Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
3D Rendering on Apple Silicon, CPU&GPU
- Thread starter vel0city
- Start date
- Sort by reaction score
I think it would be incredibly difficult. OpenGL is deprecated on macOS, but has never ran on iPad (you had OpenGL ES though). All of the UI in Blender is custom rendered using OpenGL.
Maybe it will be more likely after everything is ported to Vulkan and MoltenVK can be used.
Also there is iPad OS's slightly barbaric memory management that just kills apps that use too much memory, which would suck for something like Blender - you accidentally click the wrong button and lose all your work.
Then even if you did finally do the large amount of work to port to the iPad, the licensing / legal issues would probably make it impossible for anyone other than the blender foundation to submit to the App Store.
Maybe it will be more likely after everything is ported to Vulkan and MoltenVK can be used.
Also there is iPad OS's slightly barbaric memory management that just kills apps that use too much memory, which would suck for something like Blender - you accidentally click the wrong button and lose all your work.
Then even if you did finally do the large amount of work to port to the iPad, the licensing / legal issues would probably make it impossible for anyone other than the blender foundation to submit to the App Store.
Blender GPU doesn't even work with Big Sur so even more unlikely on iPadOS. Plus, Apple wouldn't fund it since it means you can get away with only an iPad vs buying Macbook + iPad.
Last edited:
What licensing issues?
Would it be simpler after Blender's viewport gets a Metal backend? It seems Apple will sponsor it.
The Blender roadmap explains: "Vulkan and Metal backends for Blender’s GPU API are being developed. We expect these to be ready to replace the OpenGL backend by the end of 2022."
Blender 3.x roadmap — Blender Developers Blog
Ton Roosendaal shares on what's next after Blender 3.0 and beyond.
code.blender.org
The first is Apple's ban on running interpreted code inside iPad apps, which would mean that all python scripts are out the window, which is a massive amount of Blender functionality not to mention absolutely all add-ons.
The second is the fact that Blender itself is under the GPL which means that while it's free and open source restricts what you can do with the code and any forks of the code. This might not actually be a huge issue, but I have to assume this is why despite the GPL saying anyone can freely distribute Blender, Blender is not available on the Mac app store.
I didn't know Blender was actually getting a proper Metal backend though, that's cool. Last I saw the lead developer said no to multiple viewport apis - good to see he changed his mind.
It seems Blender's license is incompatible with App Store terms. ?
Seems at least AMD had the professional courtesy to supply details post presentation about the slides used. Nvidia didn’t do that for the MBP - and a sly ”underperforming PC” reference to MacBooks on the GFN section.
Would be curious to see more detail on this - is it that the M1 Max just doesn’t do ray tracing well (5x slower?) - would Apple look to bring more ray tracing grunt (they’ve put a decent amount of Silicon down for neural side of things)
Would be curious to see more detail on this - is it that the M1 Max just doesn’t do ray tracing well (5x slower?) - would Apple look to bring more ray tracing grunt (they’ve put a decent amount of Silicon down for neural side of things)
What are you dragging this marketing BS here for? Anyone can draw slides with any numbers they want. This is utterly meaningless without detailed methodology and benchmark breakdown.
NVidia marketing bs = good and believable
Apple marketing bs = misleading and bad
Nvidia, Apple and fanboy BS = bad
That's why I like to do my own testing.
That's why I like to do my own testing.
I moved on long time ago. Now I’m all about chess benchmarks
NVidia has done these slides since the 20 series launch, maybe before. Looking at the chart it’s based on RTX (raytracing) performance. Which undoubtedly NVidia has an advantage in because of dedicated hardware (similar to the “biased” tests that utilized the M1 gpus dedicated hardware).
I hope Apple starts hopping aboard the raytracing train. It’s starting to be a trend.
Metal is clearly designed with ray tracing hardware in mind, so it’s coming. I was hoping to see it with A15/M2 but so far it does seem like we have to wait another year. But who knows, maybe M2 Pro/Max will surprise us?
I’m gonna bet on the M3 having dedicated rt hardware. I think that gives enough lead time. It’ll be interesting to see how it meshes with Apple’s gpu architecture since they use a different raster method (?).
Chess benchmarks are fine, but real pros use hashcat as the true performance metric. Also, tests where the pc is allowed hardware acceleration but the Mac isn't.
Just a baseless speculation my side, but probably not too much. RT just adds some new API to the shading language + the machinery to submit the geometry data to the GPU. One can probably write code where TBDR is utilized together with RT (e.g. via tile shaders), but on their own these features do not really interact in any meaningful way (at least according to my intuition). But who knows, you might be able to do some efficiency tricks with coalescing knowing that your shaders dispatch on the tile etc.
I am not familiar with Optix. But Metal has had a fully-featured RT API (comparable to DX12 etc.) since 2020. And from the looks of it it seems to be designed with hardware acceleration in mind. There is nothing preventing Aple from adding some dedicated BHV hardware via specialized GPU instructions.
RT is mentioned in the Metal Shading Language Specification (page 53) but to be seen how they compare.
https://developer.apple.com/metal/Metal-Shading-Language-Specification.pdf
OptiX and CUDA to get started.
https://raytracing-docs.nvidia.com/optix6/whitepaper/nvidia_optix_TOG_v29_n4.pdf
https://developer.nvidia.com/blog/accelerated-ray-tracing-cuda/
https://github.com/RayTracing/raytracing.github.io
Last edited:
Says the same guy who referred to apple's own marketing slides pre-release of the M1s as if those slides provided conclusive evidence of their performance....which has now been thoroughly debunked.
You generally won't use the Metal Shading Language directly for ray tracing (unless you want to do it all manually). The easiest way is to use the MetalPerformanceShaders framework with its own acceleration structures:
Metal Performance Shaders | Apple Developer Documentation
Optimize graphics and compute performance with kernels that are fine-tuned for the unique characteristics of each Metal GPU family.
There are also some lower level functions specifically for adding acceleration structures and intersection functions to a commandEncoder (but they're probably just using the MPS shaders I'd guess).
On Nvidia cards you can't access the RT cores via CUDA. I haven't done any stuff in OptiX before, but I assume it's fairly high-level like MetalPerformanceShaders, but obviously even faster because uses RT cores instead of shaders.
Apples marketing slides were actually fairly accurate. Do you refer to anything specific?