ARM V9 Architecture ?

[ProgRocker]

New to Apple, looking into the Mac Mini, which I'll be using mostly for Music Production.

I'm not in any rush, and want to familiarize myself first with Apple products.

I understand ARM V9 has just been released, what exactly would this mean in terms of Music Production and is it worth waiting for as opposed to the M1X or M2 ?

I use Studio One Pro which has recently been updated to native Apple Silicone.

Thanks...

 

[leman]

I understand ARM V9 has just been released, what exactly would this mean in terms of Music Production and is it worth waiting for as opposed to the M1X or M2 ?

Nothing, really. The only thing in ARMv9 that might have any relevance for music production is SVE, the new ARM vector processing instructions that might offer improved performance in some scenarios. But this requires explicit application support (which might take years to be rolled out) and you don’t even need ARMv9, since SVE can be implemented with ARMv8 as well.

 

[ProgRocker]

Nothing, really. The only thing in ARMv9 that might have any relevance for music production is SVE, the new ARM vector processing instructions that might offer improved performance in some scenarios. But this requires explicit application support (which might take years to be rolled out) and you don’t even need ARMv9, since SVE can be implemented with ARMv8 as well.

Thanks Leman. I'd thought the vector lengths of up to 2048 bits would enhance sampling and synthesis and allow for more tracks without freezing. Though as you had said, there needs to be application support. And so far most companies are not very forthcoming with info. I may look at the M1X and see what that will offer.

 

[leman]

Thanks Leman. I'd thought the vector lengths of up to 2048 bits would enhance sampling and synthesis and allow for more tracks without freezing.

I doubt you will be getting vectors of that size any time soon on consumer hardware. Too power-hungry, too expensive, too situational. That’s the realm of specialized supercomputers. Apple is much more likely to stick with 128bit or maybe 256bit vectors.

To be honest, I don’t think that’s such a big deal. M1 already has plenty of SIMD performance (comparable to AVX512 CPUs but more flexible) and if you need truly heavy number crunching on long vectors, the GPU is the obvious choice.

As a tech geek and an avid low-level programmer, I’d love to have SVE in my personal computer. But from a pragmatic point of view, I don’t know that it’s an essential thing to have. For small stuff, you don’t need SVE. For big stuff, GPU is always going to be faster and unified memory makes CPU/GPU data sharing fast abs efficient.