Does VEAI on M1 ultra utilise all the 32 neural cores?

Good day! It is my understanding that VEAI uses the neural engine on Apples M1 chip line-up. First of all, is this assumption correct? If it is, can we realistically be looking at a doubling of speeds when upscaling using the M1 Ultra chip (found in the Mac Studio), as it has double the neural cores? Furthermore, has any efforts been put in to optimise the software for this chip? Thanks in advance!

Bumping this for the hope of a response. I am interested in this as well.

Nice! I asked support about this and this was their answer: “When processing footage in Video Enhance AI, low power mode will only make use of the Neural Engine, although both the GPU and Neural Engine are used for processing, and can be fully taken advantage of in the new M1 Ultra found in the latest Mac Studio.”

Not sure if that’s the case though, as I haven’t really seen any real world benchmarks from the M1 Ultra. I might have seen one number, if I remember correctly, however that didn’t correspond to the hypothetical doubling of performance that is expected with twice as many neural cores

So far I have seen one mention of the M1 Ultra and that was a speed bump of roughly 50% on the Artemis model.

Of course a lot of factors are in this calculation. Different models use different parts of the hardware (CPU, GPU, NE) and I am sure there is still room for optimization.

1 Like