Raw CPU speed is one aspect in this. The other is the performance of the memory itself. Its latency (CAS latency, or CL) and clock speed are crucial to these operations where apps are copying large amounts of data (image pixel arrays) at a time. It's again the theme where "back in the olden (CCD) days" when we had, at max, 16MP images, and the time it took to do that stuff was a drop in the bucket compared to the hulking amount of readout time.

When a CPU is instructed to copy data in a large range of memory, it has to read the address range(s) is it copying from, find a region that is suitably large and unoccupied, and send the data there. Sometimes it can't contiguously copy a range and has to move around existing occupied areas to make room or it has to break up what it's copying to fit in ranges of open address space. The performance of the memory governs how fast this all happens. and is what's going on when an astro app has to copy or iterate over image data in order to do things with it.

This performance is why gamer and other high performance computing setups go with high clock rate (or even overclocked) memory that also has low latency specs. These determine how fast memory accesses can happen and the overall throughput of moving data in and out of memory. Low-power embedded-type systems are popular in astroimaging now, and the design of a given system will decide, in part, how fast your imaging app will be able to gulp down frames and do everything that it needs to do with that data. The issue is that a lot of these kinds of systems have their RAM soldered directly to the motherboard, and they might not be RAM parts that are the fastest that the CPU can support or have the lowest latencies. Sometimes you get lucky and end get something decent, but that's also going to show up in the cost of the system. I don't hold my breath for performant parts when it's a $250 (with Windows installed) deal off of Amazon. This is why I prefer NUC-based/alike systems that have SO-DIMM slots so you can swap out the memory.