Unmix Song in Spectralayers: time on M4 Macbook?

SpectraLayers
1

I’m looking at buying one of the M4 Macbooks that were recently released, and the only thing I’m really wondering about is how it will compare speed-wise doing e.g., Unmix Song, to my current GPU, a Nvidia RTX 3050.

Or maybe someone running an M4 Macbook can post how long it takes for a given audio file length if that’s a usable way to measure?

Thanks!

2

Without knowing anything about GPUs on PCs and without having an M4 Mac (but an M1 Studio Max) I know that AI driven processes such as ”Unmix song” unfortunately is waaay slower on the M series Mac than on PCs with the right GPU/graphics card. Way slower.

Robin has explained that Apple’s MacOS don’t take advantage of the internal GPU in the M processors. At least up to Sonoma. I don’t know how Sequoia handles that.
I’ve read that Sequoia is designed to handle specific AI tasks built into that OS but I guess that does not apply for general AI driven tasks through GPU processing. Again, I don’t know though.

But the current SpectraLayers is not able to make any significant use of the GPU part of Apple’s M series processors in comparison with the PC/Windows situation.

Robin has hinted though about an idea that could possibly speed up things. [waiting eagerly]

3

To be more precise, SpectraLayers can’t take advantage of the Apple Silicon GPU - macOS can, and Apple applications can, but Apple makes it very hard for external developers to use it too…
I have an ongoing discussion with Apple about it, I still hope that I some point the situation will get easier, but for now it is what it is… In the end yes a GeForce RTX 3050 will likely be faster than a M4 since GPU acceleration can’t be used on the mac side. Once GPU acceleration can be used with Apple Silicon machines it could be a different story, but I really can’t say when that’ll be the case.

2 Likes
4

Thanks for the clarification!

5

Thank you @Robin_Lobel! (and @Marshall) That’s the info I needed. Looks like I’ll stay with my current setup for the time being, since ‘it ain’t broke’.

1 Like
6

Most of the AI stuff, it’s better to upgrade GPU than buying mac, sad but true…

7

Yeah, that’s what it looks like! Yes, regrettable. And right now the cost of the cheapest M4 Macbook Pro is right around a Windows laptop with a good RTX card built in, which is how I came round to considering one. C’est la vie…

8

Can’t we use Metal for GPU acceleration on macOS?

9

That’s the goal - but Apple made it quite hard to use external AI models with Metal. They only provide broken tools so far, and third party framework don’t support metal acceleration very well, or not with the most important AI models.

4 Likes