3D Rendering on Apple Silicon, CPU&GPU

[sirio76]

Sure, and I would like to receive a new laptop for Christmas, it’s the 30-inch MacBookTroll Ultra, I think I’ll write a letter to Santa..

Edit, just to be clear this was a reply to a post that was deleted

 

[jujoje]

Looks like Octane are getting some nice performance benefits from the new hw raytracing:

Octane X 2024 is being optimized for new ray tracing hardware on M3 Apple GPUS. This is scene dependent, but is already showing 2x to 12x(!) speed gains in heavily instanced scenes (link)

And the other nice thing that stood out for me was rendering across different platforms/hardware

Mixed Platform Network Rendering
Until now we used a different memory layout for geometry data on Metal (macOS) and CUDA. This made the compiled geometry data incompatible between macOS and Windows/Linux, prohibiting mixed platform network rendering. In version 2024.1 we unified the memory layout, allowing now network rendering on both platforms

Don't really use Octane myself (gave it a go, but went back to Karma CPU in the end). Interested to hear how anyone finds it these days on macOS? They seem much more focused on AS and Metal performance than Redshift (although always feel Otoy overpromise a bit).

Also, as a side note, they have some kind of Black Friday deal going at the moment.

 

[Spybreakjj]

I'm really excited about Octane 2024 and plan to test it extensively as soon as my M3 Max comes in on Monday/Tuesday.

I've used it this last year on my M1 Max but mostly during the first 10%-20% of building a scene. Once it got too much for the live viewer I would move the project to my 4090+3090 PC and finish there. Looking forward to being able to do more work on the M3 and hopefully network render with both machines all from MacOS.

 

[Appletoni]

Sure, and I would like to receive a new laptop for Christmas, it’s the 30-inch MacBookTroll Ultra, I think I’ll write a letter to Santa..

That's a very good idea. I will buy this device too for chess.👍👍

 

[Appletoni]

I do lighter, more technical work, mainly in Houdini and have been using a m1max for the last years as my primary machine. It has been rough in many ways but also enough. When doing any serious work, I am usually connected to a display and have a lot of gadgets connected so I have been waiting for “the serious Mac desktop” for a while. When doing any sim or render that takes more than 5 mins, fans spin up. Not loudly, but enough to stress me about what happens with dust buildup etc. also, the battery gets worse by time so in the end it will become a desktop anyway.
The m3max uses a lot more power it seems, so fans will have to be working harder. The “lust” for this machine is high, but I will wait for the ultra. That will be a machine that can stay with me for many years. A m3max specced out enough is in the 7000$ range, way to much for a biannual investment.

Funny I'm doing exactly the same things using Houdini and ChessBase 17 and Fritz 19😁.
I will wait for the ULTRA too, hopefully something between 18- to 20-inch and until then I will stay with the M3 MAX maxed out MacBook.

 

[skippermonkey]

Funny I'm doing exactly the same things using Houdini and ChessBase 17 and Fritz 19😁.
I will wait for the ULTRA too, hopefully something between 18- to 20-inch and until then I will stay with the M3 MAX maxed out MacBook.

I don't think you'll get an M3 Ultra in a laptop.

 

[hovscorpion12]

Yeah. The Ultra chips will be reserved for the desktops. The Max chips are no slouch. With up to 128GB RAM, there should be no limitations.

 

[Xiao_Xi]

What are those differences in Mesh Shaders?

Experiments for optimizing shadow maps for the dynamic shadow maps. Idea is to add a simple depth pass which is a transform using mesh shaders, just moving the position and most likely be optimized away. This is for now only used by the Metal backend due to the differences between Mesh Shader optimizations on other platforms.

2023-11-27 EEVEE/Viewport Module Meeting

Practical Info This is a weekly video chat meeting for planning and discussion of Blender EEVEE/viewport module development. Any contributor (developer, UI/UX designer, writer, …) working on EEVEE/viewport in Blender is welcome to join. For users and other interested parties, we ask to read...

devtalk.blender.org

 

[singhs.apps]

I'm really excited about Octane 2024 and plan to test it extensively as soon as my M3 Max comes in on Monday/Tuesday.

I've used it this last year on my M1 Max but mostly during the first 10%-20% of building a scene. Once it got too much for the live viewer I would move the project to my 4090+3090 PC and finish there. Looking forward to being able to do more work on the M3 and hopefully network render with both machines all from MacOS.

Oh man yes. Will like to hook my pc to the Mac to get render tasks done via the pc. Let me know how u have set it up whenever you have the time.

 

[Xiao_Xi]

Some say Blender 4 might be slower because of the new Material nodes but not sure.

A forum user has quoted the lead Cycles developer on the differences between Blender 3.6 and Blender 4.0 in the Blender forum.

For 4.0, the main reason you shouldn’t compare it’s performance with previous versions of Blender is because of the shader and light changes. Changes to the shaders and lights will make them render differently. This makes comparisons of performance unfair because you’re not comparing the same render between versions.

