Jonathan Morrison posted a video [0] comparing a 10-core Intel i9 2020 5K iMac with 64GB RAM against an iPhone 12 Mini for 10-bit H.265 HDR video exporting and the iPhone destroyed the iMac exporting the same video, to allegedly the same quality, in ~14 seconds on the iPhone vs 2 minutes on the iMac! And the phone was at ~20% battery without an external power source. Like that is some voodoo and I want to see a lot of real world data but it is pretty damn exciting.
Now whether these extreme speed ups are limited to very specific tasks (such as H.265 acceleration) or are truly general purpose remains to be seen.
If they can be general purpose with some platform specific optimisations that is still freakin' amazing and could easily be a game changer for many types of work providing there is investment into optimising the tools to best utilise Apple Silicon.
Imagine an Apple Silicon specific version of Apple's LLVM/Clang that has 5x or 10x C++ compilation speed up over Intel if there is a way to optimise to similar gains they have been able to get for H.265.
Some very interesting things come to mind and that is before we even get to the supposed battery life benefits as well. Having a laptop that runs faster than my 200+W desktop while getting 15+ hours on battery sounds insane, and perhaps it is, but this is the most excited I have been for general purpose computer performance gains in about a decade.
[0] https://www.youtube.com/watch?v=xUkDku_Qt5c
Edit:
A lot of people seem to just be picking up on my H.265 example which is fine but that was just an example for one type of work.
As this article shows the overall single-core and multi-core speeds are the real story, not just H.265 video encoding. If these numbers hold true in the real world and not just a screenshot of some benchmark numbers that is something special imho.