Final Cut Pro, Compressor, Motion updated with better Mac Studio performance

Here comes a whole new round of Mac Studio review YouTube videos I have to watch! 🤓

What confuses me about this is that Apple‘s marketing for the M1 Ultra seems to indicate that all the performance benefits of the chip come straight from the hardware (and possibly the operating system). 

hucom2000 said:

What confuses me about this is that Apple‘s marketing for the M1 Ultra seems to indicate that all the performance benefits of the chip come straight from the hardware (and possibly the operating system). 

So how does optimizing software lead to better performance on these chips?

Is it possible that compressor didn’t take full advantage of the M1 architecture prior to this update? If that’s true it should result in performance gains not only on the ultra, but also lower tier M1 chips.

I thought the same. Even Pixelmator update and Capture One claims a boost on M1 Ultra so maybe there are a few that can be optimized. Unless they really have old legacy code.  

hucom2000 said:

What confuses me about this is that Apple‘s marketing for the M1 Ultra seems to indicate that all the performance benefits of the chip come straight from the hardware (and possibly the operating system). 

So how does optimizing software lead to better performance on these chips?

Is it possible that compressor didn’t take full advantage of the M1 architecture prior to this update? If that’s true it should result in performance gains not only on the ultra, but also lower tier M1 chips.

Yes, of course it is possible, most especially with 20 year old software. You can pretty much assume that every single piece of nontrivial software has been not been coded for the specific hardware features that it runs on, even though they have been running on the same type of hardware for decades. Look no further than Adobe apps, or MS apps, or web browsers. Every single nontrivial piece of software can be further optimized. Then, in comes down to specific hardware details that software developers won't pay attention to unless it makes their software unstable.

It basically has to be trivial for developers to implement. There have been about 4 to 5 generations of SIMD units on Intel CPUs spanning 20 years. Only the most obvious and easy workflows have been design for AVX256 or AVX2. AVX-512? It's basically only done only on a case by case basis. Then, Apple's SIMD units don't work in exactly the same way, and developers have to rearchitect their software to really get great performance gains. Developers just aren't going to spend the resources to do it. They do what they always do, rely on continuing hardware performance to make their software run faster. Getting a piece of nontrivial software to run stably and have competitive features is already a gargantuan job.