Quote:
Originally Posted by
CustomStudios
β‘οΈ
So. I replaced the computer in one of my studios with a Mac Studio M1 Ultra. I carefully installed what I needed (only VST3). I did a lot of comparison tests (or more precisely, I compared all the purchased AA plugins face to face) on the computer Mac mini i7/32RAM/1ssd with the new Mac Studio M1 Ultra.
I can say with 100% accuracy only a few of the data obtained as a result - all plugins from AA (especially their newest ones for 23/24 year), as well as CRIMSON1.5 - consume from 5% to 11% less on the M1 Ultra, in contrast from i7 Mac mini 2020. Confident and Stable (the values are different for different plugins). I donβt know what this is connected with - but this is a good result.
For example, ASH ULTRA - consumes 6% less (at any OSx/OSxAA values).
Itβs strange that ASH Ultra Flat consumes 2% more, unlike ASH Ultra. (consistently 2% more on any OS) - but itβs still better, because Flat consumes 4% less on M1 Ultra than on Mac mini i7.
And only βMagic Flowβ (as always) - which was NOT developed by AA, showed itself to be the Worst of all!
Yes - it consumes 1.5-2% less on the M1 Ultra - But..
All the "bugs" remain - as they were before. They were not fixed by DMI.
Regarding stability, working with βMagic Flowβ - there were no crashes in my DAW during Testing. On a real project, I have not yet used it on the M1 Ultra (I did not like it in general, due to the fact that it has an EQ Curve, which we cannot Disable).
In any case, I was pleased that the plugins from AA (which I purchased, and I donβt have many of them, but only those needed for my work) showed the best results in performance with the M1 Ultra chip.
I understand now that the next step I will have to switch to daw REAPER.
PS. I forgot to add an opinion about GOLD5. To be honest, I DO NOT understand what kind of crazy consumption of my M1 is this)))
As soon as I turn on OS2x, I instantly get from 4% to 24%. And this is only one plugin GOLD5)))
Of course, I donβt know who else has more powerful computers - but I know for sure that even if the computer is even more powerful, it will not reduce the load by 20%))
This is the only plugin (in my collection) that takes up 24% of the processor on OS2x.
How do you use it then (namely, a single block GOLD5/GOLD5ZL)? Only if WITHOUT enabling OS?)). Because I CANNOT imagine a large Mix using this plugin with OS2x on at least 5 channels. (What if I need 20-30 bands?). This is simply impossible.
Of course, I have nothing against it (just thoughts out loud).
By the way, GOLD5 literally destroyed my Mac miin i7))
The enhancements and improvements you see are not at all random but result from deliberate development. When comparing Intel and ARM machines, you're not just seeing C++ code recompiled on ARM with adjustments for deprecated Apple structures. Instead, we've heavily optimized specifically for ARM architecture, especially considering Apple's unique version of ARM, which behaves differently. We've also made specific optimizations for Intel, including AVX2 and AVX512, along with various SSE versions. This leads to 5-6 different optimization combinations executed by runtime dispatchers.
A significant difference we observed is in the FFT library implementation. On Intel, the Intel Performance Library outperforms, while on Mac, we support various optimized libraries from different developers, leading to ARM-specific optimizations. Apple's own FFT layer, likely more aggressively optimized, may explain performance differences.
Our products utilize dispatchers that choose the best execution method based on the hardware, continuously updating optimizations with new processor releases. This is why you're noticing performance differences and improvements. The most significant improvements will be seen with Hyper plugins, which employ even more advanced optimization techniques.