Coming from an i9? An M1 with 8GB of RAM would be significantly faster for most things.

Look - the reality is it's not going to be clear cut. How much faster it is will depend on the individual task you're working on. There are massive differences and not just in terms of architecture. For example the i9 tends to burst at high speed then throttle to a much lower speed. Apple silicon tends to start off sipping power and ramps up to high speed only after you've given it a sustained workload.

That makes them really hard to compare on a benchmark. Apple Silicon tends to really shine on tasks that take 15 minutes to complete (those often take hours on Intel). The Intel chip tends to shine on tasks that take just a few seconds (and a lot of the popular benchmarks test those, with deliberate breaks to let the CPU cool down).

On a Windows machine you'll have faster, more expensive RAM devoted to graphics (DDR 6) and still-very-fast (but not as fast) RAM devoted to memory (DDR 5). Apple's Unified memory is all the fast expensive type which means the graphics will be borrowing your main memory (though it's the slightly slower and less power-hungry kind sort designed for laptops).

Apple uses HBM memory, which is very fast, though exactly how fast depends on the model:

• An M3 Max will have HBM3 running at 3.2 Terabytes per second
• The fastest i9 Apple ever sold (sounds like you have that) used 2666Mhz DDR4 running at 90 Gigabytes per second
• That's 35x faster.

But an even bigger difference is in the caches. For example your i9 probably has 16MB of L3 cache (newer i9s have more, but you're not using a recent model right now). The early Apple Silicon chips had zero L3 cache. Instead they had massive amounts of L2 cache. More L2 than you have for L3. And L2 cache is a lot faster than L3. It's also obviously orders of magnitude faster than 2666Mhz DDR4 (which is what your i9 will use for a lot of operations that are in the L2 cache on a newer Mac). The M3 models still have more L2 cache than your i9 has L3, but on top of that they also have a bunch of L3 cache (again, more than you have now).

And when you're comparing DDR4 to L2 cache you'll see real world performance for common operations, such as low level memory management in the swift language, being thousands of times faster. Sometimes ten thousand times or more.

And of course there are some things that are slower. Once it all comes out in a wash though, I think you'll be really happy with whatever you get. Why not just try one?