likr the anim in this thread) you would benefit massively going for redshift or octane while a scene for a blockbuster movie would be a worse fit.
Also, if you use GPU, the scene has to be transferred to the GPU (etc, like kernel compiles) which slows you down when doing interactive renderng, that is, for example, tweaking the scene is some way.

 

[locksmack]

1300 36 70 70!

Those ads will forever be etched into my brain.

Impressive performance increase on the M4, especially considering the M1 is still a reasonable chip 4 years later.

 

[Sami13496]

Now YouTubers can render videos about rendering times twice as fast

Jokes aside, for people who do rendering as part of their work, this can have a huge impact. For example 1 month a year of rendering instead of two. Saved time can be used on something productive.

 

[funwithstuff]

For anyone considering Blender, the official benchmark charts have been pretty reliable as an indicator for GPU performance. With GPU, I could render 3x faster on the M3 Max compared to the M1 Max.

 

[Allen_Wentz]

Here is the animation:
View attachment 2472601

That little animation is great. I have come back to watch it about ten times.

 

[leman]

As someone who doesn't do rendering at all, what advantages does CPU rendering still have over GPU rendering?

It's generally easier to code and maintain. The infrastructure (compilers/tooling) is generally more robust. The skills and competencies are more readily available. And the environment itself is more stable (you never know what weird optimizations a GP or a driver might pull on you).

 

[sirio76]

As someone who doesn't do rendering at all, what advantages does CPU rendering still have over GPU rendering?

It’s more scalable, more reliable, easier to code and debug. GPU rendering have other advantage.

 

[ThailandToo]

I think the OP meant the project is 125 frames total, not that either M-chip is rendering at 125 FPS.

Going 55min to 24min is a pretty sizeable improvement to me. That reduces your work time by more than half!

I agree that it’s a good gain, but it isn’t time working in reality. It’s wait time. When doing these things, in reality, you just start the render when day is finished or when going for lunch or etc.

 

[MadDawg2020]

render time per frame is dependant on the complexity of the scene. I think a 100% + increase in speed from M1 to M4 is impressive.

I’d like to see the comparison between an M4 and a loaded iMac Pro. Im sure and M4 would be 10x faster or more!

 

[Meshbuilder]

I’m not sure which render engine was used in Maya, but if the rendering was done on GPU, the M3 and M4 processors would be significantly faster than the M1 due to the new ray tracing cores introduced with the M3. Here are some benchmarks (GPU-based) using Blender’s render engine Cycles and Maxon Cinebench (with Redshift) both of which are optimized for Apple Silicon. I’ve also seen similar results with Otoy’s Octane Render.

The latest M4 Max is in some cases as fast as an RTX 3090 or a 4070 Super in Cycles, which is pretty cool!

 

[biting-mammal]

Roof Seal, is that you?!