For extended runs with a demanding workload the M1 in the fanless model will throttle a lot according to a review.
This is hardly surprising as Anandtech measured the SoC consuming over 20w in the Mac Mini.
Way too much for a fanless thin and lite to maintain.
A Zen 2 with a fan and 8 cores will dominate it for extended heavy workloads.
You can't compare active and passive results. Either active with active, or passive with passive. There's no passively cooled Intel/AMD laptop that beats the M1 Macbook Air in performance, burst or sustained, single or multithreaded, regardless of thermal throttling which happens on passively cooled Intel/AMD laptops as well.
And yes, once you include actively cooled CPUs, then they do dominate in extended multithreaded workloads. Almost as if that heat has to go somewhere eventually
The M1 is excellent but some people are getting very carried away.
I have heard people saying you can get away with much less RAM because Unified and Integrated memory is somehow magical.
For CPU bound workloads that demand loads of fast memory, paging to the SSD will kill performance.
A lot of people misunderstand the RAM point, either saying you can get away with a lot less RAM in any scenario, or dismissing it entirely, saying there's no practical advantage. Both are equally wrong.
It's true that the unified memory is more size efficient versus shared memory, but most CPU or GPU specific workloads don't benefit much from it, the benefits only exist where there's a lot of communication between parts of the SoC (CPU, GPU, neural engine, decoders, etc), and in those cases the benefit can be significant, and yes, you can get away with a lot less RAM. If the data in memory remains busy in one SoC module (e.g. gaming), there's no advantage.
And automatic reference counting (ARC) is significantly more size efficient than garbage collection (by a factor of 3-4x), but not all macOS software use ARC, in fact the majority of non-Apple software don't. But most Apple software use that. Apple forces this on iOS, so they can get away with a lot less RAM easily there. but they don't have that level of control on macOS. This, however, is 100% the case on Intel macs as well, as it's a software feature. Everyone else could also use ARC if they wanted to, but they usually don't because there's a performance hit and putting more RAM is cheaper and more scalable than making faster CPUs. Apple decided to minimise the performance hit in their own CPUs, and save on RAM. Worked well on iOS, but doesn't extend to macOS.
So yeah, more size efficient depending on the workload, but no magic
