Another real life, pro scenario: an entry-level MacBook Air with 8 Gb of RAM (the subject of this thread) literally trounces… an iMac Pro in Final Cut Pro.
French article, but in a nutshell: exporting an H.265 8k video on the MBA is 158% faster than on the iMac Pro. All while remaining entirely silent. For an entry level machine.
Thanks to Apple’s history with the A-series chips I actually think we have a pretty good idea where we are on the curve: around 20-30% per year on average. There might be some expectation that Apple might be able to do better than that as it learns more about how to adapt its chips for macOS, but the strength of the M1 out of the gate makes that less likely. They’ve probably already gotten a lot of the low-hanging fruit when it comes to optimizing the chip for the Mac.
I should note that here I’m thinking about year over year improvements to the same chip (same number of CPU and GPU cores, etc.). There’s obviously still plenty of room for speed increases by increasing core count for higher end machines (16” MPB, iMac, Mac Pro).
Well - given that Windows/Parallels is essential to my mobile work and it may take a while to achive that with an M1 Mac, I am about to pull the switch and upgrade my 7-year-old MacBook Pro to a new 16" Intel model with 64 Gb RAM. I can always trade it in if Mx ARM technology advances faster than predicted with Windows virtualization.
That said, I must say this is a stunning review - if this is really “entry level” performance then this may well be a winning move for Apple:
In your use case that is an excellent choice. And I am confident that you will find a good solution when your shiny new Macbook Pro eventually will be long in the tooth years from now. 
The graphics performance is interesting:
The M1 performs much better than Intel integrated graphics, but it still gets beat by the discrete GPU in the 16". However, the difference isn’t as big as I would have thought. The biggest performance difference (on the OpencCL Geekbench benchmark) is still less than twice as fast and on most of the other tests the difference is in the 30-60% range. And this is the wimpy 7-core GPU! I could definitely see a 12 or 16 core GPU obviating the need for a discrete GPU in an Apple Silicon 16" MBP.
That’s what I’m hoping for. There are currently ARM chips out there with 80+ cores, and the ability to access terabytes of memory. It’s all high-end server stuff, granted, but it’s possible with the architecture.
I would imagine that heat dissipation (and doing it without excess noise) is the major challenge as far as putting this in the hands of consumers. But one of those big Mac Pro cases could accommodate quite a bit in the way of cooling equipment.
Indeed - my main desktop computer is a Mac Pro with lots of RAM and SSD space - no doubt Apple will eventually figure out a way to entice me to upgrade that.
Or you could purchase an inexpensive Windows PC and run it via Microsoft Remote Desktop. I abandoned VMWare/Parallels and gave my users access to real PCs, and eliminate 90% of my Windows related support calls.
That also gives the useful benefit that the Windows PC is doing the processing as well. Unless it’s something that requires graphics performance (in which case a VM is an odd choice as well) it’s kind of like adding some dedicated Windows cores to one’s Mac.
That is certainly an option.
But if do that then I lose the ability to use Coherence Mode - which really is nice for integrating the two operating systems to share all hardware and files effortlessly.
Plus my main use of Windows requires a good bit of computing power - I am using Madcap Flare to manage a large document database to be included in a writing project.
That’s true. And you would miss all the fun troubleshooting your hypervisor when a macOS update drives a stake through its heart 
There’s no perfect solution. You gotta do what’s best for you.
Ars Technica has a nice review of the M1:
“If Apple’s M1 isn’t the fastest single-thread—and quad-thread—consumer-available processor on the planet, it certainly isn’t missing it by much.”
In this vein, after a few hours using the M1 MacBook Air with a suite of not particularly demanding apps (OmniFocus, Microsoft Word, Keyboard Maestro, Fantastical, 1Password, and Obsidian) @MacSparky posted:
Even though it’s only been a few hours, I’m already using this Mac to do work and the word that just keeps jumping to my mind is “snappy”. I’ve never had a Mac that jumped to my command like this. The way apps load and leap onto the screen are reminiscent, not surprisingly, of iPad OS more than traditional macOS.
I’m open minded about this. I do know that when mainframes went from Bipolar to CMOS in the mid 1990s we went through a series of huge speed ups and then it slowed down.
I think we know where we’ve been. I’m not sure we know where we’re going - but that’s fine.
I hope the 20-30% PA speed ups continue.
Cores is an interesting question.
Fun fact: In the 1960s when mainframe went from 1-core to 2-core we got 1.1X.
Now, 50 years later we scale nicely to 190.
This is not my attempt to praise the mainframe but rather to impart what I’ve learnt from a very long timeline with hardware and software developers who are about as clever and determined as those in Apple.
The interesting thing is how you design huge multiprocessors - to minimise what we call the “MP effect” - namely how that 190th processor yields as close to as much as the first.
Inevitably, beyond a certain point you do it by ganging together multiple chips, each with many cores. (190, for example, is a 12-core PU chip with 20 of these chips working together via a sophisticated cache hierarchy, System Control chips, and communication protocol. Yes, lots of these 240 either aren’t used or are used as eg I/O processors.)
Now back to Apple Silicon: I would hope the architecture has been designed in a similar way.
I could conceive of eg Mac Pro, iMac and Mac Mini being re-architected in a similar way to get us to eg 64 or 128 cores - with good MP ratios. There is the physical space.
I could conceive of a (16”?) MacBook Pro with 2 PU chips, getting us to 32 cores - again with good MP ratios. Again, there probably is space and appetite for it.
I could imagine M2 being 16-core, or maybe just 12-core.
All of the above necessitate M2 supporting more than 16GB of memory and more I/O capability (bandwidth and ports).
What will get to be interesting is whether MacOS can drive this efficiently and effectively. (Our operating system, z/OS, has had to do much work over the decades to enable the hardware to achieve excellent MP ratios.)
(Just a few thoughts from a highly experienced / old Performance person.)
“Announcing the new Mac Pro! 64 cores, up to 512 GB of memory, and three high-speed Thunderbolt 4 ports! That’s 50% more ports than our industry leading Macbook Pro!”
« What? Only 512 Gb? That’s a third of what the previous Mac Pro did! This is in no way a pro machine and it will perform horribly! Apple is catering to the mass market! It’s dead and doomed! »
(All said in good humour, eh 
)
True. But if they take the cue from the M1, can you imagine putting 512 GB of RAM in the same package as a 64-core processor? I feel like that would be a very, very large chip.
The RAM is in the same package as the SOC, but it’s not on the same die.
For larger RAM configurations, there’s no reason that the RAM can’t be completely separate.
