That’s not how RAM works. Most of the ram in use on a desktop is not by the program but by data sets in use. Editing a 52 Megapixel photo uses the same memory no matter the architecture, be it x86, PPC, ARM, MIPS, or anything else. The iPP gets by with the workload it’s experiencing. Desktops have a much different workload.
I was holding out for this but sprung for a good deal on a used 2018 model a couple of weeks ago when the rumors were pointing to the iMac getting AS first. I have no regrets, my i7 should last a long time and looks much better than this, benchmarks pending.
I'm not saying RAM works differently, but the "need" for RAM is different. Why is RAM needed? It's an in-between medium between the CPU cache and slow storage. Normally, because storage is slow and cache is limited, we designed systems to park a lot of data in RAM even if it was not necessarily needed right that second, because loading it into RAM took too much time if you waited until the second you needed it. But I'm not sure that is still as applicable today as it was 20 years ago.
First, and I don't know this to be the case with M1, but efficient RAM compression can make do with less RAM if the compression/decompression algorithms are built-in to the SoC architecture. I've seen it used very efficiently on even Intel Atom chips, where the penalty for the compression was basically deminimis compared to the performance gains of being able to store more data in the same amount of RAM. Lots of servers make use of this today. Thus depending on the compressibility of the data stored in RAM, some gains can be had there.
Second, the difference with M1 is not just that it's ARM, but that they have more controllers into the SoC itself, which improves IO performance. The M1 contains inside it basically everything we used to have on an entire PC motherboard 20 years ago. Most of us have learned to avoid virtual memory like the plague because we grew up with rust spinners connected over slow IDE interfaces where page-outs would slow a system to a molasse drip. But if you have a blazing fast NAND chips directly connected via an on-chip interface controller (something I have not seen anywhere else but in Apple chips), the performance hit of using virtual memory may not be so bad. Paired with the larger amount of cache available on the SoC, it seems the only thing the RAM would be needed for is the immediate-term storage of data moving between storage and processor, but not needed to store medium-term datasets. In other words, you don't need a lot of RAM if the time it takes to load what your CPU needs right now is basically nothing.
Again, this is mostly speculation. But from what Apple showed today, and what Apple has done with A-series chips and iOS optimization in the past, I am not sure the RAM paradigm we're used to is applicable here. If the Macbook Air can load that 52 Megapixel photo you mentioned from storage into RAM as fast as the SoC can process it, then it really doesn't matter how much RAM you have.
