....
Assuming that the M1 in the Mac Mini will be able to handle sustained workloads without thermal throttling, I'd strongly consider shelling out $1300 for 1TB of flash storage and 16GB of RAM if my cMP kicks the bucket for some reason. Logic Pro X is my primary concern—I run lots of plugins and sampler instruments, and so far I haven't even gotten the cMP to break a sweat, and it doesn't even have X5680s. I'm still rocking the E5620s.
My only concern is that the divide between the performance cores and efficiency cores translates into fewer available threads overall for stuff like Logic. How is that likely to balance out? Can the four performance cores compete with dual Xeons?
The Apple Silicon implementations (like the M1 and A-series) have a built in performance controller that dynamically allocates work to the Performance (P) and Energy Effecient (E) cores. You shouldn't spend lots of time getting lots in the weeds of core allocations. Generally the circuits in the SoC are going to do a better job than any "back seat driving" end users are going to do.
Audio code that is just using basic math ( add subtract multiply ..... ) computations can be allocated to whatever is appropriate. If look at DAW hardware cards you don't see giant heatsinks and loud blowers to keep the card cool. For substantive numbers of workloads the E cores can be used. There are some corner cases where audio code might do heavy invocation of SIMD/Vector calls which is where they'd be restricted to just the P cores. If see warnings about how your audio software won't work with Rosetta 2 because it only leverages AVX calls on Intel ( x86-64) then there are "limited P core " issues.
Obvious compatibility issues aside, it looks like technology really has progressed so far that a 150-watt kleenex box is just as powerful as this 960-watt gargantuan thing parked under my desk.0
That isn't a particularly new phenomena. Same thing basically happened in the 9 year gap between 2000 and 2009 . Higher end laptop of 2009 was computationally superior in most aspects than some mid range desktop of 2000 . Transistor density doubles about every 1.5-2 years (Moore's Law). 9 years is about 4.5-6 iterations. If CPU vendor can't do better in a smaller system with that kind of transistor budget increase then should just quit the business.
But, the one thing that my Mac Pro can do that a Mini certainly can't is keep my room warm during the winter.......
If want a better "heater" than perhaps should wait. The M1 Mini only replaced the i3 powered Intel version. It also slid backwards on port count ( lost 2 TB ports ) . RAM capacity ( blackslide from 64GB down to 16GB ). There is a pretty decent chance ( not guarantee) that when an " M1X " shows up to power the upper MBP models replacements that the Mini will get a shot at those too.
That Apple will extent the Mini line up into the > $1K range for the "best" standard configuration.
Something in the rage of 6-8 P cores would be my guess. ( there were rurmors of a 8+4 "Mac" SoC model that hasn't shown up yet).
6-8 P cores would be more than capable of keeping up with your 8 cores from effectively a decade ago. ( those 8 cores are AVX less anyway ... so honestly outclassed a while ago on Intel even) . The clock and memory improvements , cache size increases , etc would likely make up anything the lack of threading (SMT).
This all really boils down to you not particularly needed the PCI-e and HDD bays in the Mac Pro. If those don't contribute a critical different then really don't need to be in Mac Pro class anymore. Even if Apple hadn't switched to ARM. The stuff that both AMD and Intel will have rolled on in mainstream desktop by early 2021 ( a couple of months ) would be better. (AMD stuff is already lined up for sale).
P.S. Apple's latest E cores do have SIMD/vector capability. ( they don't have the proprietary Apple Matrix stuff (AMX) ). But if put a heavy SIMD workload on them that performance allocation system will move it off if not a "short burst" kind of workload. And if there is "too much " work to allocate then will labor on the E cores the best it can.
If using Apple accelerate libraries for FFT stuff it is capped to the P cores with AMX , but AMX is so monstrously faster than what cores could do in that narrow subset computational area from 5-10 years ago it doesn't really matter there are half of them. Most "home grown" audio code probably doesn't use it, but stuff that is deeply optimized there is a huge gap where can use instructions that didn't even exist 5-10 years ago.