Cycles AMD HIP device feedback

There are two things going on here: A user earlier in this thread reported that HIP-RT in 4.0 was slower than HIP in 4.0. When looking at testing in 3.6, they observe the opposite, so this seemed concerning. What Toms hardware (and some other people) have observed. GPU rendering in Cycles...

devtalk.blender.org

 

[aytan]

A forum user has quoted the lead Cycles developer on the differences between Blender 3.6 and Blender 4.0 in the Blender forum.

Cycles AMD HIP device feedback

There are two things going on here: A user earlier in this thread reported that HIP-RT in 4.0 was slower than HIP in 4.0. When looking at testing in 3.6, they observe the opposite, so this seemed concerning. What Toms hardware (and some other people) have observed. GPU rendering in Cycles...

devtalk.blender.org

There is no significant difference between 3.6 and 4.0.1 on M1 Ultra, rendered out same scene both of them difference is 18 seconds better at 4 minutes 22 seconds total render time with Blender 4.0.1. This issue could related only with AMD GPUs. I have no clue about new shader system is slower on any Apple Mx Devices. It works just fine on Mx.

 

[leman]

There is no significant difference between 3.6 and 4.0.1 on M1 Ultra, rendered out same scene both of them difference is 18 seconds better at 4 minutes 22 seconds total render time with Blender 4.0.1. This issue could related only with AMD GPUs. I have no clue about new shader system is slower on any Apple Mx Devices. It works just fine on Mx.

According to official Blender benchmark database Mx series seem to be around 6% faster in 3.6 compared to 4.0 (with M3 being the obvious exception, of course).

Blender - Open Data

Blender Open Data is a platform to collect, display and query the results of hardware and software performance tests - provided by the public.

opendata.blender.org

 

[Rafterman]

With up to 128GB RAM, there should be no limitations.

Except the $1,000 to upgrade to that

 

[altaic]

…

 

[iPadified]

Except the $1,000 to upgrade to that

$1000 upgrade to get a 128 Gb RAM GPU render setup is amazingly cheap. Try that with NVIDIA.

 

[Appletoni]

$1000 upgrade to get a 128 Gb RAM GPU render setup is amazingly cheap. Try that with NVIDIA.

Don't forget it's also for DDR5 RAM + much faster
and not DDR4 + much slower or DDR3 + much much slower.
By the way I paid $500 for 64GB DDR3 RAM!! (Windows PC self configuration) FOR CHESS.

 

[Appletoni]

Yeah. The Ultra chips will be reserved for the desktops. The Max chips are no slouch. With up to 128GB RAM, there should be no limitations.

Depends on what you are doing CHESS could be a limitation.
I'm still waiting for my 1TB RAM inside my MacBook MAX or ULTRA.

 

[Macintosh IIcx]

I'm still waiting for my 1TB RAM inside my MacBook MAX or ULTRA.

No you’re not…

 

[Xiao_Xi]

What could those issues be?

Metal support is under development in OIDN [Open Image Denoise]. There are some technical issues regarding how Metal needs handles instead of device memory pointers, and there was some discussion about the precise implementation of that.

2023-12-12 Render & Cycles Meeting

Attendees Brecht Van Lommel (Blender) Stefan Werner (Intel) Nikita Sirgienko (Intel) Attila Áfra (Intel) Xavier Hallade (Intel) Patrick Mours (NVIDIA) Brian Savery (AMD) Notes The meeting in two weeks will be skipped due to holidays, the next one will be on January 9. OpenImageDenoise For...

devtalk.blender.org

 

[Appletoni]

No you’re not…

Okay I need 19 TB RAM to fill the Syzygy Endgame Tablebases from SSD into the RAM but that's another story

 

[sirio76]

I fill sad for you

 

[leman]

What could those issues be?

2023-12-12 Render & Cycles Meeting

Attendees Brecht Van Lommel (Blender) Stefan Werner (Intel) Nikita Sirgienko (Intel) Attila Áfra (Intel) Xavier Hallade (Intel) Patrick Mours (NVIDIA) Brian Savery (AMD) Notes The meeting in two weeks will be skipped due to holidays, the next one will be on January 9. OpenImageDenoise For...

devtalk.blender.org

Isn't that written in the notes? It seems that the code relies on unified virtual memory between the CPU and the GPU, which Apple does not support.

 

[altaic]

…

 

[diamond.g]

Isn't that written in the notes? It seems that the code relies on unified virtual memory between the CPU and the GPU, which Apple does not support.

This feels like a dumb question, but how does Apple have literal unified memory, but not support it virtually?

 

[leman]

This feels like a dumb question, but how does Apple have literal unified memory, but not support it virtually?

Those are two very different things. Virtual address is what your application sees, the hardware uses a sophisticated mechanism to translate it to the actual physical address. This enables a bunch of useful features (e.g. the address might point to a file on disk instead of physical RAM, you get additional safety, and the OS can move/compress the physical memory without you even noticing, etc. etc.).

For unified virtual memory you need to use the same address space on the CPU and the GPU, so that if you copy a memory address between devices it will be valid and point to the same data. I do not know why Apple does not support this, one would think that their hardware would be perfectly capable of sharing memory page descriptors. Maybe there is still some legacy reason, or maybe there are additional complications. But who knows, it is also possible that Metal 4 will have unified virtual memory on Apple Silicon but will drop Intel-based Macs or something like that.