Well, if Apple GPU would beat the rtx 3090 we would have another problem. Crypto miners would buy all available laptops and we couldn’t use them Nvidia and AMD have a giant problem here, that won’t be solved soon. There hardware is easily twice as expensive then it should be.
3 Likes
If no other graphics cards should run with the Mac anymore, then I demand that from Apple!!! 
2 Likes
For 3.0 (Cycles X) and later I thought progressive rendering was the fastest, no?
I haven’t found a tile size combination that is faster than setting it to progressive for 3.1.
Yes same here. This is actually in the sure code just wanted to see what if changing it and recompiling would make a difference.
1 Like
The basic question here is why you are comparing what is essentially alpha development with very, very mature tech? Why do you keep doing this? Because you are impatient?
Check M1 Max performance in optimized settings in another year and you can START to make the claims you are making. This is always how technology has worked. We all wish development were much faster but it isn’t and it never has been…
2 Likes
Did a brief test out of curiosity with the jan 23 build, its pretty stable, not any faster though but stable. Im enjoying the journey with this development being so live and direct. Gotta love blenders direct connection with the community at large,
5 Likes
Interesting dev statement.
TWINMOTION VERSION
2022.1
Note: Unreal Engine has implemented Apple’s Metal API for using their GPUs. However, we have not implemented Metal’s ray-tracing functions because Apple has not released any hardware accelerated ray-tracing support so far. The difference between hardware-accelerated ray-tracing and non-accelerated ray-tracing is the difference between rendering a single image in minutes instead of hours. We are researching ways to bring Apple computer users the speed and capabilities of our Path Tracer, so we are not waiting on Apple to solve this problem.
5 Likes
You mean like implementing Apple’s raytracing capabilities with Metal. lol
First off, glad they’re “working on it” but… their post makes zero sense, you can’t add RT cores with a software upgrade.
And, did engines not use raytracing before RT cores?
1 Like
There were conversations above that were deleted by the moderators so you didn’t understand the context, and I got annoyed with some answers that’s all, but now the subject is dead. 
2 Likes
I’m pretty sure the moderators don’t delete comments unless something is way over the top like racial slurs. Then it’s not the comment getting deleted but the user.
there was no racial slur or anything like that, i was rude and they removed it for that, that’s all, but the matter is over… let’s not relive this…
5 Likes
Ah! No problem. Welcome aboard!
5 Likes
I have no idea what this statement means…
Unless epic adds support for metal raytracing in unreal, there will be no support in twinmotion either. If there is no hw acceleration in M1 chip then there is no point to add anything, or is there?
1 Like
From my limited understanding, both Metal and Optix take care of raytracing.
Optix can, as far as I know, do raytracing on 1080 cards without RT cores. Its not as fast but it works.
I suspect Metal is similar. It can do raytracing on non-RT Gpu cores, its just not as fast but it works.
On the unreal and blender end, they still need to interface with Metal and Optix to do raytracing and benefit from the either system. Im not sure what Metal and Optix are called… Application Program Interfaces or something?
This is a bit beyond my understanding but perhaps someone else can highlight this in a clearer or more definite way
I think this is pretty correct
This is not about using metal performance shader to do raytracing
This is about a missing ability for using dedicated hardware acceleration for ray tracing !
In my understanding MetalRT is an API similar to Embree or CUDA that saves you coding a ray intersection algorithm by hand. Optix is an API that can use nVidia RTX and tensor cores for acceleration which CUDA can’t, therefore raytracing speed on non-RTX cards should be similar with CUDA and Optix.
Apple doesn’t provide RT cores yet and won’t get anywhere near RTX+Optix even with MetalRT. Also the new Blender Metal build just ditched MetalRT in favour of BVH2 “because it’s faster”.
Seems there is room for improvement of Apple’s API and hardware too.
(Currently they are adapting GPU drivers just to make Blender work better.)
Seeing the dedication of Apple software engineers here with Blender and elsewhere too I’m slightly confident for the M2 and also better APIs to come.
Considering final rendering the current stuff isn’t quite there yet, especially considering the price tag.
But for everything else Blender on M1(pro/max) is already really awesome.
7 Likes
that is indeed - I think apple realized how broadly adopted and useful blender is for content creation - which Apple knows is their creative market.
1 Like
I hadn’t heard about this, is this announced officially anywhere?
This is Apple’s first foray into GPUs for their laptops and desktops. There is probably a lot of things they didn’t think about that they didn’t have to deal with on the mobile side int-terms of api and hardware support. Hopefully they continue addressing these issues in the future. They can no longer blame AMD or Intel for issues with the GPU. They need to get this right even if it takes a few iterations.
Blender is the perfect piece of software to work these kinks out.
2 Likes
